I feel like I'm seeing a lot about how Gelsinger's decisions were the wrong move for Intel, but I don't see what he direction he should have taken the company instead. What is (or was when Gelsinger became CEO) the move for Intel to get back on top?
>I feel like I'm seeing a lot about how Gelsinger's decisions were the wrong move for Intel
The narrative machine for Intel is in full force given how much monies is in play, I wouldn't take anything that seriously.
>What is (or was when Gelsinger became CEO) the move for Intel to get back on top?
Whatever it takes to drive stonks up up up, obviously. If there was one thing Gelsinger could legitimately be blamed for, it is that he wiped out Intel's market valuation. Nothing else matters, because his "interim" replacements are both beancounters whose literal job is to drive stonks up up up.
because his "interim" replacements are both beancounters whose literal job is to drive stonks up up up.
That seems like the opposite of what I've heard before and what NVDA have been doing. Is NVDA's stock soaring because the bean counters stifled any innovation before it could happen in order to drive the stonks up?
Bean counters can bring stock value up temporarily. The trick is to get out before shit hits the fan. For a comparison, see Boeing. Where shit is hitting the (turbo) fan right now as well.
The difference between Nvidia and Intel is that Nvidia still has its founder at its head. A founder interested in seeing his company instead of just stonks go up up up.
I have no doubt Nvidia will start losing its way when Huang retires, like almost all other founder-less companies such as Intel.
I don't disagree with you that Nvidia has good leadership, and that Intel has bad leadership, but I take a bit of issue with this pedestal that founders are placed on.
Being a founder doesn't magically make you the best person for the job of CEO of a company. Huang happens to be a very good leader, but a lot of founders become a drag on their companies later on.
It's just an anecdote, I worked for one place where the company truly outgrew the founder's abilities. He was just some guy with an art degree who was at the right place at the right time with his dorm room business idea. I would argue that his lack of ability and experience squandered a lot of the later stage growth opportunities that could have made the company into a billion dollar public corporation.
I think about a company like WeWork, that's a company that would not exist today if the founder was not kicked out. Some companies like McDonald's would never have expanded to be a household name if their founders were still in charge.
Their point wasn't that founders are somehow wiser. The assertion was merely that founders care about the company.
Their decisions may be no more prescient than anyone else's, but their decisions are reliably in different directions than anyone else's.
Their goal is different, the direction they always aim is different, and that consistent aim has an effect over time regardless what mistakes and misfortunes happen along the way.
If you have a specific and clear goal on the horizon and are always aiming for that even though you have to deal with obstacles closer in front of you, over time you still get somewhere.
If your goal is just chasing whatever random butterflies float by, you get nowhere.
"Jensen Huang, the 61-year-old co-founder and CEO of Nvidia" ... There's a hint in there somewhere.
How do you get a CEO to believe innovation is the best thing for them and the company? Even if you get them to believe that how do they get their company to innovate?
Innovation is hard work. It involves failures. Just buy back your shares and make money. Then someone else comes and eats your lunch. That is how the world works.
nvidia's success has nothing to do with bean counters - if nvidia even has them.
bean counters are for industries like beans - where the market is stable and the only way to liberate shareholder value is to catch the stray beans. they are exactly the wrong answer when "disruptive innovation" is afoot.
Well said. Everybody makes mistakes. How many mistakes do you think you have to make in order to build something like Tesla, or SpaceX?
Larabee was a mess....anybody who as actually been involved with designing a chip which breaks new ground (and is not just repackaging of somebody else's IP) knows that designing and taping out ANY AND EVERY chip is a big mess. It is not easy to tell heaven from hell on this one--and arguably, Intel's management at the time made the wrong call. Pretty much every decision they were making at the time was the wrong decision, which is why Intel is in the state it is in.
When I worked at Intel, Robert Noyce had unfortunately died, but we still had Andy Grove on the business end and Gordon Moore on the technical end. All three of those people are not just good, not even merely great. They were the kind of transcendent geniuses who rewrite the world, and which come along once every 2 or 3 hundred years. Impossible to replace.
AFAIAK, If Gelsinger couldn't fix Intel, it was irreparably broken by the time he got there. We should celebrate him for having the courage to try to fix what 4 previous CEOs couldn't fix.
I understand that it's painful when one of your suppliers has financial difficulty and has to exit a business you depend on. I feel ya. I hope that nobody will be writing blog posts about the OP's decision to partner with Intel in the first place, pointing to it as one of the fatal mistakes made by the company and their leaders.
> but I don't see what he direction he should have taken the company instead.
Software and design support. As pointed out in the article.
Intel is genetically incapable of valuing anything other than "The Fab(tm)". VLSI designer ... second class citizen. Software ... second class citizen. etc.
The ARC graphics card series is an exemplar of the problem. It was always going to be a market disaster. Fine, accept it. However, Intel should have pushed a couple of cards for free to every single HPC/AI grad department in the US. And perhaps even tossed in a $50K grant to boot. That's, what, $50K*100=$5 million max and a thousand graphics cards? Sure, the top 10-20 departments are well-funded and still going to use Nvidia--the rest will chew on your graphics cards. In return you get a legion of grad students developing software for your hardware. And you will quickly get feedback as to what you need to fix for v2.
However, you have to value software for that idea to penetrate. And Intel just ... doesn't.
> In return you get a legion of grad students developing software for your hardware.
That would be worthless. It's so weird that people think academia is contributing in any meaningful way to the top lines or bottom lines here. You're free to look at the commit history of PyTorch/TF/Triton/HF/llama/whatever and judge for yourself.
I’ve worked at several companies where there have been decent HPC/GPU codebases and at all of them the staff working on that have largely come out of academia in one way or another.
> However, Intel should have pushed a couple of cards for free to every single HPC/AI grad department in the US. And perhaps even tossed in a $50K grant to boot. That's, what, $50K*100=$5 million max and a thousand graphics cards
Nvidia used to do this in the late 2000s and early 2010s.
It did wonders to drive CUDA uptake versus Vulkan, and Nvidia would sponsor plenty of researchers and students to leverage CUDA and Nvidia GPUs in Parallel Programming classes.
> you have to value software for that idea to penetrate
I'm not sure it's software per say and more so Developer Experience.
If a vendor fails to truly interact with their implementers, they lose critical feedback.
Intel used to have a very active Technical Marketing Engineer org (the old word for DevEx), but I'm not sure how active or relevant it is anymore.
> not software per [se] and more so Developer Experience
This aligns with my experiences trying to use Intel software.
The software was often technically very good, but the user experience trying to use it was horrible, and god help you if you got an error.
Larrabee all over, really.
I suppose linking to [1] and [2] is probably germane for this discussion.
I swear there was another blogpost explaining the tradeoffs made in the Larrabee/Xeon Phi development process in exacting detail, and confirming that it ran gaming workloads pretty well before being shelved, but I can't find it again.
Thanks for the links! I absolutely should follow Matt Pharr and Tom Forsyth's blog tbh - I've been out of the HPC world for sometime now but it's getting relevant again.
Vulkan didn’t even exist when I got a free card from NVidia.
For cross platform programming on GPUs people in academia/industry just gave up because performance was worse, the tooling was terrible and library support was non existent. OpenCL, OpenACC, OpenMP GPU offloading have all been disappointing.
There's just two things Intel should have been focusing on, and that's getting their fabs back to being competitive with TSMC, and getting their chip designs back to being the best IP in the world.
There is no reason why TSMC should be leading Intel on fabrication by so much. Intel has had a crazy amount of time to restructure and refocus when we consider how long their x86 intellectual property lead kept them afloat with their subpar fabrication and delayed manufacturing process updates. They spent many years having faster chips than anyone else despite having subpar fabrication. But now, they can't really escape the fact that ARM is exploding and AMD has the most flexible x86 platform where as each year goes by the number of use cases where Intel is the best choice diminishes.
If you play games, the x3D chips are basically the only sensible choice. If you are looking for servers, AMD is consistently 10% cheaper in all the cloud platforms. If you want a mobile system, anything involving ARM and/or TSMC will get you better battery life.
The fact that fabless players in the market like AMD and Apple can outpace Intel in so many product lines is, frankly, embarrassing. While I think that other processor design shops selling off fabrication divisions made sense, I think that Intel as a company basically makes no sense if they can't get their fabrication shit together. Because if Intel can't fabricate bleeding edge chips what's the point of using their IP?
I was listening to a discussion about this today, latest Accidental Tech Podcast.
The biggest problem Intel has at the moment seems to be they can’t split up and they can’t work together.
If you keep both the fab and the design, you have problems. The thing that makes Intel all the money is their CPUs, and because of the state of the fab many of them are made at TSMC.
But TSMC isn’t too keen on giving a good deal to someone who wants usurp them. So Intel is paying more per chip than they would if they just spun off the fab. If they did that, they could have higher margins/lower prices.
But if you spin off the fab, can they actually get enough business to be able to survive? They’re not on the leading process. They need a massive investment to catch up. Despite their geopolitical importance it seems like they would basically need a real continuing subsidy just to exist, which is effectively being provided by the CPU design side of the business right now right?
It’s a real damned if you do damned if you don’t kind of thing.
I don’t know what the right thing to do is. Splitting them up kind of seems like the better idea but I don’t know if Intel’s ego could take that. They were the fab company. Their advantage let them survive multiple large mistakes that could’ve killed other companies.
Now they’ve all piled up.
PS: the CHIPS act money says it goes to Intel. It sounds like if they try to split they may lose it, so would first have to convince the government to let the fab part keep it
I much agree with your assessment of the challenge, but I think Intel’s only chance is to capitalize on vertical integration of chip, design, fab. TSMC works closely with big customer (cough Apple), but there are still boundaries there.
Intel needs to step back, enjoy the runway the CHIPs act got them, and really think hard about the top to bottom design and fab of an absolutely killer 2027 / 2028 chip would look like. They aren’t as good at design as Apple, and they aren’t as good at fab as TSMC, but they are within striking distance of both and they have that vertical integration opportunity.
Wouldn’t that basically require becoming better at fabing than TSMC again?
If they can catch up, but not exceeded, then they can’t make a better product than they do right now. I guess it should cut their costs some amount (don’t know how much) which would certainly be a big plus.
But without taking the lead again… would that be enough? I suppose if they caught up they could also fab other people‘s chips as Pat wanted to. If they’re tied for best (or at least close) they could probably pick up accounts that TSMC doesn’t have space for or are perhaps too small to help fill out their capacity.
While I was listening to the podcast I was thinking about how it almost seems like sort of a historical accident that they’re both design and fabrication. Does anyone else do that at this point? I know it was common early on but maybe it’s just not that sustainable by now.
I don’t know. I guess I see splitting the company is more likely to succeed, simply because the design side works and would lose the anchor that it’s tied to.
Successfully pulling off a fix while keeping the company intact may be a better outcome if possible though.
But I don’t think they have to be better than TSMC, or even quite as good. Intel’s strength is the design + fab integration. If they are 90% as good as fab as TSMC and 95% as good as Apple or Nvidia at design… maybe that’s sufficient to be competitive.
Do you honestly believe that Intel can outperform every single company making chips? They're the last vertical chip maker. Samsung is still trying, but they're not beating the one-two punch of Apple design, TSMC production that's for sure.
* Signal Intel is in it for the long term, specifically kill the dividend.
* Be more customer centric/rebuild customer trust, specifically publicly acknowledge the 10nm slip.
Though it seems customer trust in Intel has eroded to the point where only actions will change customers' perception of Intel — I'm not sure what Intel could have communicated to Tofino/Oxide that they hadn't already.
Larrabee might be a disaster, but sometimes it takes a while to get things going.
What other market was Intel chasing that would have yielded a large new market? Oh right, Mobile CPUs. I mean besides that? An actually well-done Linux distro that helped showcase Intel's hardware without waiting for Windows to release 9x+1. Ok, besides that? Good laptop hardware. Ok Ok Ok.
But it still stands, even if 1.0 and 2.0 are a disaster, iterate and work on it.
The big problem is that Intel is a quarterly earnings focused company. Well, and even if Gelsinger is good, the middle management is pretty clearly totally dead sea. If middle management is bad, they can't attract talent, and can't retain talent, and can't support talent. Without talent, no vision or execution.
Intel doesn't value intellectual labor. It's the bottom line with the company. An arrogant technically competent leader is still better than the alternatives.
That said I have no direct or even indirect interaction with Gelsinger. All my strong opinions are derived from 30 years of being in software and PCs.
Anybody who has actually worked on a ground-breaking CPU--a really new design, with a radically new architecture--and not just recycled/repackaged third-party IP--knows that it is always a total disaster. It is always at risk of being shut down by a nervous board.
Of course, until the day that it is not. Intel's management arguably made the wrong decision when they shut down Larrabee. Lets face it, at the time they were not making very many good decisions, which is why Intel is in the mess its in now.
It’s the Kodak phenomenon. A software company like VMware would have had a hard time doing the capital investment necessary to be AWS, while simultaneously cannibalizing your real business.
That’s why VMware tried to pivot to sticky stuff like EUC and dev platforms.
IMO I just don't think it would have been anything closet to being as difficult as a Kodak phenomenon for VMWare. AWS started quite small with just one service and it was funded by an extremely low profit margin e-commerce business.
If VMWare had the product vision at the right time they could have been AWS.
No way. Bezos had the ability to do the Jedi mind trick to convince wall st that billions and billions of capital investment was ok.
It’s also easier to sell massive datacenters when you’re good at building distribution centers and have low margins. VMware was a software company with that margin expectation. They tried a partner model that didn’t really work. The only other good option was something like IBM or Sun back in the day.
I lived it - my first job was a software company that primarily served call centers in a specific vertical. The customers wanted out of the phone business and online… so they offered the full package. Fast forward a couple of years and the dot-com bubble bursts, and the investment dollars and lines of credit dry up. The founder mortgaged his house to keep the place open 60 days to complete an acquisition, which ultimately went through and everyone did ok. Two years earlier, we’d all be millionaires. 3 weeks later, holders of worthless paper for a bankrupt company. Timing matters more than anything.
Amazon has been public since 1997. There's no Jedi Mind Trick involved with convincing investors to back AWS. The public has had their real financial results available to the public for a whole decade before AWS existed. They have been under pressure from shareholders in the same way that VMWare is.
You're proposing the idea that VMWare had no idea how to build data centers as a company that builds data center abstraction virtualization software almost exclusively? VMware is a company that partnered with data center hardware companies to make VMWare-branded data center appliances. And they just...don't know how to set up a data center?
If Amazon's expertise in e-commerce was a direct line of sight totally obvious foray into cloud computing we'd see Walmart, Target, and eBay operating AWS their own competitors.
Amazon leadership saw an opportunity and understood the 20 year vision of how enterprise computing would work in the future.
VMWare's managers are solely responsible for not seeing that future and failing to empower their teams to create new products in new categories. That is particularly embarrassing since they are subject matter experts on abstracting away hardware in the data center.
Docker was a factor but there's no particular reason why VMware had to take the lead they had in on-prem container platforms and throw it on the floor for OpenShift to pick up.
This isn't a space I pretend to understand in detail, but Cantrill's objections to Larrabee here seem confusing. With hindsight, Intel abandoning Larrabee has been a disaster and they should have sustained the effort to develop a great GPGPU chip. Without taking any position at all on Gelsinger's execution his strategic vision seems to have be supported by what happened in practice after his attempt.
If Intel had been practising making discrete GPU from 2009 to 2024 instead of giving up then it does seem quite reasonable that Nvidia would be a quarter of its size. One of the obvious mis-steps Intel made over the last 20 years was failing to keep trying Larrabee-style chips until they figured out how to make them work. Ok version 1 might be a disaster and version 2 might be bad. But we're looking back over almost 20 years! There was a lot of room to figure out how to be good and tap in to the huge new market that was opening up.
Intel has a habit of coming up with great ideas, then deciding that they aren't immediately successful or immediately 60+% gross margins, so it clearly isn't worth following up. Intel's R&D department is a grave yard of great products murdered at the alter of Gross Margin.
Larrabee is one such product. It wasn't immediately great, and didn't immediately promise great returns, and didn't quite have a market yet, so it was killed. I can't imagine where Intel would be in the AI world if they had the foresight to stick with Larrabee.
Yep. Intel has no patience. They have high margin businesses throwing off lots of cash, so the finance-driven culture says “why waste money on risky new things that can’t possibly equal that level of profitability in the short term?”
Somehow they forgot that all of their highly profitably businesses were risky bets at one time.
Right. Google has the same problem. Any time they create a new product, it’s a coin flip on whether or not it’ll still be around in a year or two. I’ve steered multiple companies away from Google’s offerings because I just don’t think you should trust that any of their newer products will still be around in 5-10 years.
Good on them for sticking with their promises for firebase. But it’s a pity about stadia. Don’t bet your business on a flip of the Google coin.
In an infinite universe, anything can happen, and so "it might not work" is is always true for everything. No matter what you do, it might not work, and some of the most important and valuable things ever aren't attainable without a lot of investment before you can start to reap any return.
Not all ideas are good ideas, so of course there are reasons not to do some things, or to abandon something you started, but merely "what if it never works" is not one.
Asking that is about equal to asking what if a piano falls on you tomorrow. Sure it could happen, but so what? What does that change? If you don't walk down the street for fear of pianos, you don't get to the grocery store or your job and then you die from starving instead of from a piano.
To be clear, I think that they should have kept iterating on Larrabee (though I would have liked to see it with an open software architecture). With respect to Larrabee, my shock was that Gelsinger had such confidence that iterating on Larrabee would have resulted in an NVIDIA of one-quarter its current size (or one-quarter the size when he recorded the interview in 2019). That's the bit that's bonkers to me: yes, they should have kept iterating (though some acknowledgement from Gelsinger that the first iteration was plagued with problems, not all of them technical[0]) -- but success was emphatically not assured just because it was Intel.
>their fab advantage might have let them really go after Nvidia.
I think Cantrill spelled it out in the article, and I think people are overlooking (and for some reason really continue to overlook, or just have no idea) that nvidia's advantage is Software.
I remember Larabee as a teenager - I just graduated high school, the forums were talking about it, and I had finally saved enough money to buy an 8800GTX. The 8800GTX was also coincidentally, the first consumer GPU to support CUDA. Almost 2 decades later and it feels like nvidia has an insurmountable moat in AI due to CUDA. This wasn't an accident, nvidia spent a ton of time, with developers, getting it right. FTA:
>that NVIDIA had an indisputable lead, even in 2009; that for Intel to dominate NVIDIA it would have required conjuring software expertise and focus with which Intel has famously struggled
Ignoring this piece is huge in claming that nvidia would be a quarter of the size. Can you even imagine Intel investing in a non-x86 developer platform for 15 years while they were in the lead?
Really? It doesn't seem to me like much Intel-written software has been key to x86's success. The things I can remember from Intel are compilers, performance analysis tools, and threading and math libraries. But none of them were must-haves, or differentiated x86 chips from competing CPU architectures. Most of the software that made x86 successful came from third parties.
And more recently, the reviews I read of the Archmage graphics drivers didn't make it sound like Intel had understood the importance of the software that was essential to a pretty big hardware release.
Your examples are simply too recent. They've been doing it since the 70s: https://en.wikipedia.org/wiki/Intel_system_development_kit - that they forgot what made them the success they are is unfortunate, but also wholly self-inflicted.
What? This is really wrong. The name of the game these days is optimizing memory and FLOPs usage on these GPUs. For example, Hopper (H100) cards introduced the WGMMA (Warp-group Matrix-Multiply Accumulate) instruction, and anyone who does LLM training jumped to use it ASAP because without it you can't fully utilize the FLOPs. Anyone training AIs at the cutting age cares a lot about CUDA details. https://hazyresearch.stanford.edu/blog/2024-05-12-tk
That’s not what OP meant though. That’s caring about _hardware_ capability. You could do that yourself for different hardware. And (OP’s point) any alternative hardware provider could do that for you. It just happens to be really, really hard
I think "AI is all about Torch/TF, not details below that" directly contradicts the fact that ML people very much care about the details, to squeeze performance out of the hardware.
This could be 100% true, and the best ML engineers in the world could be completely oblivious to CUDA and it wouldn't change why nvidia is leading right now.
Anyone who thinks it's "trivial" to use AMD cards should help geohot and his tinygrad boxes. Despite emailing the CEO several times, it was a clear uphill battle to get something working, and even now the sales page lists driver quality as "mediocre".
If you open a GitHub issue or a forum post about your weird CUDA application that isn't working right, you will have an engineer help you. This is in nvidia's DNA and they did this before AI started printing money. On the flipside, you can email Lisa Su today and still not have problems fixed in AMDs drivers. I'm sure AMD is trying, this isn't to say they aren't, but having decent software culture at these large companies is hard.
If AMD can't do it, I feel even less confident about Intel. If the CUDA moat truly didn't matter, we would see AMD cards in demand. However it doesn't and it isn't for lack of trying.
I literally sat at a table where the argument was made: The Intel accelerator is faster. It provides more FLOPs. But you can’t get actual users to use it outside of our LINPack runs, since it just doesn’t provide bindings to all the software actually intended to run. Granted that was in the HPC space (ie FP64 performance), but that is literally the argument.
That doesn’t mean, that ML people don’t care about some hardware specifics btw. Of course they do! Because CUDA is very, very good at exploiting the specific features of the newer cards (I’ve forgotten, what exactly this was though). But take away the simple integration into their favorite AI-SDK and it would be completely irrelevant.
The point is just that, if Intel were to build strictly better hardware, but not provide the interface to torch, then they would still loose to Nvidia. Yes, yes, this is not completely true. Given enough advantage someone else would build the IntelCUDA. But this is literally the bet that those academics make. And it’s so risky, that only academics are doing that
Because Torch only exists on CUDA-platforms. At least in any useful form.
This was, if anything, the main theme at Supercomputing: “there is not even a point in talking about Nvidia’s new benchmarks. You all are going to buy it anyways” coupled with very few “we are betting on buying cheap hardware (Intel/AMD GPUs) and hope that we can build the relevant parts of CUDA ourselves” and the latter is pretty much a desperation move of labs/sites that simply cannot get NVIDIA GPUs (either price or availability).
And yes that is probably the seed of the end of Nvidia’s dominance. But it will take 20 years and multiple fuck ups. Just as it did with Intel
Maybe not now (I’m not in a position to know), but wasn’t it a HUGE factor in them becoming the player they are in non-graphics stuff? Because they had great APIs/libraries/documentation to play in that space?
You need to remember that was during the financial crisis, and the first chips were going to be laughably bad and uncompetitive. From the quote in that article:
> Justin Rattner (Intel Senior Fellow) demonstrated Larrabee hitting one teraflop, which is great but you could walk across the street and buy an ATI graphics board for a few hundred dollars that would do five teraflops.
Intel had no chance for v0 to do anything but lose money. They were way off track, and spent most of the time trying to get to a software rasterizer plus hardware that could work for some games.
They did pivot Larrabee into the various Knight chips (Knights Corner, Landing and whatever else) since the "it's just x86" was a pretty compelling story in the 2009 era. But ultimately, HPC wasn't that great a market in the early 2010s for accelerators yet and NVIDIA ended up totally eating their lunch over time.
I think Intel could have stuck with the HPC market and gotten lucky once ML exploded, but realistically until about 5 years ago there wasn't that much ML spending yet either. 10 years of last place discrete GPU work would have been difficult to invest in.
tl;dr: the business case for Larrabee hinged on having a successful gaming business and it wasn't even close.
Was Pat perfect, no. But Pat acknowledged Intel’s problems - something Otellini, Krzanich, and Swan never did. These CEOs, all non-technical, focused on dividends, buybacks, and next-quarter results while Intel fell behind in advanced nodes and innovation.
Gelsinger inherited a disaster: 10nm delays, TSMC pulling ahead, and no GPU strategy. He had the courage to cut buybacks and slashed dividends. He poured billions into fabs in Arizona, Ohio, Germany, and Ireland. He delivered Intel 18A, powered on first silicon, released PDK 1.0 for Microsoft, and secured Microsoft and Amazon as customers. There were even rumors Apple might join.
Contrast this with Nadella at Microsoft back in 2014. He didn’t reboot the company by tearing everything down. Instead, he built on Ballmer-era wins like Azure, Office 365 while shifting Microsoft’s focus to the cloud. Gelsinger had to start from scratch in many ways, tackling years of neglect while facing harsher challenges.
Yes, Intel’s stock dropped $150 billion, but Gelsinger was upfront - it wouldn’t turn around before 2025. He was trained by Gordon Moore and Andy Grove, and while some saw him as arrogant, that confidence came from decades of technical leadership.
The real issue? The board. Full of people like Boeing execs. They don’t get engineering. Trusting them to fix Intel is like hoping a plane door won’t pop open mid-flight. They’re the ones who should be replaced.
I was inside since Otellini (left last year). Otellini at least 'stayed the course' reasonably well. I never felt like the company was going to GROW, but at least not die.
BK is who did his best to sink the company, multiple mis-guided layoffs, bad top level hirings, and stupid direction changes every time he read a new news article. Don't even mention his absurd AMA on reddit where he couldn't stop using the ellipse and laughing like a 14 year old girl.
Bob Swan wasn't the guy who could right the ship. He was another Otellini type guy, someone you bring in to not mess stuff up. Pat was a great pickup for Intel, and his return was my first time hoping that Intel could actually survive and grow.
Unfortunately, he came by too late. Foundry STILL hadn't really made progress, and the rot in the rest of the company had set in too far. The brain drain from BK was still being felt everywhere.
Frankly, I still feel like Pat was the right guy, 10 years too late. He joined to right the ship, but by the time he came aboard, the ship had already hit the iceberg, and the bow was 30 feet in the air. And now he's being blamed for the ship sinking.
This is a great answer. Pat is not a wartime CEO. He could have been fantastic for Intel if the company was somewhat healthy and the competition was not so fierce.
Honestly, I think the thing that got Pat fired was agreeing to suspend stock buybacks for five years as part of the CHIPS act money deal. The Board needs to be able to prop up their investments.
But if buybacks are illegal and dividends are highly discouraged by taxation, what's the purpose of the stocks at all? How are stockholders supposed to get paid back if there is no flow of money from the companies? And no, "sell stocks for higher price to a greater fool^W^W^W another investor" cannot work on global level because if the company doesn't distribute it's profits to stockholders, stock price is purely speculative so it essentially becomes a Ponzi even if the is real value production getting done, because the value production doesn't find it's way back to the investors.
This is only true today because buybacks aren't taxed
Buybacks are ultimately a way of saying "we don't have a better way to spend this money." Consider that, in a world without buybacks, execs have two choices. One is to pay dividends, and eat the tax implications. The other is to find productive ways to spend the money to increase the company's earnings. The taxation of dividends strongly motivates companies to innovate. Buybacks weaken that motivator significantly, because there's no tax implication. It's financial engineering and not a productive use of money.
Buybacks are also a pretty neat way for insiders to enrich themselves at the expense of shareholders.
Example:
Insiders schedule a sell of their shares to occur right after announcing a buyback plan.
Worth noting also that dividends aren't uniquely taxed. Worst case, they're taxed the same as income, but they can also be capital gains which are taxed at a lower rate.
So any talk of incentives should include a justification of passive investment being more valuable than work for income, if someone is asking for favorable tax treatment.
Wasn't aware of that. Yes, I've no clue what I'm talking about. But if buyback gains are also taxed as income when cashing out, what is the advantage of buybacks then?
You can defer selling capital gains for decades in a buy-and-hold portfolio, while your portfolio continues to grow. Dividends are always taxed in the current tax year. One benefit here for investors is that you can defer taking capital gains until years when you are in a lower tax bracket (e.g. retired), which you cannot do with dividends. (The US capital gains tax rate is progressive, if with fewer buckets than income. There's a 0% bucket, a 15% bucket, an 18.8% bucket, and 23.8% bucket.)
My point was that GP's assertion that "Worth noting also that dividends aren't uniquely taxed" is wrong - they are uniquely and specifically taxed, the rate just happens to currently be the same as capital gains rates.
What changed in the 80’s is that a safe harbor was created. Before that companies were at risk of being charged with market manipulation - there were no clear rules on what was market manipulation and what wasn’t.
The safe harbor rule 10b-18 simply laid out requirements for buybacks fell into a “safe harbor”, presumed to not be market manipulation.
But companies could still do them before that, but it was more risky.
I've watched the talk in 2022 between Linus and Pat. It's a common talk as a company PR but I think Pat showed me that he is a real engineer to drive this huge and years-old company, not the Wall Street managers.
It's not 20 years. Just 12 years ago Intel was launching Ivy Bridge on 22nm and was absolutely on top of the world.
It's true they've completely fallen off the pace. But people tend to forget how rapidly this happened. Even as late as the semi-aborted 2018 launch of Cannon Lake it seemed like it was just a routine burp they'd correct with a process respin. Then TSMC quietly reached parity with 7nm, shipped 5nm which was a better process, and by 2021 Apple had jumped ship and Intel was falling behind even AMD.
The disaster happened fast. Boards of Directors aren't that agile.
Mobile phones were picking up a lot of steam by the mid 2000s, and it doesn’t seem like Intel bothered to even investigate developing more power efficient chips.
Seems like the leaders just lost the stomach for taking risks, a long time ago. No forays into mobile or GPUs, at least not in the billions of dollar and many years scale that was needed. No stomach to pay the competitive salaries necessary to compete with Apple, Microsoft, Alphabet, Meta, Amazon, Netflix, etc for talent.
Atom has always been a laptop chip. They tried to shoeshorn it to handhelds but it sucked for obvious reasons.
Think Apple's chips started in iPhone, then iPads and finally very recently ramped up to Macbooks. Even Snapdragon has only very recently released a laptop worthy chip because of the design they've acquihired from Nuvia.
"didn't beat" puts it mildly. Every attempt Intel made at entering the smartphone business was doomed because they were years behind ARM. Paul Thurrott confirmed this with HP when discussing the Elite x3 smartphone:
They did as recently as 2016 and then gave up on it: "Intel could be on the verge of exiting the market for smartphones and standalone tablets, wasting billions of dollars it spent trying to expand in those markets. The company is immediately canceling Atom chips, code-named Sofia and Broxton, for mobile devices, an Intel spokeswoman confirmed."
(https://www.pcworld.com/article/414673/intel-is-on-the-verge...)
Yeah, Otellini famously turned down Steve Jobs' request to make the chip for the first iPhone, thinking the market wouldn't be big enough. When he got pie in his face, he tried to correct course. By the time he needed to retire, the board wanted to give up on mobile, thinking they would never catch up, and double down on data center.
AFAICT, this was a self-serving bit of reverse myth-making from Otellini. If there really had been a single binary decision Intel got wrong—saying no to Jobs when they might have said yes—then their collapse looks like bad luck: Nobody bats 1.000.
But the way Apple insiders tells this story, there was no way Intel was even being considered in the (short!) window when the original iPhone was being built. Intel was in the middle of selling Xscale, and even that design was too power-hungry.
Intel missing mobile was a long history of poor strategic and tactical choices, not one bad call.
Jobs had two iPhone teams working in secret against each other, and was setting up things on the side. He likely approached Otellini before either team was far along.
They did more than investigate. Nokia, at that time still market leader in mobile phones, wasted a lot of time and effort because management wanted them to move to Intel. Nokia engineers did not believe that Intel would ever reach the required power efficiency. Whether it was self-fullfilling prophecy or just technically impossible is anyone's guess. (No, Nokia did not fail because of Intel, but that miss certainly made the disaster more complete.)
Intel connection was not the sole reason for Nokia's demise in phones, but it contributed on the failure of their effort to recover from the tailspin. Symbian their old mobile platform was clearly due to be replaced and they had a pretty viable in-house Linux platform, Maemo, that already shipped with N900 in 2009. Instead of iterating on that, they decided to "join forces" with Intel and merged Maemo with Intel MobLin to create MeeGo. They wasted at least a year on that and not with a lot to show for it as the Intel chips they planned for never materialized.
Obviously it was going to be very difficult to compete as a third platform with the behemoths iOS and Android become during those years. At least the MeeGo and Windows Phone cards were not the winning ones.
That's the software part of the story, which became fully public in form of MeeGo.
But there was also a hardware story how Nokia would start Intel silicon. I don't think anything of that has ever been publicly annouced before it failed. Wasting a year seems to be massive underestimate. I believe it must have been much longer. After Nokia started to fail Intel hired former Nokia engineers. I have no reliable insights what they did there, but I believe at least in the beginning they still worked with phone hardware on low-level software.
The 2017 "launch" of Cannonlake wasn't something that anyone inside or outside of Intel could have reasonably considered to be a "routine burp". It was a desperate move to avoid shareholder lawsuits and possible criminal prosecution. Intel had to ship something under the 10nm label before the end of the year, because they had made far too many (false) promises that 10nm would be working Soon. Cannonlake was a mostly-broken chip because their 10nm process did not work, and Intel never even tried to make significant revenue from it or ramp it to volume production (though they kept promising for months that they were going to ramp). And it was still two years late.
Meanwhile, Intel's chip designers kept targeting an unusable process, and wasted years that they should have been iterating on designs for the fab process that actually worked. Skylake shipped in 2015. They didn't deliver a new CPU microarchitecture on 14nm until 5.5 years later, a year and a half after they shipped that same microarchitecture in a mobile-only form when their 10nm finally started to be somewhat usable (but not fast enough for desktop).
What were the chip designers doing for all those years? In 2015, Intel knew that 14nm had been harder to bring up than any previous fab process, and they knew that 10nm was proving even harder, but they refused to try making an updated CPU design for 14nm. How could the management not have realized that spending multiple consecutive years not shipping new designs would cause long-term damage to their capability to iterate on CPU designs? Not participating in the feedback loop of actually shipping left Intel with an oversized P-core design and an E-core design that wasn't well-matched to it, making Alder Lake awkward and slapdash when they finally got 10nm working well enough for desktop CPUs.
Sure ... THIS is what I don't get. Non-technical (ie. CEO + board before Gelsinger) people are responsible for a technical disaster. They did not, of course, stop creating new technical disasters. Which ended in complete panic and Pat Gelsinger on top.
They complain about arrogance, but even if you accept that, it was arrogance BEFORE Gelsinger, with Intel under the control of MBAs that they're talking about.
And can I just say, I've seen some seriously arrogant assholes in the tech departments I've worked ... but for absolute incredible arrogance, you need MBAs.
There was always a deep cultural problem in Intel - at least since otellini when I was there. But it's not just the management that was arrogant - engineers too were a bit arrogant because they were taught that Intel was the best. The cultural problem was Intel assumed that they did not need to look outside the company about how the world and tech landscape was changing, and really they assumed they could always depend on the semiconductor process advantage to cover design inefficiencies. So the whole company was living in the past. Plus they did not hire the best - either in terms of thought leaders or in terms of senior people who were really really innovative. Intel had or has this culture of hiring lots of recent college graduates who would push new designs that were iterative but they were not of the level of new patents, or truly an outcome of research. Whereas amd was being successful with a much smaller number of employees, because of new patents and hired mostly senior people.
Once Intel lost its research focus it became an extractive company extracting the riches that were already there, instead of creating true innovation. Case in point - Intel stopped doing it's research day long time ago.
Design pipelines are deep and Intel at the time famously had very node-specific designs without industry-standard PDKs. The moment engineers were told to switch a design to 14nm, it basically reset the 5 year design-to-product pipeline. Management failed because they did not hedge the risk by starting a parallel 14nm design effort at first sign of 10nm troubles. They likely were engaged in magical thinking or some variation of the "Are YOU going to tell him?" Silicon Valley scene. It does not help that information like that is considered actionable insider trading information. I bet a lot of people working on 10nm designs first heard the news about the delays from the quarterly investor calls.
> Design pipelines are deep and Intel at the time famously had very node-specific designs without industry-standard PDKs. The moment engineers were told to switch a design to 14nm, it basically reset the 5 year design-to-product pipeline.
Right. It was well-known publicly that Intel was running their business in a way that maximized the damage any fab troubles would have on their product roadmap. It was obvious a decade ago that Intel needed more flexibility to bring their CPU designs to other fab processes. It took them too long to start working on Rocket Lake, and too long to deliver it. But they have at least made some progress on the problem, since they've been selling x86 CPU cores made at TSMC for the past year.
(On a related note: Buying Altera and forcing them to port their entire roadmap over to a broken 10nm process was made even more stupid by the fact that Intel didn't have a usable PDK that outsiders and acquisitions could work with.)
The true embarrassment was when SMIC (read: China) reached 7nm and thereby surpassed Intel last year (or was that 2022?).
Intel then proceeded to waste CHIPS monies and other aid on five digit layoffs and now ousted the one CEO who ostensibly at least had the right idea.
At this point I want to see Intel fail (and Boeing too), American Exceptionalism(tm) absolutely needs to have its longass Pinocchio nose broken in half before we have any hope of rebooting ourselves.
The "even" makes the tone of your comment feel a tiny bit disrespectful towards AMD. By 2021, it was clear to me that AMD had their gloves off and were winning. Zen 3 was released in 2020 - the third generation of nearly flawless execution by AMD that Intel failed to respond to - outside of cutting the prices on some CPUs. For a while, Intel held onto the "fastest single-core speeds". Back in 2017, my first thought after being blown away by the performance of a first-gen Zen PC build was "I should buy shares in AMD" - AMD clearly had a superior product with an even better value proposition.
I think the point is that Intel had such a lead in the Bulldozer era that for AMD to overtake them was a tremendous failure.
I would not say that the first gen of Zen is was a clear winner over Skylake. It took a couple iterations before AMD clearly took the lead. AMD was simply so far behind that several large generational improvements were needed to do better than Intel.
> I would not say that the first gen of Zen is was a clear winner over Skylake.
In 2017, I would not have said that either for Zen 1 without qualification[1]. Zen 3 on the other hand, was a winner.
That said, 1st gen Zen had better bang-for-buck than Intel, for multicore workloads - in my case, I had built a workstation and thr equivalent intel build would have cost much more, expensive Ryzen motherboards notwithstanding.
1. In my comparison as I buyer, I didn't compare Intel and AMD processors by core count, but by what I'd get with my budget. The AMD build I eent with was better than an intel build for the same amount of money.
As in many large projects, even more so with a large company,
the point is not to react when the problems are happening,
it is to preempt these problems, foresee them and prevent their happening.
So indeed, by 2018, even though Intel has not yet fallen, it's actually already late.
The roots of the problems seems to be earlier, and that's where the CEO, and the Board, should have reacted.
The ETFs actually do own the underlying securities and vote the shares. They typically have a mechanism for ETF owners to vote internally on how to vote externally, but the vast majority of ETF owners ignore it so whatever the ETF recommends (which is often from a third party) almost always is what happens.
> So when a company's ownership reaches 50%+ of index funds
well luckily, this isn't the case. And most index funds do ask index fund holders for the vote, tho not individually. But if the majority holders end up not following the board recommendations, the index fund would vote that (at least with vanguard - not sure about others).
Great, that. However the point of capitalism is that people who know the business would invest and make good decisions. In the situation that index funds hold most of the capital accountants will be making all decisions.
The point of capitalism is that the businesses that have people who know the business and make good decisions eventually win out over the businesses that don’t.
So the boards don't have totally unchecked power. But despite that policy being 22 pages long, it doesn't pay any attention to companies' individual circumstances.
Vanguard's voting policy doesn't have an opinion on EA's lootboxes, or Intel's 18A node, or Disney's approach to Star Wars.
You can’t easily dismiss the problem by pointing to history.
In those historic times, stock ownership was much more restricted to rich investors (not a good thing) who are far more opinionated in AGMs (a very good thing) than some faceless index fund or Robin Hooder who doesn’t even realize they should vote at all.
So boards used to perform but they also used to have pressure to perform. Will they still perform on autopilot? Maybe, but chaos always wins unless there’s a forcing function (your votes at the AGM).
"Most large investors vote alongside Glass Lewis and ISS recommendations, the two most prominent proxy solutions for investors. Most GL/ISS recommendations tend to vote with the board and don’t change much unless something drastic happens. If you’re a passive investor, you vote in line with the two proxy powerhouses."
>If you’re a passive investor, you vote in line with the two proxy powerhouses.
As a passive investor in SWPPX, an S&P 500 index mutual fund from Charles Schwab, "my" Intel stock votes are whatever Charles Schwab deems appropriate.
This is the correct answer. People still don't realise just how much power over virtually every public company in America (and therefore the economy) is concentrated in the hands of a handful of people at Vanguard you've never heard of, for example.
It's been a concern that I've had for a bit - if everyone recommends index funds, then you lose a lot of the underlying "value" behind them. You have fewer people making decisions about what stocks to buy. You get this really top heavy system.
It's not like most of the people making those decisions are doing any better than just throwing dice at a wall and seeing what sticks.
In the immortal words of Warren Buffett and Jack Bogle respectively: "The stock market is a device for transferring money from the impatient to the patient." and "The daily machinations of the stock market are like a tale told by an idiot, full of sound and fury, signifying nothing."
You can either gamble and blame yourself or ride the market (invest in index funds) and excuse yourself from losses. If you're just interested in making some money using some disposable cash, it makes even more sense to just ride the market.
Broad market index funds are the inflation protected asset due to US federal government bailout. Your savings account is not going to keep up with inflation, nor are TIPS or US treasuries (not for land/healthcare/education), but an SP500 fund will do a better job over the course of decades.
TIPS are literally designed to track inflation, Treasury I-Bonds are also designed to track and surpass inflation.
The US total stock market (ex-US stock market is a crapshoot) and its subset the S&P 500 index will generally do a better job than any bonds given a long enough timeframe, but that doesn't mean appropriate bonds can't do the job either.
That is why I specified land/healthcare/education. I guess I should have specified that it matters more for metropolitan areas, especially high cost of living regions.
TIPS won’t come close to making one be able to compete with other buyers in those markets for the non mass produced/imported resources.
If someone invested their money in TIPS over the last 30 or 40 years thinking they will be able to buy real estate because TIPS protected them from inflation, they would have been sorely disappointed for pretty much all non Midwest/interior northeast metros.
This is a demographic/political issue for all developed countries, they must reduce the purchasing power of their currency as a tax to be able to deliver the benefits expected by the more populous, older voting populace.
> If someone invested their money in TIPS over the last 30 or 40 years thinking they will be able to buy real estate because TIPS protected them from inflation
Does that change the reality that a 1997 30 year TIPS would have been ineffective at helping purchase what was once a below $100 per square foot home in 1997 that is now $300+ per square foot in most US population centers?
Replace home price change (or land price change) with education price change or healthcare price change. Probably even trades’ worker price change.
If a nursing home cost $x per month in 1997, and you thought putting away an equivalent amount of cash in TIPS will ensure you can afford a nursing home in 2027 or 2037, it’s probably not going to be fun to find out how much they cost now.
It worked great if you wanted to ensure being able to buy electronics, other manufactured goods, and probably groceries. But those are beneficiaries of automation and foreign labor.
3.6% annual return since Dec 2003. 1.036^20=2.0286.
Home prices have more than doubled since then, for a large portion of the US. Source for that is going to Zillow, searching a home in a major metro, and looking at price history.
For those who may not be aware of what you are talking about: 3 fund companies (BlackRock, State Street, Vanguard) are the largest shareholder in over 80% of SP500 companies and collectively own around 28% of the SP500 market cap.
Yes. If they collectively decide stock price matters more than anything else (sustainability, the planet, long term viability of the business), they get it, whatever the short or long term costs.
There was a comparison between Amundi (France based) and BlackRock, and their voting patterns, and BlackRock was consistently voting against any ESG or in any way ecology related proposals. Anything that isn't directly about making more money is just not their thing. Contrast that with Amundi who overwhelmingly voted for ESG or similar measures.
Electronic trading and passive investing drove commissions to zero so Blackrock helped create ESG because ESG requires paying commission for the service of certifying an investment as ESG.
> Also, how did Blackrock get so wealthy so fast? They've only been around since 1988.
It is not their money. They have roughly $11.5 trillion of assets under management, but their market cap is only $161.5b (on net income of about $6b). Compare that to xAI, which has existed for less than two years and has a valuation around $50b.
I am interested in learning about what you are referring to, but you linked his entire Substack and Twitter account. Would you please link a specific post?
Ok fair enough. But please explain like you might to a child, since these fund managers are exercising their power and since they are picking short term focused, parasitic, extractive boards like the Boeing/Intel ones, what impact are they trying to have on their funds? What fruit will the Vanguard 3030 fund reap when Intel and Boeing tank, say, maybe this year?
A fund manager can throw their weight behind short term gains, and when those gains dry up they can sell their Intel stake and put it in the next company to be sucked dry. They don’t have to care about long term success.
Fund managers are obligated to act according to their fund's prospectus[1], of which the specifics will vary with each fund.
For a TDF (Target Date Fund), because that was brought up (Vanguard 2030): Both actively and passively managed ones must generally be managed such that shareholders can start withdrawing adequate funds (selling shares) upon and after reaching the "target date".
For an S&P 500 index fund like the one I mentioned and hold (SWPPX), the fund manager is required to imitate the actual S&P 500 index as much as reasonably possible.
In short, "don't have to care about long term success" is not a generally usable argument for fund management.
If the prospectus says "follow the S&P 500" the fund manager is also not interested in long term success, they are interested in tracking the index.
But the fund managers tracking an index are not the main problem, they are just putting a lot of passive votes behind the funds that are actively working on electing board members in the interest of short term growth/profit (which brings more people to invest in their funds and gets them big bonuses).
You have to sell before the “gains dry up”, otherwise you won’t have much money to invest in a new company to have enough of a voice to suck it dry.
But for an index fund, there is no fund manager choosing when and if to sell. The investors of the fund are just following the markets, not really earning a lot (in real terms), but also not losing much.
The board might've been deliberately planted there to tank the stock in the long run.
Wouldn't be the first time hedge funds do this, but to be fair they prefer small pharma (famously cancer research/meds) startups or generally smaller companies to do it. Wouldn't be surprised though.
The way modern stock ownership is structured makes it almost impossible for shareholders to exercise any meaningful control.
Shareholders have no access to insider, commercially confidential information - so shareholders don't get to change the captain until after the ship's hit the iceberg. If I have shares in a video game company and the inept boss didn't organise enough testing so the game's got loads of major bugs? Well, I only find out after the damage is done. Is Gelsinger fucking up the delivery of 18A? I have no idea!
Meanwhile, individual shareholders' power is incredibly diffuse. The smart investor has a diversified portfolio, and even if I've literally got a million dollars invested in Intel, I still only own 0.0011% of the company.
Maybe I should coordinate with the other investors, you say? Get together with 1000 other similar investors, and we've got 1% between us? It's impossible, because they're all anonymous. There isn't anywhere I can rally the other shareholders.
And on top of that, loads of companies have dual-class share structures specifically designed to stop shareholders having any say. Whether you're invested in Facebook, FitBit or Ford - good luck exercising control when insiders' shares have 10x the votes yours have.
And that's without getting into passive investors and pension funds.
If I don't like Intel's current board, just selling my shares is far, far simpler than exercising any sort of meaningful active governance.
Note this is only true for companies registered under US law, especially under Delaware law. (And UK law to a lesser extent)
Which puts the onus of proof on the minority shareholders, to demonstrate they have a bonafide need for such information.
Most US states do so, which ironically makes the US one of the most authoritarian and dictatorial countries when it comes to minority shareholder rights.
Compare it with say Japan or China where the onus of proof is on the company to demonstrate why the requests of minority shareholders should be denied.
And the only real restriction is that any group of shareholders making such requests have to own at least 3% to 5% of the total shares.
From Fabricated Knowledge[0]:
"Meet Greg. He’s the former CFO and EVP of operations at Boeing. He’s been on the board since 2017 and was an interim CEO at Boeing during 2020. He also sits on the American Airlines board and is Chairman there. He sits on the Sierra Nevada Space Corporation board as well.
He has almost no semiconductor experience and could probably be directly involved with the Boeing fiasco. He’s been on the board for the entire Intel disaster and, at one point, was interim CEO of Boeing, so he's likely not the most focused member."
It's very, very difficult to shrink a public company until the board fully accepts that it has to happen.
The problem Gelsinger faced is one that many similarly placed CEOs face. A company is bloated when they come in; they have to make the company smaller; they become the sacrificial lamb for reducing its size and temporarily its income; so another CEO is brought in who recognizes the need to do the same thing; they are also fired; finally, the board brings in a white knight (typically someone with a moniker like "Chainsaw Al") who makes them see that the company has to shrink even more to get the turnaround. That CEO gets to keep the job and, if successful, win all the plaudits.
The point you’re missing is urgency. Convincing the board is also an important part of the CEOs job. They need to use all the tools at their disposal. They have to build the necessary relationships to smoothen out the board approvals for their plans. Better yet, they could reach out to major shareholders, pitch their plan to them, and use their influence to get the board in line. There is more than one way to achieve this. Pat must have tried some of this but didn’t succeed.
No offense, but what you say sounds like an argument for the tech approach.
Intel didn't lose the cloud, and what you describe as power efficiency was the fab/euv problem (that Pat fixed). AI: okay, but are you sure that Intel should aim to own every market? What if Intel fabs produce a lot of AI chips for fabless vendors?
No one has offered any real explanation for why the board would do this, now. 18a is about to go to hvm, which would carry large gains for server and consumer products, in addition to third parties. some mention "not listening to customers", which is peculiar, since customers (defined by current and past revenue) want faster, cheaper products (18a). about the only think I can imagine is that the board wanted a lower defect rate on large foundary chips (a potential product, not relevant to current products, except perhap Gaudi and maybe Altera). but firing Pat isn't going to improve defect rates...
It seems that “Non technical people don’t belong in management at technical companies” is an impossible lesson to learn for many, no matter how rich or credentialed.
There should be representation in management from every realm in which a company does battle on a complex field. A technical company should have strong representation from engineering, but it cannot and should not neglect lawyers in various specialties (tax, tort, labor) and include those knowledgable in finance, marketing and (when applicable) the supply chain. The members should ALL have "leadership" pixie dust. A companies leaders need to have insight and experience in every domain in which the firm faces either existential threat or growth opportunity.
I don't think that is necessarily true. But if you are non-technical, you need people advising you that are technical and actually listen to their advice.
The problem with non-technical decision makers is that they will mostly listen to the MBAs who think like them and not the technical people. They tend to waste their time on stupid things to prop up their egos, compensating for lack of understanding of their business fundamentals and what direction to go in.
I wonder whether that is a more general problem. E.g., maybe technical decision-makers tend to listen to other engineers and not give enough considerations to financials and market expectations.
Check out X these days. There is this "founder mode" meme that is about company founders are the culprit of growing bigger and fix how the company works.
Imho this is bullshit. It is not about being the founder, but most people just cannot change their ways of working. Like hired managers would not be able with the same chance. At the same time, when founders or engineers cannot change to accomodate some business reality, only time will tell wether they can and manage to lead through that - or the McKinsey usual mantra "selling off unprofitable business unit" would have been proved better.
I think the takeaway should be that if we want Western economies to keep growing, we need to cut these useless rich bureaucrats out and give power to the engineers.
Elon is an idiot but has a gift for convincing people to spend money building new things, particularly physical. There's a lot of people who could do great things if the funding was available and not just thrown into unicorn wannabe startups.
I’m convinced a long term successful business needs a >50% owner dictator who’s focused on more than the next quarters dividend payout. That sole owner dictator doesn’t even need to be particularly smart. They can play Diablo all day for all it matters. They just need to be able to make decisions that are longer term than the next balance sheet which the traditional executive class and shareholder structures are failing to do.
It's not the size of the technical insight that matters, it's what you know what do with it. Point being, a smart and capable CEO can still have misaligned incentives (read: optimize for personal benefit).
I think the question isn't just whether Gelsinger was wrong for Intel... It's whether Intel is capable of empowering any leader to do what’s necessary to rebuild...
If what you described above is true, well, good luck for competing with China's semiconductor sector given the US government is on its all out war with China on semiconductors.
Surely the Chinese can repeat what they achieved in EVs.
Just a small side comment: I don’t think that being technical and focusing on the right things are equivalent.
There are examples of people like Steve Jobs who knew how to surround themselves with the right technical people without being too technical themselves.
And then there are also the technical leaders that ignore the business side…
Steve Jobs wasn’t a programmer, but he was absolutely a techie. Go watch some of his marketing and keynote videos from the NeXT days: He understood the technology and more importantly why it mattered to the product and customers.
Also, think about the corollary: if that was true then most certainly the opposite is also true that technical people are terrible at business decisions.
Hardly my experience. Some are, but some are excellent business people.
>the opposite is also true that technical people are terrible at business decisions.
I'd say thst is generally true, yes. That's why the CTO isn't the CEO.
The big issue people are hinting at is that there's not much balance in this C suite these days. Just a bunch of people who all put money first, that isn't the CFO.
> These CEOs, all non-technical, focused on dividends, buybacks, and next-quarter results [..] The real issue? The board. Full of people like Boeing execs. They don’t get engineering. Trusting them to fix Intel is like hoping a plane door won’t pop open mid-flight.
And that economical model of maximum profit we have "chosen" as a society is exactly why we are possibly heading towards failure, as western societies and as economic entities in the very near future.
The problem is that once the cracks start to show, whether at Intel with the current crisis or at Boeing with doors blowing out and airplanes crashing due to gross negligence, so much engineering excellence and knowledge has been lost that the tide is pretty much impossible to stop.
Enshitification, whether applied to online platforms or to consumer goods is a similar expression of the same problem. Once the quality is lost, the rot has set in and there is simply no way back for the companies involved.
We may not have the power to fire anyone, but the overwhelming support for Pat shows we can still make an impact. Let the board know they made a mistake and leave a record of it: https://www.intc.com/board-and-governance/contact-the-board.
He was frequently on CNBC discussing Intel's prospects. At first I thought it was great for transparency, but it eventually became 'The lady doth protest too much, methinks.'
This is the classic innovator's dilemma. Any interesting new innovation from Intel gets snuffed out due to the fact it takes time to get to market. The success of x86 was blinding. Intel was impatient, so it killed products in their infancy (which leads to customer mistrust).
Intel did a great job coming up with great ideas before everyone else, then killing them off once someone decided they weren't going to immediately make 60% gross margins. I'm constantly amazed at how quickly the company managed to throw away good ideas because it could only make 40% gross margin, or was going to take another few years to develop a market.
That said Google could be doing worse. They at least have something around phones and AI. They're trying to attack some new markets. Google was structured so shareholders have no power but share price is still something that is going to drive decisions and hasn't been that amazing recently.
Intel has actually had many quarters of declining revenue, Google has been far more consistent in the last 15 years. Intel has had over 20 quarters where revenue declined since 2010 and they are spread out across 2010-2024, while Google only lost revenue in two quarters.
I would argue that Intel's poor performance is quite detached from Google's consistently exceptional performance. Heck, they aren't even in the same business (primarily hardware vs. primarily media (advertising) and software).
Google is:
- A member of the smartphone app store duopoly
- A member of the cloud compute big 3
- Nearly a pure monopoly in Internet search in Western countries
- Has a higher office productivity software marketshare than Microsoft Office
- Internet cable TV provider with the highest subscriber count
I'd like to understand how it "hasn't been that amazing recently."
YTD S&P 500 has a 26.9% return and GOOG has 25.7%.
Intel has been around since 1968 and Google since 1998. As you say they're also in a different business. But Intel has executed pretty well up to maybe 2007 so over about 50 years. vs. Google's 26 years of existence.
Most of your bullet points aren't anything new, this is Google riding on work that's been done a while back. The question is what will be the new bullet points and whether it will be able to stave competition eating into its existing bullet points. The pressure to get the stock price higher is going to take its toll as the bean counters try to artificially inflate it.
I’m not sure what less than one percent difference from the S&P 500 is supposed to prove good or bad. The S&P 500 is heavily weighted toward Alphabet, it literally is a big part of the S&P 500 itself.
And, you know, Exxon Mobil hasn’t really made a new product in quite some time but they’re doing just fine.
Still, I’d say that Google has had plenty of new products and bets that it has been making, along with great recent performance in many ways.
They’re finally hitting their stride with consumer hardware, with the FitBit acquisition clearly boosting their wearables business and selling a lot more pixel watches than any previous smartwatch effort.
Similarly, Pixel phones themselves have jumped from around 4% marketshare to 8-12% marketshare in the US just this year. Success in that arena is very new to Google, so I would call that a long term investment that is paying off.
Google is investing in custom Tensor processors which is also a recent product launch for them, starting in 2020.
Stadia was a new product that obviously failed in the end but did represent a decent technical achievement and major bet on a new product.
YouTube TV isn’t that old, either, it launched 2017, and being the top subscribed digital cable product is a big achievement.
The elephant in the room for newest major product launch is Bard AI and Gemini. It’s fair to say AI is less than proven and that these may be copycat products, but it’s also true that it’s going to be an area where your big competitors are basically Microsoft, Apple, and Google with little room for anyone else.
And let’s not forget that large companies can boost their value proposition and mitigate competitive threats through acquisitions. Even IBM’s lost decade was overcome, and a lot of it was through acquisitions and business unit spinoffs and closures that helped modernize their offerings (companies like Red Hat and Hashicorp).
Culture is not important for their revenue, most of the company is not. The ads business is their cash cow which finances the rest of the company. They can have horrible culture and mismanagement everywhere and revenue will still grow - if they don‘t fuck up their teams working on ads.
So what you're telling me is that I should be listening to you, a random person on the Internet, and expect that one of the most valuable companies in the world with decades of consistent revenue growth, a monopoly in internet search, duopoly in smartphone operating systems/app stores, one of the big 3 social media companies, one of the big 3 cloud compute companies, one of the big 3 AI companies, and the owner of the digital cable TV provider with the highest subscriber count is just going to magically not be successful in the future?
You might as well be telling me that you think that Saudi Aramco is going to go under because they need to fix their "culture."
You could argue Microsoft had the same disease. Ultimately, it's not as terminal because turning around a software-sized cruise ship is alot easier than turning around a hardware one.
Google can be refocused in a 4 years. Intel, at minimum, seems to need a decade.
Microsoft has a worse case of the same problem but greater monopoly force. Entirely dependent on the captive markets but rapidly snatching up everything they can to enforce it on other markets.
Gelsinger did not entirely do what needed to be done: cut the fat, _encourage_ the lean -- rebuild the culture of growth that made Intel strong. Re-light the fire.
Perhaps he wished us to identify the real culprits: the negligent board.
Or, he DID do what was right -- take the brunt of the blame for the hard things, and set the stage for the next act.
While gracefully giving a middle finger to the board. Let them eat their stock price they so desperately cling to as a talisman
Call me hopelessly naive, but other than hiring and firing the CEO, can a board really be blamed for a corporate culture and technical failings? Other than the CEO, I don't think I can name a board member of any company I've ever worked for and neither can I come up with an example about how they impacted the company.
Intel's board certainly has a lot to do with things like acquisitions, whether the company goes into certain fields. Foundary is a good example: either it was a good idea (and Pat definitely got it going with EUV), or it's a bad idea (and the board's to blame for letting the company work in that direction years ago).
It's pretty common for a board to trigger big restructurings - especially of firings, killing of product lines, divestiture.
All sorts of strategic decisions (M&A, new product direction, entry to new markets, significant financing events, vendor purchases above a certain $ threshold, even the hire of new executives above a certain level) require the CEO to get board approval.
> Intel needed a leader that could confront this cultural problem directly — who could work to undo an accretion of generations of entitlement — but if Gelsinger’s narrative for himself was any indicator, it felt like he would instead be feeding the company’s worst impulses about its own exceptionalism
Knowing both Pat and Bryan pretty well professionally, I'm not sure what the right answer here was. I did a "scare them" presentation to the senior Intel sales force a long time ago but it was mostly about multi-core Opteron. (And, I suppose it's long enough ago now, to say that Pat told me Gates was basically afraid of Windows requiring multi-core performance.) Intel probably needed to get into GPUs (or other AI-enablement) seriously earlier but I'm not sure how predictable that was. I did tell AMD a long time ago to head off in directions Intel's inertia wouldn't let them. But so much is about execution.
ADDED: And, yes, Pat was an Intel insider but I think he was still a pretty good choice for the job. Not sure who else I would go with.
> Intel probably needed to get into GPUs (or other AI-enablement) seriously earlier but I'm not sure how predictable that was.
You certainly couldn’t know AI was coming. But GPUs have been a thing since the 90s. And they only kept getting better. They were also the only thing other than CPUs that gamers would pay a high margin on.
Nvidia and AMD both made plenty of money on them. Intel had it easy. They had the best fab and could’ve easily bundled things for system integrators, netting lots of sales.
Instead they had no external GPU and the integrated one when they finally added it was a joke for a very long time. Their newest efforts sound like maybe they’re at least reasonable (though they don’t compete at the top end) but that’s like 15 years too late at least.
I don’t know. Seems to me like it would be a pretty obvious money making venture for them, and if they had gotten over the nonsense of Larrabee/making it out of x86 seems like they should’ve been able to have a really good chunk of the market easily.
The big win no one could’ve predicted is that if you had a good modern competitive GPU five years ago, then you were well positioned to pivot some for AI. At a minimum they would have a horse somewhere near the race.
In the old days gamers paid extra for sound cards etc., all these multimedia addons gradually got commoditised. I don't think it would've been a completely crazy bet to assume the same was going to happen to the GPU.
There were even rumors about "IGP is all you need" and Intel "canceling" Nvidia by eliminating PCIe so there would be nowhere to attach a discrete GPU. (Intel did cancel nForce.) In retrospect that was completely ridiculous.
The big problem for them is AMD. Anytime they do something stupid (Itanium) they have a competitor right there to take advantage of it.
Same problem IBM had. They didn’t like people using “their“ PC architecture, so they decided to make the MCA bus and other proprietary stuff. Force everyone to do what they want.
IBM‘s AMD was Compaq and all the clone makers. They all got together and used their power to prevent the market from being screwed up by IBM.
I could really see Intel wanting to make that kind of move but it never would’ve worked in reality.
I'm curious why you think there was not effort behind ia64.
I still think ia64 was sunk by the let-the-compiler-do-it tarpit. There were ambivalent aspects such as, at the time, questions of power or chip area. But things like memory latency are just not predictable enough to do strictly in-order pipelines. OoO won.
Had AMD not been around with AMD64, the industry would have had to get it working no matter what, HP and Microsoft were already on the Itanium train, and if production of x86 got slowly replaced by Itanium that was it, Windows and HP-UX would drag the ecosystems into it, and eventually the remaining issues would be sorted out.
No. HP was behind Itanium. In fact, it was basically their idea in the first place. I'm still not entirely sure why Intel went along although I suspect at least some of it was getting away from various x86 cross-licensing agreements.
I'm inclined to agree. Intel would have been loathe to go the 64-bit x86 route on their own. But a somewhat resurgent AMD pretty much forced their hand given the sorry state of Itanium development. Absent AMD, I'm pretty sure they'd have found a path to a "good enough" Itanium to sell successfully.
The compilers were fine, even if not producing maximally optimized code. Intel’s decision to add x86 compatibility circuitry to the die was probably the fatal one, it slowed everything down; made for terrible comparisons with existing x86 performance and generally signalled a lack of confidence. Something like Rosetta was out of the question, but they could have just had better transitioning tools for those code bases that couldn’t be recompiled easily.
Maybe early (say 94-99?). To some degree that happened. Most computers using an integrated GPU at this point, it’s mostly gamers or people who really need the power that have one on an expansion card.
Of course it’s not actually the CPU doing the work, it’s just in the same package.
Part of the difference is that audio can only get so good. But graphics don’t have much of a limit yet. We kept adding more colors and higher resolutions and higher refresh rates, more polygons, and even a second display for true 3-D.
It’s the true embarrassingly parallel problem we’re throwing more silicon at it just keeps making it better.
By the time Intel decided to get “serious“ and develop Larrabee, they clearly saw the market was there and would keep growing otherwise they wouldn’t have even bothered.
And honestly, whether it was on the same package as the CPU or an actual part of it the way normal vector units or the branch predictor are, who would benefit more from having a good graphics story to go on their GPU than the king of CPUs. They could use their advantages to just further cement their moat.
IMO the main reason why dedicated audio cards aren't a thing anymore is that Microsoft killed hardware accelerated audio as part of their driver API changes in Vista. The result is that games today have worse spatial audio than games that came out 20 years ago. If Microsoft hadn't done that, I think audio cards would have been in the same position as video cards where most users have their needs met by integrated solutions and gamers/high end users get dedicated cards (of course dedicated audio cards would have been more of a niche than GPUs).
Yeah there were old dedicated sound cards with support for spatial audio, and they were killed. But CPU's also just got really really fast in the meantime. You don't need dedicated hardware for spatial audio processing, it's cheaper and easier to do in software.
There have been enough attempts post-Vista to do spatial audio on GPUs that I don't think the argument that CPUs were fast enough really holds water. I think it's mainly that spatial audio hasn't been a must-have feature, so omitting it or half-assing it hasn't been a problem.
I wouldn't call a private implementation of spatial sound "solving the problem". If tech were that simple we'd have solve a lot of non-trivial problems.
Before AI it was HPC that was obviously becoming huge. AI is the most recent and buzziest form of HPC, but people have been buying AI-style clusters for a while. The Knights [*] products were a great entry point into that market if they would just figure out the software of it. Knights [*] essentially had Atom cores with huge SIMD units, and fast forward to today, an AMD GPU is essentially a small scalar core attached to a huge SIMD unit.
IMO it seems like they had the Google problem: if you can't sell a billion units it has to be canceled.
That's because Larrabee wasn't totally canceled, it was just pivoted from a gaming GPU to an HPC accelerator in the form of Knights Corner/Landing. The idea of having lots of x86 cores with wide SIMD units didn't change, but it was a lot more successful in the HPC world because anybody could compile an old MPI or OpenMP application with AVX512 and it just worked.
Larrabee itself was a complete disaster. Xeon Phi resurrected the Larrabee trash heap into something sellable. Sadly, it was never really that easy to program a Xeon Phi.
Intel did a good job on iGPUs, which satisfy most customers (even back in the i915 days).
But in retrospect, of course it was a missed market. It's hard to disentagle that choice from others in the same era (including the refusal to jump onto EUV or chiplets).
They did a good job in the sense that they provide the minimum neccessary baseline to run a graphical Windows environment without additional hardware. The attitude that it's a good job because it satisfied most of their current customers is exactly the problem with the whole companies approach to product development.
A lot of the technology behind EUV is just insane, but it's not like ASML has done so great financially. CMOS process shrinks have been a pretty impressive unicorn. There are various levers but not clear to me what will happen going forward.
Cell required Assembly programming Demoscene skill level to make good use of them, there was a C sdk, but really you wanted to write Assembly, and relatively tricky Assembly at it.
It is one of the reasons why Playstation 3 is the less celebrated one, and the console generation where XBox was able to pull ahead.
Larrabee, I was at a GDCE 2009 session where it was demoed, and while the idea was promising, it didn't seem to actually pull off in practice, plus there was hardly much hardware where we could actually try it out ourselves.
Xeon Phi didn't made it better regarding adoption, nor did AVX.
> Cell required Assembly programming Demoscene skill level to make good use of them, there was a C sdk, but really you wanted to write Assembly, and relatively tricky Assembly at it.
> It is one of the reasons why Playstation 3 is the less celebrated one, and the console generation where XBox was able to pull ahead.
In retrospect, Cell was a great architecture. In typical Sony fashion it was simply a few years too early. At the time games were single-threaded and many console games still expected CPU and GPU to be in sync.
On the PS2, the difference between using the vector accelerator or not was 1.4x – many developers never bothered. But on the PS3 this was simply not an option, as the performance difference was 6x (!).
The big issue at the time was dev tooling. Most existing engines had no support for dispatching jobs or any form of pipelining. Adding that required redesigning the engine from the ground up. On top of that the dispatcher/manager and the jobs use two different µarchs. Obviously, game and engine developers at the time hated cell with a passion.[1]
But even on the PC and Xbox, frequency scaling ended with the Pentium 4. Multi-core CPUs became mainstream. As CPUs stalled, most of the performance increase on PCs came from GPUs, that had moved from a fixed-function pipeline to shaders.
Slowly, even PC games had to do the bulk of their work in jobs dispatched across multiple cores. Engines that had adapted to Cell got a massive head start.
Today, the very textbook for game development – "Game Engine Architecture", written by Jason Gregory – uses the architecture of the Uncharted games on PS3 as the prime example for how to build a modern engine. As result, nowadays even small indie games are built that way.[2]
________________________
1. Gabe Newell: "Investing in the Cell, investing in the SPE gives you no long-term benefits. There's nothing there that you're going to apply to anything else" https://www.wired.com/2007/10/valves-gabe-new/
Yes, in that regard it was ahead of its time, and nowadays many game developers are exploring the possiblities to get rid of everything and explore pure compute for a full software rendering stack, running on GPGPU, agnostic from the actual rendering API.
Although I think it would still be better to use specific shading languages as the hardware isn't the same architecture, thus there is only so much one can do with traditional languages without extensions, or GPGPU specific algorithms/data structures, but that is me as outsider with interest in the field.
The experts seem to be enjoying exploring CUDA, SYCL, MSL, and similar for such purposes.
> required Assembly programming Demoscene skill level to make good use of them
Yeah, but writing efficient CUDA kernels, or D3D compute shaders, is not particularly straightforward either. At least GPU cores have direct access to global memory but still, for good throughput that memory access needs to be coalesced. Then there’re manually managed groupshared memory, atomics, wave/warp intrinsics in D3D12/CUDA, and now these special low-precision matrix multiplication instructions for AI.
People accepted all that complexity because performance is too good to ignore, both flops/dollar and flops/watt. Contemporary mainstream GPUs were just better than SPEs on Cell, or AVX-512 on Xeon Phis.
Specifically, PS3 (2006) delivered up to 200 GFlops FP32, first-generation Xeon Phi (incredibly expensive devices released in 2010) 750 GFlops, but GeForce 460 (released in 2010 for the price of $200) delivered up to 900 GFlops.
Same for AVX instrinsics. Very fun to use (for some of us), powerful when correctly used, but so hard to actually get right and to get the expected performance. I wish Intel had put more into ISPC back then.
Larrabee (or something similar) could have been an amazing real-time operating system target. Each little (Pentium!) core having truly its own local memory, no contention. But it was pitched as a GPU and the raw performance was not there.
Intel sucks at software and should have leaned in to their strengths - documenting the hell out of the low level workings of Larrabee. Workloads tuned to it would have been a great moat.
Another way to see this is NVIDIA invested heavily, heavily in making a language and its compiler useable (for some level of useable.......) on enthusiast-affordable and poor-scrappy-researcher-affordable hardware on a complex weird-to-program-from-asm architecture and provided good-enough performance (and showcasing lots of very good optimizations) on standard HPC libraries and a bit of HW/SW co-design and kept at it for a long long time... with some level of backwards compatibility too. The recipe is here.
I remember having this argument 10+ years ago. Discrete GPUs like what gamers were buying were thought to be a small market with no real routes to bigger markets. Workstation and HPC could be addressed with entirely separate product lines, hence the disaster that was larrabee. It was Nvidia that pushed the unified workstation/consumer architecture. ATI reused core architecture, but they had very different software stacks above with their firestream stuff.
If you were to design a GPU several years ago, people would probably tell you designing it for AI is crazy and cryptocurrencies and raw hashing power is where the money's at.
In the end, Nvidia kind of got lucky with their bet on AI. While CUDA has been a thing for a while, it was limited to mostly academic and workstation workloads. It wasn't until LLMs and things like Stable Diffusion when it really became popular.
Unfortunately for non-NV users, most of these AI workloads when run locally assume CUDA is present and it's a considerable effort to make them work on non-NV GPUs. It's fairly okay for common workloads like running basic LLMs and SD models, but as soon as you get into more advanced/obscure stuff the harder it gets without CUDA.
I see flavors of this narrative quite often. Nvidia has been positioning their GPU's for AI since the late 2000's. And their GPU's were being used for AI at least as early as 2011
To me, it doesn't seem fair to discount Nvidia's foresight. A lot of smart people saw this opportunity even before the invention of LLM's. That's a bit like building a light bulb before discovering electricity.
I do think it's unfair to discount Nvidia's success as "having gotten lucky." On the other hand, I also think the back-to-back 1-2 punch of crypto and LLMs was probably hard to foresee as something way more important than a handful of high-end gamers and maybe some commercial linear algebra applications.
> The big win no one could’ve predicted is that if you had a good modern competitive GPU five years ago, then you were well positioned to pivot some for AI
Nvidia pivoted to AI almost 2 decades ago (late 2000's). And big tech recognized GPU's potential around the same time.
Google brain was 2011 and tensorflow was publicly released in 2015 (much earlier internally).
The "Attention is all you need" paper came out in 2017. That was for many folks, another smoking gun that GPU's are fundamental to AI.
The point is: Intel has had a looong time to get into GPU's. They just chose not too. And it wasn't because "nobody saw ai coming".
With products just leaving behind the old Apple days and Foundry, in fact, not a 1st tier supplier anymore what is there to salvage do you think? For the latter money was not enough even in, much, better days.
I find it interesting that the X2 chip they are using as Tofino's successor, made by Xsight. One of the investors of Xsight is Intel Capital. So seems Intel had a hedge against themselves.
The Emeryville Eye mentions a night of looting (probably opportunistic rather than planned by protest organizers) but doesn't mention any fires burning (a) building(s) to the ground.
I think there's merit to the argument that Gelsinger was unwilling to kill the culture to save the business. Intel and he seemed trapped by pride, nostalgia, complacency, and fear.
But some of the points feel a tad personal. I've found Pat and Bryan to be similar as professionals, so I'm not sure how much of this is a mote in one's eye, and a beam in the other's.
* Can't win over AMD, Nvidia, Apple, Broadcom as fab customers
* Fab cancellations and delays, wasting a ton of money
* Not hiring the right people to steer Intel's internal fab culture to external quick enough
* Not simplifying product roadmap. No one knows what the lake code names are. Expect 30% of roadmap to get cancelled. 30% get delayed. 30% switch node tech. Compare this to the simplified AMD roadmap, which is easy to understand and makes sense.
* Didn't stop paying dividends until August 2024
* He cut fab funding to pay dividends in 2022 [0]
* Didn't see that Intel was swimming naked during the covid boom and that after the boom, Intel would be in huge trouble due to inferior products
* He hired 20% more workers since he joined but 54% less revenue
* Intel still has more employees in 2024 than in 2019. Instead of trimming fat, he added more fat.
Sure, Intel was on the decline no matter who stepped in as CEO in 2021. However, Pat definitely accelerated the decline and made a mess of things with little progress in any area.
Gelsinger was a failure all around and it's time someone with no ties to the original decades-long Intel IDM strategy step into the role. Intel needs an outsider who can come in and objectively see the current situation - not the rosy glorious Intel of the past.
Kind of disagree. Have read on forums that by all accounts, the tech in the 18A process is amazing (Gate-all-around transistors, backside power-delivery, 3d packaging tech...), and that they are on the path to catchup/pass TSMC in terms of chip-fabrication performance/power-expenditure in a year or two. Yeah, this is an amazing achievement for such a large company.
But have also ready that to make this profitable is like a 5-8 year plan, not something that is profitable in 3. To me, this is his biggest mistake, he needed to keep the ship sailing at least somewhat smoothly while waiting for this big gamble to pay off.
And eventhough I do agree with you that he made a lot of mistakes, still he got the the most important issue right, regaining manufacturing leadership.
This forum is full of engineers and they talk about 18A (I myself am only a software engineer but who took hardware design classes waay back in college, so can understand 60% what they're are saying but am no expert).
And obviously, many of these posters are like me and have a bias towards Intel succeeding. Still, it sounds like 18A seems like it's on track in terms of the technologies (at least for now)
No they are not, but one big part of their problem they have in their attempts to catch up to TSMC is that they can't get reliable yields above 30%, while TSMC has a better node AND they have 60% yields.
Yeah, that is why some outsiders say this is more like an 8 year strategy, it takes time to convert customers.
IMHO, if you have a great product and a market with significant demand, seems like eventually the customers will come - and it seems like this is possible if Intel can retain manufacturing leadership for a few years and build out a better ecosystem to support external chip development by foundry customers, but this of course take time.
> * Intel still has more employees in 2024 than in 2019. Instead of trimming fat, he added more fat.
Intel hiring practices under Gelsinger were really puzzling. Did he think that by hiring thousands of random university graduates there will be the next Gordon Moore among them? It was also discussed on HN: https://news.ycombinator.com/item?id=42300028
I can kind of understand the extensive poaching of TSMC engineers in Arizona. Though making TSMC upset is playing with fire as much of Intel's business now depends on them.
> Can't win over AMD, Nvidia, Apple, Broadcom as fab customers
Not surprising, and arguably not to be expected. His 5 nodes in 4 years plan ends in 2025, and only then is the fab "good enough" for external customers.
Specifically, they want to see how well Intel's own product on 18A performs before committing. So the real test is in 2025-2026.
> Fab cancellations and delays, wasting a ton of money
The cancellations and delays are on the expanded fab roadmap that he introduced. Compared to before he joined, the fabs are in a much better place. Essentially, this is complaining that he committed to opening 5 new fabs and only opened 3 (or whatever the number is).
> Not hiring the right people to steer Intel's internal fab culture to external quick enough
Perhaps. 5 nodes in 4 years, which Intel may actually achieve, is unprecedented, though. No other company has done it (and likely never will again - it's crazy expensive). So at least as far as execution goes, the fab is in much better shape than when he joined.
> Not simplifying product roadmap.
Definitely agree.
> Didn't stop paying dividends until August 2024
They cut them significantly in early 2023. Yes, they should have done it all earlier, but keep in mind that the board needs to approve it. Or at least, not fire him for doing it. On the plus, he did stop stock buybacks almost immediately after joining.
> Didn't see that Intel was swimming naked during the covid boom and that after the boom, Intel would be in huge trouble due to inferior products
I'm sure he knew the products were inferior. What his failure was is to see that the overall PC market would drop significantly. As an example, in one year it dropped globally 25%. Intel's share dropped 33%, with AMD picking up the remaining 8%. That extra 8% for AMD was something like a 30-50% increase in sales.
> He hired 20% more workers since he joined but 54% less revenue
Totally agreed. What's worse, after the 5% cuts in 2022, Intel was back at the same headcount before the current layoffs. How do you lay off 5% of the company, claiming you need to save money and in 18 months grown the company to the same level?
I think the employee count blowup was his biggest mistake because it shortened Intel's runway towards being an external fab company.
Lip Bu Tan, who was on the Intel board, recently resigned because he did not agree with Pat's employee count plan and company culture. Tan was likely right.
Specifically, they want to see how well Intel's own product on 18A performs before committing. So the real test is in 2025-2026.
Another mistake of Pat's is that he did not split the company fully. AMD, Nvidia, Apple, Qualcomm aren't voluntarily going to fund a competitor by using Intel fabs. Furthermore, Intel not splitting means there will always be small conflicts of interests with Intel designs teams. Pat wanted the cake and eat it too. He wanted to compete with TSMC in addition to Nvidia, Apple, AMD, Qualcomm. Now Intel is at risk with competing against neither group.
I think not having any major external customer signed onto 18A in 2024, approaching 2025, is a huge failure on Pat's part.
> I think the employee count blowup was his biggest mistake because it shortened Intel's runway towards being an external fab company.
It was a mistake, but in terms of runway, would not have made a large difference. The capital costs of running a fab dwarf the cost to pay salaries. That's why they still do fabs in the US - outsourcing to another country doesn't save much money.
> Another mistake of Pat's is that he did not split the company fully.
I don't know that anyone knows what his plans were on this. Internally within the company (and perhaps externally), he was clear on a few things:
- If Foundry cannot become profitable, he'll shut the fabs down and kill Foundry altogether.
- The Products side will be charged market rates, and they can choose if they want to use Intel's fabs or TSMC
- If an external company is willing to pay more than Intel Products, they'll prioritize the external company.
So he was trying to set it up so that it was effectively two separate companies. I guess it was either harder than expected, or there was significant pushback. When he joined, the two sides were heavily intertwined. His goal for 2023 was to get things separated enough so we could get separate P&L statements for Foundry and Products. He achieved it - but I don't know if it could have been done quicker.
The other (obvious) reason he didn't want to break the company up: Intel Foundry is unprofitable. They need the revenue that the Products brings in to keep it alive. The goal was to hit profitability in 2027-2028, when there are enough customers. Splitting it up now will kill Foundry altogether (and now with the CHIPS act, they are not allowed to kill Foundry).
As for the concern about funding competition - not sure how much that is true. The vast majority of TSMC's customers (including AMD) do not want to use the leading edge node. It's more expensive, and has higher quality problems. They are happy to let Apple take it. Likewise, the narrative was that most of these companies will prefer not using Intel's latest node, and would be happy to let Intel use that capacity. So I think the plan was to make 18A mature, move to 14, and then external customers will use the mature 18A node.
Because Intel fired Gelsinger, I suspect that in 18 months, there will be no US based foundry doing advanced logic technology development. Maybe that’s better for the industry because all the money will be concentrated on one or two competently run foundries (TSMC and Samsung) but it’s unfortunate for Intel fab employees who have both ultra specialized skills and will be stigmatized for coming from a failed semiconductor manufacturer.
Could you explain stigma? I don't understand why skilled workers would be looked down upon for business decisions out of their hands, being at the wrong place at the wrong time. Aren't there smaller fabs that would appreciate their skills?
In addition smaller fabs tend to have broader/less specialized roles. For example an integration engineer at an analog fab would have much broader scope than their counterparts at Intel and thus the Intel engineers would/could struggle to adapt.
You are paying for two TD teams rather than three TD teams so the costs are amortized better. It also means that employees are more likely to stick to one employer which leads to more preservation of knowledge, etc.
Articles like this are so easy to write after the event. It would be a lot more compelling of an article if it had been published a year ago and had instead predicted what was going to happen given what the author claimed was obvious at the time. Fact is Intel was in a really tough situation and Pat seemed to many (including many here) like the best candidate at the time to be able to turn things around, with the acknowledgement that the situation was dire and there might be nothing anyone could do to save it.
There are reasons I didn't write this a year ago (perhaps obvious ones?), but yes, it was pretty clear. And in fact, in going back through my own timeline, I found a DM between my co-founder and me on March 23, 2023 in which I predicted that Gelsinger would be fired on November 1, 2024 -- and my co-founder predicted March 1, 2025.
>It’d be more impressive if these predictions were shared beforehand
It’s hard to disagree with that.
Based on your comment above, it sound like “the receipts” are messages between you and one other person. Messaging a prediction to a friend is very different than publishing it for the world to see.
398 comments
[ 4.7 ms ] story [ 320 ms ] threadThe narrative machine for Intel is in full force given how much monies is in play, I wouldn't take anything that seriously.
>What is (or was when Gelsinger became CEO) the move for Intel to get back on top?
Whatever it takes to drive stonks up up up, obviously. If there was one thing Gelsinger could legitimately be blamed for, it is that he wiped out Intel's market valuation. Nothing else matters, because his "interim" replacements are both beancounters whose literal job is to drive stonks up up up.
Bean counters can bring stock value up temporarily. The trick is to get out before shit hits the fan. For a comparison, see Boeing. Where shit is hitting the (turbo) fan right now as well.
I have no doubt Nvidia will start losing its way when Huang retires, like almost all other founder-less companies such as Intel.
Being a founder doesn't magically make you the best person for the job of CEO of a company. Huang happens to be a very good leader, but a lot of founders become a drag on their companies later on.
It's just an anecdote, I worked for one place where the company truly outgrew the founder's abilities. He was just some guy with an art degree who was at the right place at the right time with his dorm room business idea. I would argue that his lack of ability and experience squandered a lot of the later stage growth opportunities that could have made the company into a billion dollar public corporation.
I think about a company like WeWork, that's a company that would not exist today if the founder was not kicked out. Some companies like McDonald's would never have expanded to be a household name if their founders were still in charge.
Their decisions may be no more prescient than anyone else's, but their decisions are reliably in different directions than anyone else's.
Their goal is different, the direction they always aim is different, and that consistent aim has an effect over time regardless what mistakes and misfortunes happen along the way.
If you have a specific and clear goal on the horizon and are always aiming for that even though you have to deal with obstacles closer in front of you, over time you still get somewhere.
If your goal is just chasing whatever random butterflies float by, you get nowhere.
How do you get a CEO to believe innovation is the best thing for them and the company? Even if you get them to believe that how do they get their company to innovate?
Innovation is hard work. It involves failures. Just buy back your shares and make money. Then someone else comes and eats your lunch. That is how the world works.
bean counters are for industries like beans - where the market is stable and the only way to liberate shareholder value is to catch the stray beans. they are exactly the wrong answer when "disruptive innovation" is afoot.
Larabee was a mess....anybody who as actually been involved with designing a chip which breaks new ground (and is not just repackaging of somebody else's IP) knows that designing and taping out ANY AND EVERY chip is a big mess. It is not easy to tell heaven from hell on this one--and arguably, Intel's management at the time made the wrong call. Pretty much every decision they were making at the time was the wrong decision, which is why Intel is in the state it is in.
When I worked at Intel, Robert Noyce had unfortunately died, but we still had Andy Grove on the business end and Gordon Moore on the technical end. All three of those people are not just good, not even merely great. They were the kind of transcendent geniuses who rewrite the world, and which come along once every 2 or 3 hundred years. Impossible to replace.
AFAIAK, If Gelsinger couldn't fix Intel, it was irreparably broken by the time he got there. We should celebrate him for having the courage to try to fix what 4 previous CEOs couldn't fix.
I understand that it's painful when one of your suppliers has financial difficulty and has to exit a business you depend on. I feel ya. I hope that nobody will be writing blog posts about the OP's decision to partner with Intel in the first place, pointing to it as one of the fatal mistakes made by the company and their leaders.
Software and design support. As pointed out in the article.
Intel is genetically incapable of valuing anything other than "The Fab(tm)". VLSI designer ... second class citizen. Software ... second class citizen. etc.
The ARC graphics card series is an exemplar of the problem. It was always going to be a market disaster. Fine, accept it. However, Intel should have pushed a couple of cards for free to every single HPC/AI grad department in the US. And perhaps even tossed in a $50K grant to boot. That's, what, $50K*100=$5 million max and a thousand graphics cards? Sure, the top 10-20 departments are well-funded and still going to use Nvidia--the rest will chew on your graphics cards. In return you get a legion of grad students developing software for your hardware. And you will quickly get feedback as to what you need to fix for v2.
However, you have to value software for that idea to penetrate. And Intel just ... doesn't.
That would be worthless. It's so weird that people think academia is contributing in any meaningful way to the top lines or bottom lines here. You're free to look at the commit history of PyTorch/TF/Triton/HF/llama/whatever and judge for yourself.
> PyTorch/TF/Triton/HF/llama/whatever and judge for yourself.
Where do you think the creators of most of these current models came from and started their careers?
More critically, the ecosystem the arose for these to exist was subsidized by research for decades.
I’ve worked at several companies where there have been decent HPC/GPU codebases and at all of them the staff working on that have largely come out of academia in one way or another.
isn't the same as in academia.
Nvidia used to do this in the late 2000s and early 2010s.
It did wonders to drive CUDA uptake versus Vulkan, and Nvidia would sponsor plenty of researchers and students to leverage CUDA and Nvidia GPUs in Parallel Programming classes.
> you have to value software for that idea to penetrate
I'm not sure it's software per say and more so Developer Experience.
If a vendor fails to truly interact with their implementers, they lose critical feedback.
Intel used to have a very active Technical Marketing Engineer org (the old word for DevEx), but I'm not sure how active or relevant it is anymore.
This aligns with my experiences trying to use Intel software.
The software was often technically very good, but the user experience trying to use it was horrible, and god help you if you got an error.
Larrabee all over, really.
I suppose linking to [1] and [2] is probably germane for this discussion.
I swear there was another blogpost explaining the tradeoffs made in the Larrabee/Xeon Phi development process in exacting detail, and confirming that it ran gaming workloads pretty well before being shelved, but I can't find it again.
[1] - https://pharr.org/matt/blog/2018/04/18/ispc-origins
[2] - http://tomforsyth1000.github.io/blog.wiki.html#%5B%5BWhy%20d...
For cross platform programming on GPUs people in academia/industry just gave up because performance was worse, the tooling was terrible and library support was non existent. OpenCL, OpenACC, OpenMP GPU offloading have all been disappointing.
From https://en.wikichip.org/wiki/intel/process#Timeline those are the 14nm, 10nm, and 7nm respectively.
There is no reason why TSMC should be leading Intel on fabrication by so much. Intel has had a crazy amount of time to restructure and refocus when we consider how long their x86 intellectual property lead kept them afloat with their subpar fabrication and delayed manufacturing process updates. They spent many years having faster chips than anyone else despite having subpar fabrication. But now, they can't really escape the fact that ARM is exploding and AMD has the most flexible x86 platform where as each year goes by the number of use cases where Intel is the best choice diminishes.
If you play games, the x3D chips are basically the only sensible choice. If you are looking for servers, AMD is consistently 10% cheaper in all the cloud platforms. If you want a mobile system, anything involving ARM and/or TSMC will get you better battery life.
The fact that fabless players in the market like AMD and Apple can outpace Intel in so many product lines is, frankly, embarrassing. While I think that other processor design shops selling off fabrication divisions made sense, I think that Intel as a company basically makes no sense if they can't get their fabrication shit together. Because if Intel can't fabricate bleeding edge chips what's the point of using their IP?
The biggest problem Intel has at the moment seems to be they can’t split up and they can’t work together.
If you keep both the fab and the design, you have problems. The thing that makes Intel all the money is their CPUs, and because of the state of the fab many of them are made at TSMC.
But TSMC isn’t too keen on giving a good deal to someone who wants usurp them. So Intel is paying more per chip than they would if they just spun off the fab. If they did that, they could have higher margins/lower prices.
But if you spin off the fab, can they actually get enough business to be able to survive? They’re not on the leading process. They need a massive investment to catch up. Despite their geopolitical importance it seems like they would basically need a real continuing subsidy just to exist, which is effectively being provided by the CPU design side of the business right now right?
It’s a real damned if you do damned if you don’t kind of thing.
I don’t know what the right thing to do is. Splitting them up kind of seems like the better idea but I don’t know if Intel’s ego could take that. They were the fab company. Their advantage let them survive multiple large mistakes that could’ve killed other companies.
Now they’ve all piled up.
PS: the CHIPS act money says it goes to Intel. It sounds like if they try to split they may lose it, so would first have to convince the government to let the fab part keep it
Intel needs to step back, enjoy the runway the CHIPs act got them, and really think hard about the top to bottom design and fab of an absolutely killer 2027 / 2028 chip would look like. They aren’t as good at design as Apple, and they aren’t as good at fab as TSMC, but they are within striking distance of both and they have that vertical integration opportunity.
If they can catch up, but not exceeded, then they can’t make a better product than they do right now. I guess it should cut their costs some amount (don’t know how much) which would certainly be a big plus.
But without taking the lead again… would that be enough? I suppose if they caught up they could also fab other people‘s chips as Pat wanted to. If they’re tied for best (or at least close) they could probably pick up accounts that TSMC doesn’t have space for or are perhaps too small to help fill out their capacity.
While I was listening to the podcast I was thinking about how it almost seems like sort of a historical accident that they’re both design and fabrication. Does anyone else do that at this point? I know it was common early on but maybe it’s just not that sustainable by now.
I don’t know. I guess I see splitting the company is more likely to succeed, simply because the design side works and would lose the anchor that it’s tied to.
Successfully pulling off a fix while keeping the company intact may be a better outcome if possible though.
But I don’t think they have to be better than TSMC, or even quite as good. Intel’s strength is the design + fab integration. If they are 90% as good as fab as TSMC and 95% as good as Apple or Nvidia at design… maybe that’s sufficient to be competitive.
* Signal Intel is in it for the long term, specifically kill the dividend.
* Be more customer centric/rebuild customer trust, specifically publicly acknowledge the 10nm slip.
Though it seems customer trust in Intel has eroded to the point where only actions will change customers' perception of Intel — I'm not sure what Intel could have communicated to Tofino/Oxide that they hadn't already.
What other market was Intel chasing that would have yielded a large new market? Oh right, Mobile CPUs. I mean besides that? An actually well-done Linux distro that helped showcase Intel's hardware without waiting for Windows to release 9x+1. Ok, besides that? Good laptop hardware. Ok Ok Ok.
But it still stands, even if 1.0 and 2.0 are a disaster, iterate and work on it.
The big problem is that Intel is a quarterly earnings focused company. Well, and even if Gelsinger is good, the middle management is pretty clearly totally dead sea. If middle management is bad, they can't attract talent, and can't retain talent, and can't support talent. Without talent, no vision or execution.
Intel doesn't value intellectual labor. It's the bottom line with the company. An arrogant technically competent leader is still better than the alternatives.
That said I have no direct or even indirect interaction with Gelsinger. All my strong opinions are derived from 30 years of being in software and PCs.
Of course, until the day that it is not. Intel's management arguably made the wrong decision when they shut down Larrabee. Lets face it, at the time they were not making very many good decisions, which is why Intel is in the mess its in now.
The fact that VMWare didn't become the market leader in cloud computing like AWS says a lot.
That’s why VMware tried to pivot to sticky stuff like EUC and dev platforms.
If VMWare had the product vision at the right time they could have been AWS.
It’s also easier to sell massive datacenters when you’re good at building distribution centers and have low margins. VMware was a software company with that margin expectation. They tried a partner model that didn’t really work. The only other good option was something like IBM or Sun back in the day.
I lived it - my first job was a software company that primarily served call centers in a specific vertical. The customers wanted out of the phone business and online… so they offered the full package. Fast forward a couple of years and the dot-com bubble bursts, and the investment dollars and lines of credit dry up. The founder mortgaged his house to keep the place open 60 days to complete an acquisition, which ultimately went through and everyone did ok. Two years earlier, we’d all be millionaires. 3 weeks later, holders of worthless paper for a bankrupt company. Timing matters more than anything.
You're proposing the idea that VMWare had no idea how to build data centers as a company that builds data center abstraction virtualization software almost exclusively? VMware is a company that partnered with data center hardware companies to make VMWare-branded data center appliances. And they just...don't know how to set up a data center?
If Amazon's expertise in e-commerce was a direct line of sight totally obvious foray into cloud computing we'd see Walmart, Target, and eBay operating AWS their own competitors.
Amazon leadership saw an opportunity and understood the 20 year vision of how enterprise computing would work in the future.
VMWare's managers are solely responsible for not seeing that future and failing to empower their teams to create new products in new categories. That is particularly embarrassing since they are subject matter experts on abstracting away hardware in the data center.
If Intel had been practising making discrete GPU from 2009 to 2024 instead of giving up then it does seem quite reasonable that Nvidia would be a quarter of its size. One of the obvious mis-steps Intel made over the last 20 years was failing to keep trying Larrabee-style chips until they figured out how to make them work. Ok version 1 might be a disaster and version 2 might be bad. But we're looking back over almost 20 years! There was a lot of room to figure out how to be good and tap in to the huge new market that was opening up.
Larrabee is one such product. It wasn't immediately great, and didn't immediately promise great returns, and didn't quite have a market yet, so it was killed. I can't imagine where Intel would be in the AI world if they had the foresight to stick with Larrabee.
Somehow they forgot that all of their highly profitably businesses were risky bets at one time.
Good on them for sticking with their promises for firebase. But it’s a pity about stadia. Don’t bet your business on a flip of the Google coin.
And that all of their highly profitable businesses are risky bets long term.
In an infinite universe, anything can happen, and so "it might not work" is is always true for everything. No matter what you do, it might not work, and some of the most important and valuable things ever aren't attainable without a lot of investment before you can start to reap any return.
Not all ideas are good ideas, so of course there are reasons not to do some things, or to abandon something you started, but merely "what if it never works" is not one.
Asking that is about equal to asking what if a piano falls on you tomorrow. Sure it could happen, but so what? What does that change? If you don't walk down the street for fear of pianos, you don't get to the grocery store or your job and then you die from starving instead of from a piano.
[0] https://brightsideofnews.com/blog/an-inconvenient-truth-inte...
In an absolute best case scenario, he may be right.
But that seems incredibly unlikely.
Still, they should’ve had a much MUCH better GPU story than they did.
I think Cantrill spelled it out in the article, and I think people are overlooking (and for some reason really continue to overlook, or just have no idea) that nvidia's advantage is Software.
I remember Larabee as a teenager - I just graduated high school, the forums were talking about it, and I had finally saved enough money to buy an 8800GTX. The 8800GTX was also coincidentally, the first consumer GPU to support CUDA. Almost 2 decades later and it feels like nvidia has an insurmountable moat in AI due to CUDA. This wasn't an accident, nvidia spent a ton of time, with developers, getting it right. FTA:
>that NVIDIA had an indisputable lead, even in 2009; that for Intel to dominate NVIDIA it would have required conjuring software expertise and focus with which Intel has famously struggled
Ignoring this piece is huge in claming that nvidia would be a quarter of the size. Can you even imagine Intel investing in a non-x86 developer platform for 15 years while they were in the lead?
You’re dead right that Nvidia is where they are because of their software story and all the work they put into it.
I don’t know if Intel could have learned that lesson.
Intel invented this model with x86.
And more recently, the reviews I read of the Archmage graphics drivers didn't make it sound like Intel had understood the importance of the software that was essential to a pretty big hardware release.
AI hasn't cared about Cuda for years - sure, Torch on NV will use it, but AI is all about Torch/TF, not details below that.
This could be 100% true, and the best ML engineers in the world could be completely oblivious to CUDA and it wouldn't change why nvidia is leading right now.
Anyone who thinks it's "trivial" to use AMD cards should help geohot and his tinygrad boxes. Despite emailing the CEO several times, it was a clear uphill battle to get something working, and even now the sales page lists driver quality as "mediocre".
If you open a GitHub issue or a forum post about your weird CUDA application that isn't working right, you will have an engineer help you. This is in nvidia's DNA and they did this before AI started printing money. On the flipside, you can email Lisa Su today and still not have problems fixed in AMDs drivers. I'm sure AMD is trying, this isn't to say they aren't, but having decent software culture at these large companies is hard.
If AMD can't do it, I feel even less confident about Intel. If the CUDA moat truly didn't matter, we would see AMD cards in demand. However it doesn't and it isn't for lack of trying.
That doesn’t mean, that ML people don’t care about some hardware specifics btw. Of course they do! Because CUDA is very, very good at exploiting the specific features of the newer cards (I’ve forgotten, what exactly this was though). But take away the simple integration into their favorite AI-SDK and it would be completely irrelevant.
The point is just that, if Intel were to build strictly better hardware, but not provide the interface to torch, then they would still loose to Nvidia. Yes, yes, this is not completely true. Given enough advantage someone else would build the IntelCUDA. But this is literally the bet that those academics make. And it’s so risky, that only academics are doing that
This was, if anything, the main theme at Supercomputing: “there is not even a point in talking about Nvidia’s new benchmarks. You all are going to buy it anyways” coupled with very few “we are betting on buying cheap hardware (Intel/AMD GPUs) and hope that we can build the relevant parts of CUDA ourselves” and the latter is pretty much a desperation move of labs/sites that simply cannot get NVIDIA GPUs (either price or availability).
And yes that is probably the seed of the end of Nvidia’s dominance. But it will take 20 years and multiple fuck ups. Just as it did with Intel
You need to remember that was during the financial crisis, and the first chips were going to be laughably bad and uncompetitive. From the quote in that article:
> Justin Rattner (Intel Senior Fellow) demonstrated Larrabee hitting one teraflop, which is great but you could walk across the street and buy an ATI graphics board for a few hundred dollars that would do five teraflops.
Intel had no chance for v0 to do anything but lose money. They were way off track, and spent most of the time trying to get to a software rasterizer plus hardware that could work for some games.
They did pivot Larrabee into the various Knight chips (Knights Corner, Landing and whatever else) since the "it's just x86" was a pretty compelling story in the 2009 era. But ultimately, HPC wasn't that great a market in the early 2010s for accelerators yet and NVIDIA ended up totally eating their lunch over time.
I think Intel could have stuck with the HPC market and gotten lucky once ML exploded, but realistically until about 5 years ago there wasn't that much ML spending yet either. 10 years of last place discrete GPU work would have been difficult to invest in.
tl;dr: the business case for Larrabee hinged on having a successful gaming business and it wasn't even close.
Was Pat perfect, no. But Pat acknowledged Intel’s problems - something Otellini, Krzanich, and Swan never did. These CEOs, all non-technical, focused on dividends, buybacks, and next-quarter results while Intel fell behind in advanced nodes and innovation.
Gelsinger inherited a disaster: 10nm delays, TSMC pulling ahead, and no GPU strategy. He had the courage to cut buybacks and slashed dividends. He poured billions into fabs in Arizona, Ohio, Germany, and Ireland. He delivered Intel 18A, powered on first silicon, released PDK 1.0 for Microsoft, and secured Microsoft and Amazon as customers. There were even rumors Apple might join.
Contrast this with Nadella at Microsoft back in 2014. He didn’t reboot the company by tearing everything down. Instead, he built on Ballmer-era wins like Azure, Office 365 while shifting Microsoft’s focus to the cloud. Gelsinger had to start from scratch in many ways, tackling years of neglect while facing harsher challenges.
Yes, Intel’s stock dropped $150 billion, but Gelsinger was upfront - it wouldn’t turn around before 2025. He was trained by Gordon Moore and Andy Grove, and while some saw him as arrogant, that confidence came from decades of technical leadership.
The real issue? The board. Full of people like Boeing execs. They don’t get engineering. Trusting them to fix Intel is like hoping a plane door won’t pop open mid-flight. They’re the ones who should be replaced.
BK is who did his best to sink the company, multiple mis-guided layoffs, bad top level hirings, and stupid direction changes every time he read a new news article. Don't even mention his absurd AMA on reddit where he couldn't stop using the ellipse and laughing like a 14 year old girl.
Bob Swan wasn't the guy who could right the ship. He was another Otellini type guy, someone you bring in to not mess stuff up. Pat was a great pickup for Intel, and his return was my first time hoping that Intel could actually survive and grow.
Unfortunately, he came by too late. Foundry STILL hadn't really made progress, and the rot in the rest of the company had set in too far. The brain drain from BK was still being felt everywhere.
Frankly, I still feel like Pat was the right guy, 10 years too late. He joined to right the ship, but by the time he came aboard, the ship had already hit the iceberg, and the bow was 30 feet in the air. And now he's being blamed for the ship sinking.
Companies should focus more on their nominal products/services and spend less time playing games with money.
Already is. There's a reason why many quant finance books model stock prices as martingales/random walks.
Dividends
> dividends are highly discouraged by taxation
This is only true today because buybacks aren't taxed
Buybacks are ultimately a way of saying "we don't have a better way to spend this money." Consider that, in a world without buybacks, execs have two choices. One is to pay dividends, and eat the tax implications. The other is to find productive ways to spend the money to increase the company's earnings. The taxation of dividends strongly motivates companies to innovate. Buybacks weaken that motivator significantly, because there's no tax implication. It's financial engineering and not a productive use of money.
Buybacks are also a pretty neat way for insiders to enrich themselves at the expense of shareholders.
Example: Insiders schedule a sell of their shares to occur right after announcing a buyback plan.
So any talk of incentives should include a justification of passive investment being more valuable than work for income, if someone is asking for favorable tax treatment.
What changed in the 80’s is that a safe harbor was created. Before that companies were at risk of being charged with market manipulation - there were no clear rules on what was market manipulation and what wasn’t.
The safe harbor rule 10b-18 simply laid out requirements for buybacks fell into a “safe harbor”, presumed to not be market manipulation.
But companies could still do them before that, but it was more risky.
The Arrow Lake launch certainly didn’t make things easier for Pat.
For anyone interested, https://www.youtube.com/watch?v=0m4hlWx7oRk.
It's cool to see how Pat’s work helped developers like Linus.
Literally, Gregory Smith, former Boeing CFO is a member of the board.
It's true they've completely fallen off the pace. But people tend to forget how rapidly this happened. Even as late as the semi-aborted 2018 launch of Cannon Lake it seemed like it was just a routine burp they'd correct with a process respin. Then TSMC quietly reached parity with 7nm, shipped 5nm which was a better process, and by 2021 Apple had jumped ship and Intel was falling behind even AMD.
The disaster happened fast. Boards of Directors aren't that agile.
Seems like the leaders just lost the stomach for taking risks, a long time ago. No forays into mobile or GPUs, at least not in the billions of dollar and many years scale that was needed. No stomach to pay the competitive salaries necessary to compete with Apple, Microsoft, Alphabet, Meta, Amazon, Netflix, etc for talent.
Well, it did good enough in netbooks. It could probably have been good in tablets if they kept trying (and if non-iPad tablets really caught on).
https://www.thurrott.com/hardware/64677/elite-x3-hp-takes-wi...
But the way Apple insiders tells this story, there was no way Intel was even being considered in the (short!) window when the original iPhone was being built. Intel was in the middle of selling Xscale, and even that design was too power-hungry.
Intel missing mobile was a long history of poor strategic and tactical choices, not one bad call.
Obviously it was going to be very difficult to compete as a third platform with the behemoths iOS and Android become during those years. At least the MeeGo and Windows Phone cards were not the winning ones.
But there was also a hardware story how Nokia would start Intel silicon. I don't think anything of that has ever been publicly annouced before it failed. Wasting a year seems to be massive underestimate. I believe it must have been much longer. After Nokia started to fail Intel hired former Nokia engineers. I have no reliable insights what they did there, but I believe at least in the beginning they still worked with phone hardware on low-level software.
Meanwhile, Intel's chip designers kept targeting an unusable process, and wasted years that they should have been iterating on designs for the fab process that actually worked. Skylake shipped in 2015. They didn't deliver a new CPU microarchitecture on 14nm until 5.5 years later, a year and a half after they shipped that same microarchitecture in a mobile-only form when their 10nm finally started to be somewhat usable (but not fast enough for desktop).
What were the chip designers doing for all those years? In 2015, Intel knew that 14nm had been harder to bring up than any previous fab process, and they knew that 10nm was proving even harder, but they refused to try making an updated CPU design for 14nm. How could the management not have realized that spending multiple consecutive years not shipping new designs would cause long-term damage to their capability to iterate on CPU designs? Not participating in the feedback loop of actually shipping left Intel with an oversized P-core design and an E-core design that wasn't well-matched to it, making Alder Lake awkward and slapdash when they finally got 10nm working well enough for desktop CPUs.
They complain about arrogance, but even if you accept that, it was arrogance BEFORE Gelsinger, with Intel under the control of MBAs that they're talking about.
And can I just say, I've seen some seriously arrogant assholes in the tech departments I've worked ... but for absolute incredible arrogance, you need MBAs.
The leadership (not technical) are disconnected from reality.
Did engineers know there were problems? Of course, they are smart, but the leadership doesn’t listen
Once Intel lost its research focus it became an extractive company extracting the riches that were already there, instead of creating true innovation. Case in point - Intel stopped doing it's research day long time ago.
Design pipelines are deep and Intel at the time famously had very node-specific designs without industry-standard PDKs. The moment engineers were told to switch a design to 14nm, it basically reset the 5 year design-to-product pipeline. Management failed because they did not hedge the risk by starting a parallel 14nm design effort at first sign of 10nm troubles. They likely were engaged in magical thinking or some variation of the "Are YOU going to tell him?" Silicon Valley scene. It does not help that information like that is considered actionable insider trading information. I bet a lot of people working on 10nm designs first heard the news about the delays from the quarterly investor calls.
Right. It was well-known publicly that Intel was running their business in a way that maximized the damage any fab troubles would have on their product roadmap. It was obvious a decade ago that Intel needed more flexibility to bring their CPU designs to other fab processes. It took them too long to start working on Rocket Lake, and too long to deliver it. But they have at least made some progress on the problem, since they've been selling x86 CPU cores made at TSMC for the past year.
(On a related note: Buying Altera and forcing them to port their entire roadmap over to a broken 10nm process was made even more stupid by the fact that Intel didn't have a usable PDK that outsiders and acquisitions could work with.)
The true embarrassment was when SMIC (read: China) reached 7nm and thereby surpassed Intel last year (or was that 2022?).
Intel then proceeded to waste CHIPS monies and other aid on five digit layoffs and now ousted the one CEO who ostensibly at least had the right idea.
At this point I want to see Intel fail (and Boeing too), American Exceptionalism(tm) absolutely needs to have its longass Pinocchio nose broken in half before we have any hope of rebooting ourselves.
The "even" makes the tone of your comment feel a tiny bit disrespectful towards AMD. By 2021, it was clear to me that AMD had their gloves off and were winning. Zen 3 was released in 2020 - the third generation of nearly flawless execution by AMD that Intel failed to respond to - outside of cutting the prices on some CPUs. For a while, Intel held onto the "fastest single-core speeds". Back in 2017, my first thought after being blown away by the performance of a first-gen Zen PC build was "I should buy shares in AMD" - AMD clearly had a superior product with an even better value proposition.
I would not say that the first gen of Zen is was a clear winner over Skylake. It took a couple iterations before AMD clearly took the lead. AMD was simply so far behind that several large generational improvements were needed to do better than Intel.
In 2017, I would not have said that either for Zen 1 without qualification[1]. Zen 3 on the other hand, was a winner.
That said, 1st gen Zen had better bang-for-buck than Intel, for multicore workloads - in my case, I had built a workstation and thr equivalent intel build would have cost much more, expensive Ryzen motherboards notwithstanding.
1. In my comparison as I buyer, I didn't compare Intel and AMD processors by core count, but by what I'd get with my budget. The AMD build I eent with was better than an intel build for the same amount of money.
So indeed, by 2018, even though Intel has not yet fallen, it's actually already late. The roots of the problems seems to be earlier, and that's where the CEO, and the Board, should have reacted.
Too many retail investors just vote with the "board recommendations" all the way down the ticket every year, if they even bother to vote.
well luckily, this isn't the case. And most index funds do ask index fund holders for the vote, tho not individually. But if the majority holders end up not following the board recommendations, the index fund would vote that (at least with vanguard - not sure about others).
That seems like what is happening.
So the boards don't have totally unchecked power. But despite that policy being 22 pages long, it doesn't pay any attention to companies' individual circumstances.
Vanguard's voting policy doesn't have an opinion on EA's lootboxes, or Intel's 18A node, or Disney's approach to Star Wars.
Passive investing is cheaper, this is what “passive” means.
Historically, in the aggregate, boards of US public companies are competent enough to create good returns without strategic investor direction.
In those historic times, stock ownership was much more restricted to rich investors (not a good thing) who are far more opinionated in AGMs (a very good thing) than some faceless index fund or Robin Hooder who doesn’t even realize they should vote at all.
So boards used to perform but they also used to have pressure to perform. Will they still perform on autopilot? Maybe, but chaos always wins unless there’s a forcing function (your votes at the AGM).
- The Death of Intel: When Boards Fail
https://www.fabricatedknowledge.com/p/the-death-of-intel-whe...
As a passive investor in SWPPX, an S&P 500 index mutual fund from Charles Schwab, "my" Intel stock votes are whatever Charles Schwab deems appropriate.
In the immortal words of Warren Buffett and Jack Bogle respectively: "The stock market is a device for transferring money from the impatient to the patient." and "The daily machinations of the stock market are like a tale told by an idiot, full of sound and fury, signifying nothing."
You can either gamble and blame yourself or ride the market (invest in index funds) and excuse yourself from losses. If you're just interested in making some money using some disposable cash, it makes even more sense to just ride the market.
The US total stock market (ex-US stock market is a crapshoot) and its subset the S&P 500 index will generally do a better job than any bonds given a long enough timeframe, but that doesn't mean appropriate bonds can't do the job either.
TIPS won’t come close to making one be able to compete with other buyers in those markets for the non mass produced/imported resources.
If someone invested their money in TIPS over the last 30 or 40 years thinking they will be able to buy real estate because TIPS protected them from inflation, they would have been sorely disappointed for pretty much all non Midwest/interior northeast metros.
This is a demographic/political issue for all developed countries, they must reduce the purchasing power of their currency as a tax to be able to deliver the benefits expected by the more populous, older voting populace.
TIPS have been available less than 30 years.
Replace home price change (or land price change) with education price change or healthcare price change. Probably even trades’ worker price change.
If a nursing home cost $x per month in 1997, and you thought putting away an equivalent amount of cash in TIPS will ensure you can afford a nursing home in 2027 or 2037, it’s probably not going to be fun to find out how much they cost now.
It worked great if you wanted to ensure being able to buy electronics, other manufactured goods, and probably groceries. But those are beneficiaries of automation and foreign labor.
Claims that inflation adjusted bonds don't actually track (average) inflation need evidence.
3.6% annual return since Dec 2003. 1.036^20=2.0286.
Home prices have more than doubled since then, for a large portion of the US. Source for that is going to Zillow, searching a home in a major metro, and looking at price history.
Also, how did Blackrock get so wealthy so fast? They've only been around since 1988.
There was a comparison between Amundi (France based) and BlackRock, and their voting patterns, and BlackRock was consistently voting against any ESG or in any way ecology related proposals. Anything that isn't directly about making more money is just not their thing. Contrast that with Amundi who overwhelmingly voted for ESG or similar measures.
So an oligopoly that presents the illusion of free market?
It is not their money. They have roughly $11.5 trillion of assets under management, but their market cap is only $161.5b (on net income of about $6b). Compare that to xAI, which has existed for less than two years and has a valuation around $50b.
xAI has yet to actually move the market needle, other than getting venture capitalists to sign away money they would have done otherwise.
For a TDF (Target Date Fund), because that was brought up (Vanguard 2030): Both actively and passively managed ones must generally be managed such that shareholders can start withdrawing adequate funds (selling shares) upon and after reaching the "target date".
For an S&P 500 index fund like the one I mentioned and hold (SWPPX), the fund manager is required to imitate the actual S&P 500 index as much as reasonably possible.
In short, "don't have to care about long term success" is not a generally usable argument for fund management.
[1]: https://www.investopedia.com/terms/p/prospectus.asp
But the fund managers tracking an index are not the main problem, they are just putting a lot of passive votes behind the funds that are actively working on electing board members in the interest of short term growth/profit (which brings more people to invest in their funds and gets them big bonuses).
You have to sell before the “gains dry up”, otherwise you won’t have much money to invest in a new company to have enough of a voice to suck it dry.
But for an index fund, there is no fund manager choosing when and if to sell. The investors of the fund are just following the markets, not really earning a lot (in real terms), but also not losing much.
Wouldn't be the first time hedge funds do this, but to be fair they prefer small pharma (famously cancer research/meds) startups or generally smaller companies to do it. Wouldn't be surprised though.
Shareholders have no access to insider, commercially confidential information - so shareholders don't get to change the captain until after the ship's hit the iceberg. If I have shares in a video game company and the inept boss didn't organise enough testing so the game's got loads of major bugs? Well, I only find out after the damage is done. Is Gelsinger fucking up the delivery of 18A? I have no idea!
Meanwhile, individual shareholders' power is incredibly diffuse. The smart investor has a diversified portfolio, and even if I've literally got a million dollars invested in Intel, I still only own 0.0011% of the company.
Maybe I should coordinate with the other investors, you say? Get together with 1000 other similar investors, and we've got 1% between us? It's impossible, because they're all anonymous. There isn't anywhere I can rally the other shareholders.
And on top of that, loads of companies have dual-class share structures specifically designed to stop shareholders having any say. Whether you're invested in Facebook, FitBit or Ford - good luck exercising control when insiders' shares have 10x the votes yours have.
And that's without getting into passive investors and pension funds.
If I don't like Intel's current board, just selling my shares is far, far simpler than exercising any sort of meaningful active governance.
Which puts the onus of proof on the minority shareholders, to demonstrate they have a bonafide need for such information.
Most US states do so, which ironically makes the US one of the most authoritarian and dictatorial countries when it comes to minority shareholder rights.
Compare it with say Japan or China where the onus of proof is on the company to demonstrate why the requests of minority shareholders should be denied.
And the only real restriction is that any group of shareholders making such requests have to own at least 3% to 5% of the total shares.
He has almost no semiconductor experience and could probably be directly involved with the Boeing fiasco. He’s been on the board for the entire Intel disaster and, at one point, was interim CEO of Boeing, so he's likely not the most focused member."
[0] - https://open.substack.com/pub/mule/p/the-death-of-intel-when...
The problem Gelsinger faced is one that many similarly placed CEOs face. A company is bloated when they come in; they have to make the company smaller; they become the sacrificial lamb for reducing its size and temporarily its income; so another CEO is brought in who recognizes the need to do the same thing; they are also fired; finally, the board brings in a white knight (typically someone with a moniker like "Chainsaw Al") who makes them see that the company has to shrink even more to get the turnaround. That CEO gets to keep the job and, if successful, win all the plaudits.
It is clear how it is setup that "to survive we must act now, not in a week, and we must do this expensive sell-off"
The whole movie is very well worth watching: https://www.imdb.com/title/tt1615147/
There have been so many bad technical and operational decisions at Intel that are clearly the responsibility of the CEO and Senior Leadership Team.
And it doesn’t require an engineering degree to understand that they missed the boat on power efficiency, AI, cloud etc.
No one has offered any real explanation for why the board would do this, now. 18a is about to go to hvm, which would carry large gains for server and consumer products, in addition to third parties. some mention "not listening to customers", which is peculiar, since customers (defined by current and past revenue) want faster, cheaper products (18a). about the only think I can imagine is that the board wanted a lower defect rate on large foundary chips (a potential product, not relevant to current products, except perhap Gaudi and maybe Altera). but firing Pat isn't going to improve defect rates...
People tend to hug their ruts.
Imho this is bullshit. It is not about being the founder, but most people just cannot change their ways of working. Like hired managers would not be able with the same chance. At the same time, when founders or engineers cannot change to accomodate some business reality, only time will tell wether they can and manage to lead through that - or the McKinsey usual mantra "selling off unprofitable business unit" would have been proved better.
Zuck, Elon, Bezos etc.
But financial engineering is often at odds with real engineering and can harm the product.
If what you described above is true, well, good luck for competing with China's semiconductor sector given the US government is on its all out war with China on semiconductors.
Surely the Chinese can repeat what they achieved in EVs.
Him or the CHIPS act?
There are examples of people like Steve Jobs who knew how to surround themselves with the right technical people without being too technical themselves.
And then there are also the technical leaders that ignore the business side…
There are so very few non-tech people who knows how to run a tech company.
Also, think about the corollary: if that was true then most certainly the opposite is also true that technical people are terrible at business decisions.
Hardly my experience. Some are, but some are excellent business people.
I'd say thst is generally true, yes. That's why the CTO isn't the CEO.
The big issue people are hinting at is that there's not much balance in this C suite these days. Just a bunch of people who all put money first, that isn't the CFO.
And that economical model of maximum profit we have "chosen" as a society is exactly why we are possibly heading towards failure, as western societies and as economic entities in the very near future.
The problem is that once the cracks start to show, whether at Intel with the current crisis or at Boeing with doors blowing out and airplanes crashing due to gross negligence, so much engineering excellence and knowledge has been lost that the tide is pretty much impossible to stop.
Enshitification, whether applied to online platforms or to consumer goods is a similar expression of the same problem. Once the quality is lost, the rot has set in and there is simply no way back for the companies involved.
Google revenue has only declined twice, and never by more than 3%.
https://www.macrotrends.net/stocks/charts/GOOG/alphabet/reve...
That said Google could be doing worse. They at least have something around phones and AI. They're trying to attack some new markets. Google was structured so shareholders have no power but share price is still something that is going to drive decisions and hasn't been that amazing recently.
I would argue that Intel's poor performance is quite detached from Google's consistently exceptional performance. Heck, they aren't even in the same business (primarily hardware vs. primarily media (advertising) and software).
Google is:
- A member of the smartphone app store duopoly
- A member of the cloud compute big 3
- Nearly a pure monopoly in Internet search in Western countries
- Has a higher office productivity software marketshare than Microsoft Office
- Internet cable TV provider with the highest subscriber count
I'd like to understand how it "hasn't been that amazing recently."
Intel has been around since 1968 and Google since 1998. As you say they're also in a different business. But Intel has executed pretty well up to maybe 2007 so over about 50 years. vs. Google's 26 years of existence.
Most of your bullet points aren't anything new, this is Google riding on work that's been done a while back. The question is what will be the new bullet points and whether it will be able to stave competition eating into its existing bullet points. The pressure to get the stock price higher is going to take its toll as the bean counters try to artificially inflate it.
And, you know, Exxon Mobil hasn’t really made a new product in quite some time but they’re doing just fine.
Still, I’d say that Google has had plenty of new products and bets that it has been making, along with great recent performance in many ways.
They’re finally hitting their stride with consumer hardware, with the FitBit acquisition clearly boosting their wearables business and selling a lot more pixel watches than any previous smartwatch effort.
Similarly, Pixel phones themselves have jumped from around 4% marketshare to 8-12% marketshare in the US just this year. Success in that arena is very new to Google, so I would call that a long term investment that is paying off.
Google is investing in custom Tensor processors which is also a recent product launch for them, starting in 2020.
Stadia was a new product that obviously failed in the end but did represent a decent technical achievement and major bet on a new product.
YouTube TV isn’t that old, either, it launched 2017, and being the top subscribed digital cable product is a big achievement.
The elephant in the room for newest major product launch is Bard AI and Gemini. It’s fair to say AI is less than proven and that these may be copycat products, but it’s also true that it’s going to be an area where your big competitors are basically Microsoft, Apple, and Google with little room for anyone else.
And let’s not forget that large companies can boost their value proposition and mitigate competitive threats through acquisitions. Even IBM’s lost decade was overcome, and a lot of it was through acquisitions and business unit spinoffs and closures that helped modernize their offerings (companies like Red Hat and Hashicorp).
It is possible that Google maintains current revenue levels. It seems unlikely they will have significant growth unless they fix their culture.
Even assuming the teams working on ads are immune to bad culture, how do they grow revenue when their inventory declines?
Google is facing a substantial crisis, and their culture issues reduce their odds of making the necessary transformations.
1. https://www.gartner.com/en/newsroom/press-releases/2024-02-1...
You might as well be telling me that you think that Saudi Aramco is going to go under because they need to fix their "culture."
Google can be refocused in a 4 years. Intel, at minimum, seems to need a decade.
Perhaps he wished us to identify the real culprits: the negligent board.
Or, he DID do what was right -- take the brunt of the blame for the hard things, and set the stage for the next act.
While gracefully giving a middle finger to the board. Let them eat their stock price they so desperately cling to as a talisman
It's pretty common for a board to trigger big restructurings - especially of firings, killing of product lines, divestiture.
Who else is responsible for the DIRECTION of a company but the board of DIRECTORS?
> Intel needed a leader that could confront this cultural problem directly — who could work to undo an accretion of generations of entitlement — but if Gelsinger’s narrative for himself was any indicator, it felt like he would instead be feeding the company’s worst impulses about its own exceptionalism
ADDED: And, yes, Pat was an Intel insider but I think he was still a pretty good choice for the job. Not sure who else I would go with.
You certainly couldn’t know AI was coming. But GPUs have been a thing since the 90s. And they only kept getting better. They were also the only thing other than CPUs that gamers would pay a high margin on.
Nvidia and AMD both made plenty of money on them. Intel had it easy. They had the best fab and could’ve easily bundled things for system integrators, netting lots of sales.
Instead they had no external GPU and the integrated one when they finally added it was a joke for a very long time. Their newest efforts sound like maybe they’re at least reasonable (though they don’t compete at the top end) but that’s like 15 years too late at least.
I don’t know. Seems to me like it would be a pretty obvious money making venture for them, and if they had gotten over the nonsense of Larrabee/making it out of x86 seems like they should’ve been able to have a really good chunk of the market easily.
The big win no one could’ve predicted is that if you had a good modern competitive GPU five years ago, then you were well positioned to pivot some for AI. At a minimum they would have a horse somewhere near the race.
Same problem IBM had. They didn’t like people using “their“ PC architecture, so they decided to make the MCA bus and other proprietary stuff. Force everyone to do what they want.
IBM‘s AMD was Compaq and all the clone makers. They all got together and used their power to prevent the market from being screwed up by IBM.
I could really see Intel wanting to make that kind of move but it never would’ve worked in reality.
I still think ia64 was sunk by the let-the-compiler-do-it tarpit. There were ambivalent aspects such as, at the time, questions of power or chip area. But things like memory latency are just not predictable enough to do strictly in-order pipelines. OoO won.
Had AMD not been around with AMD64, the industry would have had to get it working no matter what, HP and Microsoft were already on the Itanium train, and if production of x86 got slowly replaced by Itanium that was it, Windows and HP-UX would drag the ecosystems into it, and eventually the remaining issues would be sorted out.
UNIX systems, I also doubt, as we were still in the days each vendor had their own CPUs.
AFAICT, they expected Itanium to own the high end server/HPC market, while 32-bit x86 would continue supporting PC’s for another decade.
That seems remarkably short-sighted in retrospect.
For what it's worth, I followed Itanium as an IT industry analyst and wrote a short book recently on the topic. https://s3.amazonaws.com/bitmasons.com/docs/Lessons+from+the...
Of course it’s not actually the CPU doing the work, it’s just in the same package.
Part of the difference is that audio can only get so good. But graphics don’t have much of a limit yet. We kept adding more colors and higher resolutions and higher refresh rates, more polygons, and even a second display for true 3-D.
It’s the true embarrassingly parallel problem we’re throwing more silicon at it just keeps making it better.
By the time Intel decided to get “serious“ and develop Larrabee, they clearly saw the market was there and would keep growing otherwise they wouldn’t have even bothered.
And honestly, whether it was on the same package as the CPU or an actual part of it the way normal vector units or the branch predictor are, who would benefit more from having a good graphics story to go on their GPU than the king of CPUs. They could use their advantages to just further cement their moat.
This was already solved 7 years ago, but few games actually make use of it: https://valvesoftware.github.io/steam-audio/
IMO it seems like they had the Google problem: if you can't sell a billion units it has to be canceled.
But in retrospect, of course it was a missed market. It's hard to disentagle that choice from others in the same era (including the refusal to jump onto EUV or chiplets).
It only seems nonsense in retrospect. No one expected nVidia to improve flops/dollar in their GPUs that quickly.
And Larrabee wasn’t the only architecture deprecated by GPGPUs, Cell Broadband Engine by Sony+Toshiba+IBM was another one.
It is one of the reasons why Playstation 3 is the less celebrated one, and the console generation where XBox was able to pull ahead.
Larrabee, I was at a GDCE 2009 session where it was demoed, and while the idea was promising, it didn't seem to actually pull off in practice, plus there was hardly much hardware where we could actually try it out ourselves.
Xeon Phi didn't made it better regarding adoption, nor did AVX.
> It is one of the reasons why Playstation 3 is the less celebrated one, and the console generation where XBox was able to pull ahead.
In retrospect, Cell was a great architecture. In typical Sony fashion it was simply a few years too early. At the time games were single-threaded and many console games still expected CPU and GPU to be in sync.
On the PS2, the difference between using the vector accelerator or not was 1.4x – many developers never bothered. But on the PS3 this was simply not an option, as the performance difference was 6x (!).
The big issue at the time was dev tooling. Most existing engines had no support for dispatching jobs or any form of pipelining. Adding that required redesigning the engine from the ground up. On top of that the dispatcher/manager and the jobs use two different µarchs. Obviously, game and engine developers at the time hated cell with a passion.[1]
But even on the PC and Xbox, frequency scaling ended with the Pentium 4. Multi-core CPUs became mainstream. As CPUs stalled, most of the performance increase on PCs came from GPUs, that had moved from a fixed-function pipeline to shaders.
Slowly, even PC games had to do the bulk of their work in jobs dispatched across multiple cores. Engines that had adapted to Cell got a massive head start.
Today, the very textbook for game development – "Game Engine Architecture", written by Jason Gregory – uses the architecture of the Uncharted games on PS3 as the prime example for how to build a modern engine. As result, nowadays even small indie games are built that way.[2]
________________________
1. Gabe Newell: "Investing in the Cell, investing in the SPE gives you no long-term benefits. There's nothing there that you're going to apply to anything else" https://www.wired.com/2007/10/valves-gabe-new/
2. e.g., Tiny Glade: https://www.youtube.com/watch?v=jusWW2pPnA0.
Although I think it would still be better to use specific shading languages as the hardware isn't the same architecture, thus there is only so much one can do with traditional languages without extensions, or GPGPU specific algorithms/data structures, but that is me as outsider with interest in the field.
The experts seem to be enjoying exploring CUDA, SYCL, MSL, and similar for such purposes.
https://home.otoy.com/render/octane-render/
If you prefer actual code,
https://advances.realtimerendering.com/s2015/aaltonenhaar_si...
and
"Mesh Shaders - The Future of Rendering"
https://www.youtube.com/watch?v=3EMdMD1PsgY
"GPU driven Rendering with Mesh Shaders in Alan Wake 2"
https://www.youtube.com/watch?v=EtX7WnFhxtQ
Yeah, but writing efficient CUDA kernels, or D3D compute shaders, is not particularly straightforward either. At least GPU cores have direct access to global memory but still, for good throughput that memory access needs to be coalesced. Then there’re manually managed groupshared memory, atomics, wave/warp intrinsics in D3D12/CUDA, and now these special low-precision matrix multiplication instructions for AI.
People accepted all that complexity because performance is too good to ignore, both flops/dollar and flops/watt. Contemporary mainstream GPUs were just better than SPEs on Cell, or AVX-512 on Xeon Phis.
Specifically, PS3 (2006) delivered up to 200 GFlops FP32, first-generation Xeon Phi (incredibly expensive devices released in 2010) 750 GFlops, but GeForce 460 (released in 2010 for the price of $200) delivered up to 900 GFlops.
Intel sucks at software and should have leaned in to their strengths - documenting the hell out of the low level workings of Larrabee. Workloads tuned to it would have been a great moat.
?? People have been buying GPUs since the late 90's, it was always an obvious growth vector.
And forget AI, we were using GPUs to run statistical programs since the late 00's...
In the end, Nvidia kind of got lucky with their bet on AI. While CUDA has been a thing for a while, it was limited to mostly academic and workstation workloads. It wasn't until LLMs and things like Stable Diffusion when it really became popular.
Unfortunately for non-NV users, most of these AI workloads when run locally assume CUDA is present and it's a considerable effort to make them work on non-NV GPUs. It's fairly okay for common workloads like running basic LLMs and SD models, but as soon as you get into more advanced/obscure stuff the harder it gets without CUDA.
I see flavors of this narrative quite often. Nvidia has been positioning their GPU's for AI since the late 2000's. And their GPU's were being used for AI at least as early as 2011
To me, it doesn't seem fair to discount Nvidia's foresight. A lot of smart people saw this opportunity even before the invention of LLM's. That's a bit like building a light bulb before discovering electricity.
Nvidia pivoted to AI almost 2 decades ago (late 2000's). And big tech recognized GPU's potential around the same time.
Google brain was 2011 and tensorflow was publicly released in 2015 (much earlier internally).
The "Attention is all you need" paper came out in 2017. That was for many folks, another smoking gun that GPU's are fundamental to AI.
The point is: Intel has had a looong time to get into GPU's. They just chose not too. And it wasn't because "nobody saw ai coming".
The entire 6 square kilomtres burnt to a crisp and no mention on the wikipedia page?
https://en.wikipedia.org/wiki/Emeryville,_California
The Emeryville Eye mentions a night of looting (probably opportunistic rather than planned by protest organizers) but doesn't mention any fires burning (a) building(s) to the ground.
But some of the points feel a tad personal. I've found Pat and Bryan to be similar as professionals, so I'm not sure how much of this is a mote in one's eye, and a beam in the other's.
* Can't win over AMD, Nvidia, Apple, Broadcom as fab customers
* Fab cancellations and delays, wasting a ton of money
* Not hiring the right people to steer Intel's internal fab culture to external quick enough
* Not simplifying product roadmap. No one knows what the lake code names are. Expect 30% of roadmap to get cancelled. 30% get delayed. 30% switch node tech. Compare this to the simplified AMD roadmap, which is easy to understand and makes sense.
* Didn't stop paying dividends until August 2024
* He cut fab funding to pay dividends in 2022 [0]
* Didn't see that Intel was swimming naked during the covid boom and that after the boom, Intel would be in huge trouble due to inferior products
* He hired 20% more workers since he joined but 54% less revenue
* Intel still has more employees in 2024 than in 2019. Instead of trimming fat, he added more fat.
Sure, Intel was on the decline no matter who stepped in as CEO in 2021. However, Pat definitely accelerated the decline and made a mess of things with little progress in any area.
Gelsinger was a failure all around and it's time someone with no ties to the original decades-long Intel IDM strategy step into the role. Intel needs an outsider who can come in and objectively see the current situation - not the rosy glorious Intel of the past.
[0]https://semianalysis.com/2022/07/29/intel-cuts-fab-buildout-...
But have also ready that to make this profitable is like a 5-8 year plan, not something that is profitable in 3. To me, this is his biggest mistake, he needed to keep the ship sailing at least somewhat smoothly while waiting for this big gamble to pay off.
And eventhough I do agree with you that he made a lot of mistakes, still he got the the most important issue right, regaining manufacturing leadership.
https://semiwiki.com/forum/index.php?threads/intel-18a-too-g...
And obviously, many of these posters are like me and have a bias towards Intel succeeding. Still, it sounds like 18A seems like it's on track in terms of the technologies (at least for now)
TSMC not only has the best fabs, the chips fly out of those fabs at blinding speed.
IMHO, if you have a great product and a market with significant demand, seems like eventually the customers will come - and it seems like this is possible if Intel can retain manufacturing leadership for a few years and build out a better ecosystem to support external chip development by foundry customers, but this of course take time.
Intel hiring practices under Gelsinger were really puzzling. Did he think that by hiring thousands of random university graduates there will be the next Gordon Moore among them? It was also discussed on HN: https://news.ycombinator.com/item?id=42300028
I can kind of understand the extensive poaching of TSMC engineers in Arizona. Though making TSMC upset is playing with fire as much of Intel's business now depends on them.
> Can't win over AMD, Nvidia, Apple, Broadcom as fab customers
Not surprising, and arguably not to be expected. His 5 nodes in 4 years plan ends in 2025, and only then is the fab "good enough" for external customers.
Specifically, they want to see how well Intel's own product on 18A performs before committing. So the real test is in 2025-2026.
> Fab cancellations and delays, wasting a ton of money
The cancellations and delays are on the expanded fab roadmap that he introduced. Compared to before he joined, the fabs are in a much better place. Essentially, this is complaining that he committed to opening 5 new fabs and only opened 3 (or whatever the number is).
> Not hiring the right people to steer Intel's internal fab culture to external quick enough
Perhaps. 5 nodes in 4 years, which Intel may actually achieve, is unprecedented, though. No other company has done it (and likely never will again - it's crazy expensive). So at least as far as execution goes, the fab is in much better shape than when he joined.
> Not simplifying product roadmap.
Definitely agree.
> Didn't stop paying dividends until August 2024
They cut them significantly in early 2023. Yes, they should have done it all earlier, but keep in mind that the board needs to approve it. Or at least, not fire him for doing it. On the plus, he did stop stock buybacks almost immediately after joining.
> Didn't see that Intel was swimming naked during the covid boom and that after the boom, Intel would be in huge trouble due to inferior products
I'm sure he knew the products were inferior. What his failure was is to see that the overall PC market would drop significantly. As an example, in one year it dropped globally 25%. Intel's share dropped 33%, with AMD picking up the remaining 8%. That extra 8% for AMD was something like a 30-50% increase in sales.
> He hired 20% more workers since he joined but 54% less revenue
Totally agreed. What's worse, after the 5% cuts in 2022, Intel was back at the same headcount before the current layoffs. How do you lay off 5% of the company, claiming you need to save money and in 18 months grown the company to the same level?
If you want to see his positives, look here:
https://news.ycombinator.com/item?id=42336143
I think the employee count blowup was his biggest mistake because it shortened Intel's runway towards being an external fab company.
Lip Bu Tan, who was on the Intel board, recently resigned because he did not agree with Pat's employee count plan and company culture. Tan was likely right.
Another mistake of Pat's is that he did not split the company fully. AMD, Nvidia, Apple, Qualcomm aren't voluntarily going to fund a competitor by using Intel fabs. Furthermore, Intel not splitting means there will always be small conflicts of interests with Intel designs teams. Pat wanted the cake and eat it too. He wanted to compete with TSMC in addition to Nvidia, Apple, AMD, Qualcomm. Now Intel is at risk with competing against neither group.I think not having any major external customer signed onto 18A in 2024, approaching 2025, is a huge failure on Pat's part.
It was a mistake, but in terms of runway, would not have made a large difference. The capital costs of running a fab dwarf the cost to pay salaries. That's why they still do fabs in the US - outsourcing to another country doesn't save much money.
> Another mistake of Pat's is that he did not split the company fully.
I don't know that anyone knows what his plans were on this. Internally within the company (and perhaps externally), he was clear on a few things:
- If Foundry cannot become profitable, he'll shut the fabs down and kill Foundry altogether.
- The Products side will be charged market rates, and they can choose if they want to use Intel's fabs or TSMC
- If an external company is willing to pay more than Intel Products, they'll prioritize the external company.
So he was trying to set it up so that it was effectively two separate companies. I guess it was either harder than expected, or there was significant pushback. When he joined, the two sides were heavily intertwined. His goal for 2023 was to get things separated enough so we could get separate P&L statements for Foundry and Products. He achieved it - but I don't know if it could have been done quicker.
The other (obvious) reason he didn't want to break the company up: Intel Foundry is unprofitable. They need the revenue that the Products brings in to keep it alive. The goal was to hit profitability in 2027-2028, when there are enough customers. Splitting it up now will kill Foundry altogether (and now with the CHIPS act, they are not allowed to kill Foundry).
As for the concern about funding competition - not sure how much that is true. The vast majority of TSMC's customers (including AMD) do not want to use the leading edge node. It's more expensive, and has higher quality problems. They are happy to let Apple take it. Likewise, the narrative was that most of these companies will prefer not using Intel's latest node, and would be happy to let Intel use that capacity. So I think the plan was to make 18A mature, move to 14, and then external customers will use the mature 18A node.
How would this be better for the industry?
>It’d be more impressive if these predictions were shared beforehand
It’s hard to disagree with that.
Based on your comment above, it sound like “the receipts” are messages between you and one other person. Messaging a prediction to a friend is very different than publishing it for the world to see.
Some people have been saying this stuff for years.