Their competitors fabs are already open, and better.
Every design Intel has ever done has had significant flaws. The basis of their last 20 years of chips was taken from AMD, then they put their own stink on it and made the worlds' computers vulnerable in many ways.
Weather the storm for the next few years and then keep going?
Consider that Zen took about 5 years to do, when AMD were in a much weaker position than Intel are now. Intel's process problems are harder to deal with but they're not insurmountable as long as they get ahead for the next generation (Which they definitely have the cash to do).
Intel aren't anywhere near as doomed as commenters would have you believe on here or Reddit. Ultimately this is down to whether their leadership keep their heads up or are cynical.
Apple M1 isn't just a shot across Intel's bow. It is giant crater in the middle of their product portfolio. AWS, Azure and Google Compute are absolutely going to pursue their own equivalents.
Microsoft x86s deep enterprise legacy is going to take time to be replaced. But product replacement cycles are pretty short. Microsoft could easily move a lot of their customers to ARM if they wanted to.
If I were Andy Grove, I would have bought a controlling interest in Microsoft years ago. But that is water under the bridge.
"Weather the storm" implies that there is some short / medium term disruption and if they just work a bit harder then they will recover (most of their) their previous competitive position. Trouble is AMD and TSMC aren't going away or stop competing and at the moment there is no evidence that Intel are going to 'fix' their process issues.
Plus,as you know, they have a serious strategic challenge to their business in the rise of firms like Apple and AWS developing their own Arm SoCs.
Intel will probably never have a run like they did last this last decade, but they can still carve a huge chunk of what will always be a huge market.
Also, one thing worth saying is that if ARM really takes a proper share of the market it will obviously make x86 less popular - this works in both ways because it also clears the way for Intel to try other things of their own (This is highly speculative obviously, but it's impossible).
The canary for Arm is if it's in the hands of developers (and not just M1).
Completely agree on the first point - x86 will be around for a long time - but I do think they need to get it onto a cutting edge process, and that means TSMC and quickly.
I don't know much about chip architecture but could they buy a an existing RISC architecture like PowerPC and use that as a competitor to ARM? Allow licensing of PowerPC for SOC and provide the fab as well.
Spend 50 billion trying to get their fab up to TSMC level, probably contract out this fab like TSMC does, and license x86 so that google and amazon hopefully don't migrate the cloud to ARM.
Somebody at Intel needs to play some serious hardball with market competitors and an internally ossified product development process. INTC has a deep patent portfolio, incredible talent, fab capability and market access. Even moreso, it has capital market access. INTC could raise billions and billions for any endeavor in a short period of time.
There are any number of scenarios and playbooks certainly gamed out by investment bankers over the years that Intel could pursue.
What is lacking is a deep competitive drive driven by existential paranoia.
> The entire cost of Fab 18 will cost TSMC $17.08 billion.
Mind you, that's the 5nm production facility which is now complete and churning pretty good CPUs for Apple. They are also building a 3nm fab for 19.6 billion USD. Source https://www.taiwannews.com.tw/en/news/3805032
To leapfrog TSMC, we are quite probably looking at something like 40-50 billion USD in R&D and fab costs and that's a gambit because they are now behind. It's one hell of a gambit for a 200B market cap company.
I always had trouble finding wafer starts per months for Intel and TSMC both, if anyone have such data please share -- nonetheless, it seems like the demand for TSMC far outstrips supply and so there is plenty of room for Intel yet. In the worst case... there's a lot of money to be made as #2 chipmaker even if it's less than they are making as #1 chipmaker?
I am not sure I have seen evidence for what you say. That is a complicated assertion because it speaks to wafer's started, wafer's finished, yield, die size, and technology among other things.
Honestly, I suspect that Intel's struggles (as per typical) with new technologies are around ossification of technical paths, struggles with yield, and copy-stupid. That leaves them struggling to produce the quantity that customers need, and has opened avenues for customers to find other routes that provide them with more control. They are also seeing the benefits of custom silicon and the possibilities of full stack optimization, as opposed to optimization of just one or two parts which largely drove older approaches to software/hardware/battery efficiency.
My other comment is that SPENDING on R&D is not always the best metric...
17 billion $ is a joke the Intel share buyback program is 20 billion $....
instead of buying their own share they can do that without raising or borrowing a single cent.
But ... this article is a sign that investors are losing patience and so they might not have a free hand. It might seem straightforward to cancel buybacks but if you lose your investors then you might find its someone else implementing a new strategy.
I’m not sure how this split would go, since Intel’s disadvantage is their poor fab performance and technology compared to TSMC - would Apple buy M1 from Intel onshore fab at 14nm? No. Would they buy at 10nm? No. They want the best - 5nm.
I am interested in seeing what Intel’s designers could achieve on TSMC 5nm, but that’s hardly the idea behind the deal here to secure on-shore chip manufacturing
No, what you do is spin out the fab business understanding that it's going to hit a brick wall once it's on its own, freeing up the rest of the company to start purchasing capacity from TSMC without internal political issues.
>> What surprises me is the lack of urgency on the part of the US government. A lack of competitive chip manufacturing in the US is a threat.
> My bet is they consider Taiwan, Japan, and South Korea well within their sphere of influence (and control), so no urgency there.
That's pretty foolish and short-sighted, though, since all of those countries are within spitting distance of the US's main geopolitical competitor, and that competitor even has a disputed territorial claim to one of them (and vows to retake it by force, if necessary).
I'd consider the area pretty fortified. The US has military bases in both South Korea and Japan, and they just sold billions of dollars worth of cutting edge weapons to Taiwan.
But not invincibly so. You usually want your defense industries located in the rear, not near your front lines. The US relying on critical defense suppliers in Taiwan is like the Soviets doing the same in Cuba (which they didn't do).
Also, IIRC, the China team has won every recent US war game simulation of a shooting war in the Taiwan Straight. Also the Taiwanese don't get the really good US weapons, because of legitimate fears of Chinese espionage.
TSMC has proven that the fab model works well, and Apple has proven a design model works well.
Intel could sell their design group to MS and open fabs to all. The u.s govt can require second source (eg spacex and ULA, envelope though the latter is more expensive) to help fund intels EUV transition.
I don't see what options Intel has to dramatically increase their future profits and thus market capitalization, even if an activist gives them a kick up the butt.
Option 1: Strip mine the company. This is more of the traditional PE/activist play. Once you stop spending all those R&D dollars, then Intel could extract a lot of profits from the x86 over the next decade or so. But this is only a short-medium term play. All mines run out of gold eventually.
Option 2: Split the Fab and design (x86) business. I'm not sure this helps that much, I think x86 is F'ed even if it could use the TSMC fabs. This is what AMD is already doing. The major important players (Apple, Amazon, MS, Goog) want to go ARM and then maybe RISC-V for their own strategic reasons anyway.
Option 3: Spend unheard of amounts of money to try and win back a process advantage over TSMC. This would entail spending tens of billions of dollars to basically make a bet that might pay off. It would be fun to watch, but success is not guaranteed, or even likely.
"AWS Graviton processors are custom built by Amazon Web Services"
and
"Apple Silicon"
It's not just about x86 -> Arm, it's about bringing more control of the design in house. This only gets more important as we move towards more SoC type designs. Imagine, for instance, AWS doing an entire SoC with Nitro, disc controllers, memory etc on one chip. I'm not sure that they could, or would want to do such a thing with Intel and x86.
Option 4: spend some of their massive amounts of money to rebuild their R&D capabilities. Simultaneously give up their outdated idea that their core architecture is so valuable and secret that it must never be fabbed elsewhere. And spend the money to keep up with fab technology. And possibly put an actual engineer at the helm. Right now Intel seems to be going the way of Boeing — plenty of money, but less innovation than they ought to have.)
Keeping fab and design under one roof does not require that the fab and the design sides of the business exclusively with with each other.
(I think Intel should keep Altera, too. A paired CPU and FPGA solution has genuine value. One thing that Intel appears to still be good at is software tooling. If Intel were to provide tooling for their FPGAs that was anywhere near as good as their x86 tooling, then Altera could take off as everyone’s favorite platform. Right now they’re at substantial risk of losing new developer mindshare to Lattice because of the IceStorm project. I have trouble imagining that more than a few $M of total resources have been spent there.
As it stands, nVidia is a major player in big chips that do serious computation. Whatever one may think about CUDA, it’s in a whole different league from the garbage currently available from Xilinx and Altera.)
Crazy idea: Why doesn't Intel launch their own cloud IaaS offering?
Value proposition: Intel needs to scale manufacturing and R&D, in a way that doesn't leave them in a performance bind (ie without performance, they can't sell at scale, and thus can't afford to buy performance). Abstracting compute behind an IaaS offering would let them do exactly that while providing an alternate sink for binned chips.
Furthermore, they've got decent HW/SW integration talent in-house, and they could offer a key feature other clouds couldn't: vendor agnosticism. Building {PaaS product} on an Intel cloud seems strategically smarter than building on one of the continually-vertically-integrating clouds.
I think an IaaS-only cloud is a losing proposition. Customers want a variety of services at different levels on the same account. There may be room for innovation in federation/reselling higher-layer services but I'm not sure.
Also, Intel-only would be limited compared to other clouds offering Intel/AMD/ARM.
It's more likely that Intel will increase market development funding to the existing cloud providers as their market share declines.
Is x86 likely to stay as the dominant cloud architecture though? My understanding of ARM on AWS is that if your stuff supports it then it’s better performance per dollar and basically a no brainer. I don’t see why x86 would continue to be the dominant deployment target long term, especially with Apple moving their line to ARM meaning that developers will increasingly demand compatibility with it across the stack.
It seems like Intel could create a pretty compelling offering. Especially since AWS, Azure, and GCP all seem to be giving them the strategic middle finger.
1. As IBM, HP and others discovered, the expertise required to run cloud services is very different from what's required to sell servers (or CPUs).
2. To compete with Amazon and MS, you would need to spend several tens of billions of dollars just to get a similar infrastructure footprint.
A variation on your proposal would be for Intel to buy a bunch of smaller cloud players, like DigitalOcean, and CloudFlare. Intel has the huge amounts of cash that these companies need to compete with AWS and Azure, Intel would then be vertically integrated, just like AWS is, and they would get the customer base and expertise from those acquisitions.
I think Intel's core offering would be value-add neutrality, which is attractive compared to building on (and competing with) first party AWS/GCP.
DigitalOcean, CloudFlare, etc would definitely have to be involved... but it's a tricky balancing act to avoid their becoming dominant solutions, at which point you would have AWS, except as a combination of companies.
Not sure how to square that circle. Maybe an extension of the Open Compute group + FRAND licensing of Apache-style preferred value add solutions on top, with Intel operating the hardware as a certified implementation? Essentially enabling MVNOs for cloud.
Intel's key distinction from foundry-less competitors, is that they can scale capacity to whatever they want, at least on older nodes. And the more volume they scale to, the more they ammortize their own costs (ie fab version of Amazon's AWS rationale).
The key would be establishing an ecosystem where innovation at the higher levels, and FRAND licensing of same, financially rewarded and incentivized players to invest in the platform.
If China encroaches on Taiwan, the US is economically obligated to intervene; losing access to TSMC would be a massive hit. (It seems that TSMC is getting increasingly-advanced over their competitors, but maybe that's just my media blindspot.)
I expect the federal government to take action here, probably by giving huge monetary incentives to Intel/AMD for creating a state-of-the-art fab in the US.
Disclaimer: I work for a competitor to Intel in one of Intel's many markets. I would indirectly stand to gain from Intel spinning out their fabs.
I would love to see Intel open up their fabs. More fabs in more locations would be beneficial to the chip world overall. However, I don't know if Intel can do it successfully.
The vertically integrated strategy pays off when you have enough volume in the money maker (CPUs) to justify the heavy investment in the bottom of the stack (fabs). Intel is nearly at a threshold where they risk losing hold of the CPU market.
I see about 4 paths. I don't know which one we're closest to.
1. The ideal future for everyone (except for their their competitors in the manufacturing space) is that Intel successfully spins out their fabs and we have a viable near-leading process available for the chip industry in the large. There's still a lot of money in not-quite-bleeding-edge processes.
2. There's a future in which Intel doesn't, they bounce back, and everyone calling for them to spin out their fabs looks like an idiot.
3. There's also a future in which Intel doesn't, they don't bounce back, and their process falls into irrelevance as they lose the volume to fund their fab development.
4. Then there's a future in which Intel does, they do or don't bounce back, and Intel's fabs are still not competitive enough to be used by anyone but Intel. It's still an open question whether Intel can reach the volume they need to make this pay off, for example. It's also not obvious that Intel's process, on a technical level, is ideal for anyone but Intel.
The ideal time for this to have happened was years ago when Intel still had a process advantage and could leverage that to bring in new customers despite the other risks, but it was precisely because they had a process advantage that they weren't going to do it. In fact, they went the other direction and let go of much of their custom foundry expertise, which was quickly scooped up by (now) more successful companies, and purchased firms that moved to the Intel process rather than maintain a customer relationship. Now they have to contend with the fact that they don't have an advantage and they aren't as accustomed to working with partners for manufacturing. It's an uphill battle, but it's doable. The geopolitics might be compelling enough that they can overcome this with pressure and funding from three-letter agencies.
Culturally, this might be better regardless of any of the technical details. I've often heard that manufacturing runs the ship at Intel and that they suffer from severe internal politics between orgs. Splitting up might be the right tool to reign all of those problems in. But, it might also be that Intel truly has been infected with the managerial MBA rot that technical folks like me like to complain about, or Intel has irreversibly shed its manufacturing talent. In that case, I don't know if manufacturing can be saved.
You're using two different bits of language, "open up their fabs" and "spin out their fabs". "Open up" implies Intel maintaining ownership but contracting out fab capacity to other chip makers like TSMC. "Spin out" implies divesting their fabs like AMD did with Global Foundries. Might help to clarify which one you mean.
Thanks for pointing that out. I think closing the door on ICF and doubling down on vertical integration is a mistake on Intel's part (but likely forced because of their production issues), but the extent to which Intel should open up is a question likely only Intel employees knee deep in the R&D can answer since it comes down to how integrated manufacturing is with the rest of the company. There's a continuum from ICF to Samsung-style conglomerate to GloFo, with different kinds of customers at each point. Any point on that spectrum is a less risky place to be than Intel relying on Intel alone.
All told, I'd rather be Intel today than, say, SMIC today, and I'd much rather be the Intel of today than the AMD of 8-12 years ago. Intel is down, but in no way out, and I find most of the commentary hyperbolic. A full GloFo-style spin out might be inevitable, but I think Intel is sufficiently positioned to take it more slowly than AMD did.
Intel's advantage..... MASSIVE advantage over ARM is that ARM chips tend to have secrets and parts that are undocumented, and don't have the compatibility of Intel.
A random Intel device probably runs the software you want.
A random ARM device almost certainly doesn't run the software you want
There are now 4 large companies that design both CPUs and GPUs. Intel, AMD, nVidia (owns ARM) and Apple.
As the RISC-V thing was getting going a few years ago I thought the commoditization of CPUs would make GPUs the next interesting IP. It seems the next thing is AI acceleration with Apple and nVidia ahead AFAICT.
Meanwhile RISC-V V is nearing completion and decent chips are starting to become available.
I dont see anything but fierce competition for Intel for the next 5 years or more.
> Intel customers, such as Apple Inc, Microsoft Corp and Amazon.com Inc, are developing their own in-house silicon solutions and sending those designs to be manufactured in East Asia, Loeb wrote.
This is the key point right here, and one our Government should be worried about. It's a national security issue if we aren't able to make our own chips.
It's a tough choice, because in the name of free trade you want to allow companies to build wherever makes the most sense cost-wise, but in the name of national security, you want to retain that critical skill here.
This has been a while coming, there were significant issues at 14nm and some good talent left. Since then some really interesting people have come and gone, most notably of course BK finally got pushed out the door. By the point he was gone though they were truly neck deep in the mud. Having failed to get a CEO who would actually take the company on they now have Bob Swan (former CFO) which gets any Intel employees' heads twitching since the HR department spent the last decade telling anyone who would listen that BK was in charge because Intel is a manufacturing company.
Intel have now burned an enormous amount of capital through failed acquisitions (Mccafee, Altera, Nervana etc) - mainly because their attitude was "Oh we can do that too, and that's a really big market". They can't make FPGAs better than an FPGA company, they can't make security software better than a software company, they can't make AI chips better than a company that does AI.
What would make them successful again is to focus on their core business (a business they departed because the growth wasn't there). So that means spinning out fab (which let's be fair, will probably die), spinning out Networking, I was going to say spinning out FPGA but I just went and looked up the head of that group and he works at AMD now (unlucky for AMD). Spin out the data centre group. Spin out the IOTG group - who can then operate on reasonable scales rather than refusing to work on anything smaller than 100 million units. Spin out MobilEye (they fucked up the BMW project that got them acquired in the first place). Then what you're left with in Intel proper is just Martin from HR, and he was useless anyway.
65 comments
[ 0.20 ms ] story [ 123 ms ] thread* Offer co-design services to Amazon, Microsoft, Apple, Google
Every design Intel has ever done has had significant flaws. The basis of their last 20 years of chips was taken from AMD, then they put their own stink on it and made the worlds' computers vulnerable in many ways.
Consider that Zen took about 5 years to do, when AMD were in a much weaker position than Intel are now. Intel's process problems are harder to deal with but they're not insurmountable as long as they get ahead for the next generation (Which they definitely have the cash to do).
Intel aren't anywhere near as doomed as commenters would have you believe on here or Reddit. Ultimately this is down to whether their leadership keep their heads up or are cynical.
Microsoft x86s deep enterprise legacy is going to take time to be replaced. But product replacement cycles are pretty short. Microsoft could easily move a lot of their customers to ARM if they wanted to.
If I were Andy Grove, I would have bought a controlling interest in Microsoft years ago. But that is water under the bridge.
Plus,as you know, they have a serious strategic challenge to their business in the rise of firms like Apple and AWS developing their own Arm SoCs.
Also, one thing worth saying is that if ARM really takes a proper share of the market it will obviously make x86 less popular - this works in both ways because it also clears the way for Intel to try other things of their own (This is highly speculative obviously, but it's impossible).
The canary for Arm is if it's in the hands of developers (and not just M1).
There are any number of scenarios and playbooks certainly gamed out by investment bankers over the years that Intel could pursue.
What is lacking is a deep competitive drive driven by existential paranoia.
> The entire cost of Fab 18 will cost TSMC $17.08 billion.
Mind you, that's the 5nm production facility which is now complete and churning pretty good CPUs for Apple. They are also building a 3nm fab for 19.6 billion USD. Source https://www.taiwannews.com.tw/en/news/3805032
To leapfrog TSMC, we are quite probably looking at something like 40-50 billion USD in R&D and fab costs and that's a gambit because they are now behind. It's one hell of a gambit for a 200B market cap company.
I am not sure I have seen evidence for what you say. That is a complicated assertion because it speaks to wafer's started, wafer's finished, yield, die size, and technology among other things.
Honestly, I suspect that Intel's struggles (as per typical) with new technologies are around ossification of technical paths, struggles with yield, and copy-stupid. That leaves them struggling to produce the quantity that customers need, and has opened avenues for customers to find other routes that provide them with more control. They are also seeing the benefits of custom silicon and the possibilities of full stack optimization, as opposed to optimization of just one or two parts which largely drove older approaches to software/hardware/battery efficiency.
My other comment is that SPENDING on R&D is not always the best metric...
I am interested in seeing what Intel’s designers could achieve on TSMC 5nm, but that’s hardly the idea behind the deal here to secure on-shore chip manufacturing
> My bet is they consider Taiwan, Japan, and South Korea well within their sphere of influence (and control), so no urgency there.
That's pretty foolish and short-sighted, though, since all of those countries are within spitting distance of the US's main geopolitical competitor, and that competitor even has a disputed territorial claim to one of them (and vows to retake it by force, if necessary).
Yeah, but from the US perspective, it's the weakest zone of that protection. So they'll be firmly beholden until suddenly they're not.
Also, IIRC, the China team has won every recent US war game simulation of a shooting war in the Taiwan Straight. Also the Taiwanese don't get the really good US weapons, because of legitimate fears of Chinese espionage.
TSMC has proven that the fab model works well, and Apple has proven a design model works well.
Intel could sell their design group to MS and open fabs to all. The u.s govt can require second source (eg spacex and ULA, envelope though the latter is more expensive) to help fund intels EUV transition.
Option 1: Strip mine the company. This is more of the traditional PE/activist play. Once you stop spending all those R&D dollars, then Intel could extract a lot of profits from the x86 over the next decade or so. But this is only a short-medium term play. All mines run out of gold eventually.
Option 2: Split the Fab and design (x86) business. I'm not sure this helps that much, I think x86 is F'ed even if it could use the TSMC fabs. This is what AMD is already doing. The major important players (Apple, Amazon, MS, Goog) want to go ARM and then maybe RISC-V for their own strategic reasons anyway.
Option 3: Spend unheard of amounts of money to try and win back a process advantage over TSMC. This would entail spending tens of billions of dollars to basically make a bet that might pay off. It would be fun to watch, but success is not guaranteed, or even likely.
Is there an option 4?
Source? I don't think they care, I think they just don't want to be beholden to a single supplier, like they were to Intel for a while.
"AWS Graviton processors are custom built by Amazon Web Services"
and
"Apple Silicon"
It's not just about x86 -> Arm, it's about bringing more control of the design in house. This only gets more important as we move towards more SoC type designs. Imagine, for instance, AWS doing an entire SoC with Nitro, disc controllers, memory etc on one chip. I'm not sure that they could, or would want to do such a thing with Intel and x86.
Edit: Adding my sources ;)
Graviton quote: https://aws.amazon.com/ec2/graviton/
Apple silicon quote: https://www.apple.com/mac/m1/ (for fun, try to CTRL+F this page for "arm")
Keeping fab and design under one roof does not require that the fab and the design sides of the business exclusively with with each other.
(I think Intel should keep Altera, too. A paired CPU and FPGA solution has genuine value. One thing that Intel appears to still be good at is software tooling. If Intel were to provide tooling for their FPGAs that was anywhere near as good as their x86 tooling, then Altera could take off as everyone’s favorite platform. Right now they’re at substantial risk of losing new developer mindshare to Lattice because of the IceStorm project. I have trouble imagining that more than a few $M of total resources have been spent there.
As it stands, nVidia is a major player in big chips that do serious computation. Whatever one may think about CUDA, it’s in a whole different league from the garbage currently available from Xilinx and Altera.)
1. Mobile (I include cars, IoT etc)
2. Server/cloud (most new servers will be deployed to AWS, MS, Google, and maybe Alibaba)
3. Graphics / ML (you mentioned Nvidia and CUDA)
How would your suggested option 4 help Intel gain a significant foothold in any of those markets? (or stop the Arm transition in the case of #2)
Value proposition: Intel needs to scale manufacturing and R&D, in a way that doesn't leave them in a performance bind (ie without performance, they can't sell at scale, and thus can't afford to buy performance). Abstracting compute behind an IaaS offering would let them do exactly that while providing an alternate sink for binned chips.
Furthermore, they've got decent HW/SW integration talent in-house, and they could offer a key feature other clouds couldn't: vendor agnosticism. Building {PaaS product} on an Intel cloud seems strategically smarter than building on one of the continually-vertically-integrating clouds.
Also, Intel-only would be limited compared to other clouds offering Intel/AMD/ARM.
It's more likely that Intel will increase market development funding to the existing cloud providers as their market share declines.
It seems like Intel could create a pretty compelling offering. Especially since AWS, Azure, and GCP all seem to be giving them the strategic middle finger.
My bear case for Intel Web Services would be:
1. As IBM, HP and others discovered, the expertise required to run cloud services is very different from what's required to sell servers (or CPUs). 2. To compete with Amazon and MS, you would need to spend several tens of billions of dollars just to get a similar infrastructure footprint.
A variation on your proposal would be for Intel to buy a bunch of smaller cloud players, like DigitalOcean, and CloudFlare. Intel has the huge amounts of cash that these companies need to compete with AWS and Azure, Intel would then be vertically integrated, just like AWS is, and they would get the customer base and expertise from those acquisitions.
DigitalOcean, CloudFlare, etc would definitely have to be involved... but it's a tricky balancing act to avoid their becoming dominant solutions, at which point you would have AWS, except as a combination of companies.
Not sure how to square that circle. Maybe an extension of the Open Compute group + FRAND licensing of Apache-style preferred value add solutions on top, with Intel operating the hardware as a certified implementation? Essentially enabling MVNOs for cloud.
Intel's key distinction from foundry-less competitors, is that they can scale capacity to whatever they want, at least on older nodes. And the more volume they scale to, the more they ammortize their own costs (ie fab version of Amazon's AWS rationale).
The key would be establishing an ecosystem where innovation at the higher levels, and FRAND licensing of same, financially rewarded and incentivized players to invest in the platform.
Their fabs need more customers to justify the astronomical investment. They must expand their foundry business.
I expect the federal government to take action here, probably by giving huge monetary incentives to Intel/AMD for creating a state-of-the-art fab in the US.
TMSC is planning to build a fab factory in the US. https://www.forbes.com/sites/willyshih/2020/05/15/tsmcs-anno...
I think this is partly because they can easily get a license to use ARM or RISC-V in custom chips, but doing that with x86 is pretty much impossible.
I would love to see Intel open up their fabs. More fabs in more locations would be beneficial to the chip world overall. However, I don't know if Intel can do it successfully.
The vertically integrated strategy pays off when you have enough volume in the money maker (CPUs) to justify the heavy investment in the bottom of the stack (fabs). Intel is nearly at a threshold where they risk losing hold of the CPU market.
I see about 4 paths. I don't know which one we're closest to.
1. The ideal future for everyone (except for their their competitors in the manufacturing space) is that Intel successfully spins out their fabs and we have a viable near-leading process available for the chip industry in the large. There's still a lot of money in not-quite-bleeding-edge processes.
2. There's a future in which Intel doesn't, they bounce back, and everyone calling for them to spin out their fabs looks like an idiot.
3. There's also a future in which Intel doesn't, they don't bounce back, and their process falls into irrelevance as they lose the volume to fund their fab development.
4. Then there's a future in which Intel does, they do or don't bounce back, and Intel's fabs are still not competitive enough to be used by anyone but Intel. It's still an open question whether Intel can reach the volume they need to make this pay off, for example. It's also not obvious that Intel's process, on a technical level, is ideal for anyone but Intel.
The ideal time for this to have happened was years ago when Intel still had a process advantage and could leverage that to bring in new customers despite the other risks, but it was precisely because they had a process advantage that they weren't going to do it. In fact, they went the other direction and let go of much of their custom foundry expertise, which was quickly scooped up by (now) more successful companies, and purchased firms that moved to the Intel process rather than maintain a customer relationship. Now they have to contend with the fact that they don't have an advantage and they aren't as accustomed to working with partners for manufacturing. It's an uphill battle, but it's doable. The geopolitics might be compelling enough that they can overcome this with pressure and funding from three-letter agencies.
Culturally, this might be better regardless of any of the technical details. I've often heard that manufacturing runs the ship at Intel and that they suffer from severe internal politics between orgs. Splitting up might be the right tool to reign all of those problems in. But, it might also be that Intel truly has been infected with the managerial MBA rot that technical folks like me like to complain about, or Intel has irreversibly shed its manufacturing talent. In that case, I don't know if manufacturing can be saved.
All told, I'd rather be Intel today than, say, SMIC today, and I'd much rather be the Intel of today than the AMD of 8-12 years ago. Intel is down, but in no way out, and I find most of the commentary hyperbolic. A full GloFo-style spin out might be inevitable, but I think Intel is sufficiently positioned to take it more slowly than AMD did.
A random Intel device probably runs the software you want.
A random ARM device almost certainly doesn't run the software you want
Intel is failing to play this advantage.
As the RISC-V thing was getting going a few years ago I thought the commoditization of CPUs would make GPUs the next interesting IP. It seems the next thing is AI acceleration with Apple and nVidia ahead AFAICT.
Meanwhile RISC-V V is nearing completion and decent chips are starting to become available.
I dont see anything but fierce competition for Intel for the next 5 years or more.
1. They need to fend off AMD in x86 and in the absense of a process solution that means TSMC.
2. Opening up fabs would maximise the value of manufacturing and support ongoing investment as x86 declines.
3. Sell everything that is non-core and a distraction.
4. Fix the culture.
Note that 1. and 2. don't necessarily mean splitting the company.
Not simple though - especially 4. (easy from the armchair though!)
This is the key point right here, and one our Government should be worried about. It's a national security issue if we aren't able to make our own chips.
It's a tough choice, because in the name of free trade you want to allow companies to build wherever makes the most sense cost-wise, but in the name of national security, you want to retain that critical skill here.
It's a tough balancing act.
Intel have now burned an enormous amount of capital through failed acquisitions (Mccafee, Altera, Nervana etc) - mainly because their attitude was "Oh we can do that too, and that's a really big market". They can't make FPGAs better than an FPGA company, they can't make security software better than a software company, they can't make AI chips better than a company that does AI.
What would make them successful again is to focus on their core business (a business they departed because the growth wasn't there). So that means spinning out fab (which let's be fair, will probably die), spinning out Networking, I was going to say spinning out FPGA but I just went and looked up the head of that group and he works at AMD now (unlucky for AMD). Spin out the data centre group. Spin out the IOTG group - who can then operate on reasonable scales rather than refusing to work on anything smaller than 100 million units. Spin out MobilEye (they fucked up the BMW project that got them acquired in the first place). Then what you're left with in Intel proper is just Martin from HR, and he was useless anyway.