Their competitors eat them. I would not be surprised to see Oracle's cloud business get absorbed by IBM or Microsoft. Maybe Amazon. The extra DC capacity is valuable to a couple companies right now.
Is it me or do none of the AI companies have a "moat" in the Ben Grahmm sense.
I use their services, but I frankly don't care who provides it. I'll chase the chepest/best and have no issue switching from one to another.
The only moat I can see is Microsoft providing its services to companies in its Azure system. Nervous IT departments probably like that it's not leaving their control if Bob in the SAP team spins up some AI crap.
> I use their services, but I frankly don't care who provides it. I'll chase the chepest/best and have no issue switching from one to another.
For the hyperscalers, there is an ease of remaining in the Azure/AWS/GCP fabric from a data provenance perspective, particularly for regulated industries or large, risk-averse enterprises. There's also, of course, a certain network egress tax in most cases.
I thought that Nvidia's moat was more in CUDA? Hardware is hard but we've already seen other companies like Google design neural processors with compute efficiency close to Nvidia.
I am about to spend $20M, if I buy anything other than Nvidia, and things go wrong, I am going to get blamed, and if things go right I will get no credit. This is why AMD is making no progress outside of very narrow cases and supercomputing.
I've been thinking for a while, there's not real winners here except the incumbent technology providers. Hear me out: all models are converging towards the same level, gains are getting smaller and harder to come by. The models are commodities nothing more.
This is the leap, nobody really wants to front a model for someone else. If i build an agent, or a service that requires a model, I'd prefer to push the model onto someone else, preferably at no cost. This is a leap as I'm sure right now, most people / businesses are thinking actually i do want to own / front the model.
However, if you accept the leap the easiest way to do this is to make the model the users problem.
From a business point of view that makes things really easy, from a customer point of view, they simply have to accept whatever their vendor of choice is pushing down their throats.
So as a business I build for whatever model Google makes available to android, and whatever model windows bundles, and whatever model Apple bundles, and, excluding the long tail of Chinese vendors and Linux (sorry, its always left out) and that's it, problem solved, and the customer picks up the tab for the tokens
The adoption of standards like skills and agent setup helps a ton. Nobody wants to be locked into an AI vendor like with cloud systems in general. And companies can't hold on to the #1 spot across multiple areas for very long, so users are even more motivated to move their process and stack between coding tools and AI companies behind them like Claude code.
The moat is shifting from technology to access to proprietary training data. It doesn't matter how good your LLM platform is if you don't have good data to feed the training run. Public Internet data and published media is already mined out. Now the frontier LLM vendors have shifted to licensing proprietary data that's locked up behind corporate firewalls, and even hiring human domain experts specifically to create new training content in target verticals. You'll see the effects of this next year, although it might not be obvious to those who mostly only use LLMs for coding tasks in popular programming languages for which there was already a lot of training data.
> Now the frontier LLM vendors have shifted to licensing proprietary data that's locked up behind corporate firewalls, and even hiring human domain experts specifically to create new training content in target verticals.
That's a losing proposition for any token provider - it's expensive and slow, and when you're done everyone with money to rent a last-gen H100 is going to distill your "closed" model anyway.
> That's a losing proposition for any token provider
The specialized models for targeted verticals being discussed may well not be sold by tokens, but instead be behind the scenes powering dedicated packaged solutions where the customers don't have raw access to the model. Token providers still won’t have a moat, but AI isn't just selling tokens.
I think anthropic with its enterprise strategy and google
with its integration in everything have a bit of a moat.
But I switched from ChatGPT to Claude 3 months ago because my account was down for like 6 hours. I haven’t used it since. It’s too easy to switch away from chatbots on a whim. There is no moat for that.
I guess I’m thinking a lot of companies seem to be getting Claude code subscriptions. It usually takes some time and effort for an org to switch away from one solution. In the meantime a lot of workflows get more and more tied to Claude in particular.
It’s not much of a moat, but it’s more than a lot of orgs have.
Observationally, for people that /aren't/ using models to code but to just do their white-collar job, claude.ai /is/ AI, now. The entire perspective for how to use AI is through claude skills, claude projects, claude cowork, etc. They've massively won the corp buy-in at the moment I believe.
obligatory correction: the semiconductor layer is still owned by TSMC and Samsung. Google sketches chip designs for them to implement - that's the lowest layer they control
The narrative that superintelligence is imminent is partially at fault here.
There are competing definitions of what intelligence even is, and the one that I find most striking is from Francois Chollet which is that intelligence can be boiled down to skill acquisition efficiency. This type of definition makes intelligence more akin to polishing a ball than growing a watermelon.
The superintelligence doomers warn that the watermelon is going to start growing exponentially and crush everyone. But what might actually be happening is that we are not growing a watermelon but rather polishing the ball until its really smooth and shiny. There's a point where you can get it to micron levels of polish but for most tasks (white collar text domains tasks), it's smooth enough! You will be able to go to the ball store and buy a low cost made in china ball for most tasks.
The real challenge is actually branching out domains and modalities to tackle things like blue collar labor. Over time, white collar work automatable or able to be made hyperefficient by LLMs will see LLM commoditization.
google might have tons of integration. But if it invested too heavily into AI then it will also suffer when increased competition causes returns to fall:
AWS and Google at least own their own hardware (Trainium and TPUs, respectively). It's a moat in the sense that designing, building, and deploying your own chips at scale is quite a feat and not easily replicated. The vertical integration will allow them to continue to be profitable once the models get good enough and competitors' prices race to the bottom. Google has Gemini; AWS may not deploy its own models (yet?), but that's not necessarily a losing position, as long as the market is able to run models sourced elsewhere on Trainium and the price is right.
Isn't specialized hardware also a big risk? GPUs are more amenable to any big changes that may happen in the next 5, 10 years of AI research. Maybe we won't even be talking about LLMs anymore. Maybe matrix multiplication won't be enough.
Maybe it isn't that far fetched, but I could see them specializing for some super high dimension multiplication and meanwhile 5 years later turns out "all you need" are 3x3 matrices and suddenly 90% of your specialized hardware is now dark silicon :)
You may not care, but a lot of people I know care what brand chat bot they use personally,. usually it's tied to trust and reputation more than anything else. People are fickle.
I've found that there is value in consuming AI services from your existing cloud provider. Customers and auditors have less of an issue with "we use AI services from AWS/Azure/GCP" if the data was already in those clouds and it doesn't expand the risks of data being breached, or trained on, by some other provider.
When you are already trusting 100% of your data, and computing on that data, to someone like AWS, it doesn't meaningfully increase risk to use an additional service, even if it is an AI service.
True. The linked article's title says that. I wonder if that was a typo by the OP or one of those HN quirks where the title was automatically changed when it shouldn't have been.
There is AI data center overcapacity already. The KOSPI crashed last week, and it's a leading indicator for the cyclical hardware industry. It already had been that indicator in the 2000 bubble.
I don't know what possessed Ellison to ruin a functioning company, but it will be interesting if he gets a margin call for ORCL's other debt exposures, which are Ellison's massive loans against his ORCL stock.
The ability of Silicon Valley to hype itself up into a frenzy is unparalleled. Apparently nothing was learned from "blockchain for everything" and "we're going to live in the metaverse".
The KOSPI went up already 125% in the past year, so some sort of correction was inevitable, even if the underlying companies are healthy. The crash has been exacerbated by South Koreans levering up heavily in the past few months and now getting wiped out.
This is surprising to me. Judging by what appears to be the common sentiment here on HN - which is that AI inference is already profitable, and OpenAI is fairly valued by private markets.
Given that Oracle and Microsoft are major counterparties of OpenAI, it seems odd that their stocks have been performing so poorly recently. Can anyone square this circle for me?
Inference might be profitable, but it does not mean the profits of AI datacenters will rise in future. Open weight models and local AI already put the pressure on the AI datacenter profit margins, and local AI is set to become much more efficient in the future.
The general fallacy of the “but inference is profitable” argument is that it tends to ignore all the costs of building and training the model. Given the fact that 1) that’s not trivial, and 2) the arms race underway means one can’t stop training, then it ruins the financial picture.
It’s like saying a new apartment building is “profitable” because the monthly income covers the monthly running costs, but ignoring the giant mortgage that covers the cost of building the building. That thinking is a good way to go bankrupt in real estate and a good way to go bankrupt in AI.
> The general fallacy of the “but inference is profitable” argument is that it tends to ignore all the costs of building and training the model. Given the fact that 1) that’s not trivial, and 2) the arms race underway means one can’t stop training, then it ruins the financial picture.
Or that it’s all hearsay and no one has released financials yet?
xAI financials are public, and OpenAI financials leaked a short while ago.
That's the best possible interpretation of them.
The other possible interpretation is that they are manipulating the numbers (that they have to show to investors) and inference isn't actually profitable either. If they are not manipulating the numbers right now, both companies have a serious case of uncontrolled operational costs that they have to solve too.
> That's the best possible interpretation of them.
Correct. Because a less charitable interpretation will squint at their marketing spend and wonder if they are just sweeping a lot of their expenses on categories they don't belong to make their business look less bonkers to those that prefer to not ask the ugly questions.
“Marketing” typically includes all the people in solutions architecture, developer advocacy, compute credits, and many other things. It’s a ton more than just advertising spend, which is often a minority of what appears on that line item. Given that it’s very plausible those sort of sums are realistic.
If the company who holds the mortgage wanted to own the building, they would have just bought it themselves. They don't, for whatever reason, so to some extent they have an incentive to help their customer succeed.
That's why it's so hard to get a residential mortgage, for example. It's more of a partnership, with more mutual vulnerability, than most people think. Same thing seems to be true here.
Given what happened with xAI’s excess capacity lease to Anthropic, and Meta’s noises about doing the same, seems likely that the demand for inference will continue to slope upwards for a while. If I’m Oracle, I’m not worried about being able to utilize the data centers I’ve built for some price, almost certainly a profitable one.
I’m guessing, though, that Oracle made their capital investments on assumptions of a higher price & return. Possibly because it wasn’t clear when these decisions were made how much competition OpenAI would have at the frontier.
I don’t think this math is all that hard. Capital markets have everything they need to start to figure it out, most especially a year or two of history to project forward.
it is isn't enough for inference to be profitable, the whole organization has to be profitable enough to keep investors from looking elsewhere for a return.
I agree but I don’t want to concede that point. There’s no evidence that inference is profitable which makes the rest of the conversation a non-starter.
When dealing with sophists you have to pin them down on specific points and prevent them from shifting the conversation.
I looked this up yesterday triggered by their threat to move the combined company out of CA. Oracle’s stock price, at least, which is way off its 52-week high, is about the same as it was at the time the WBD deal was announced.
What does that tell me? Just one of many things about the prospects of the deal still happening. That one in particular says to me they won’t be deterred. Bond rating may suggest the opposite. Lots more complexity than those two things but “fun” to speculate.
When we tried to do a pilot with their cloud we couldn't even sign-up. None of the corporate credit cards were accepted.
In addition to that the form basically only worked in Edge. We emailed support, they changed something on the backend. It still did not work. We gave up.
In retrospective that was a very clear warning sign that their priorities were misguided. I'm glad we did not waste any further time and effort on them.
Good to know it's not only problematic on the free tier. I wanted to sign up to get the free credits but couldn't finish the setup. I tried again now and it accepted/charged my card ($1 verification test) but then after the account was created it said I need a credit card?
For the longest time they were a piñata for free compute with people making multiple accounts for their free ARM instance, but with the AI crunch they're clamping down.
I'm guessing they don't care if actual business gets caught up in that because from their POV actual business comes from an account manager, and self-serve is just them cargo culting AWS/GCP
Oracle Cloud sometimes feels like an elaborate prank that I'm not in on. I know people and companies on AWS (obviously), Azure, Google Cloud, Hetzner, CloudFlare's various PaaS offerings, etc., but I can't name a single thing running on Oracle Cloud. Somebody out there is clearly using but I'll be damned if I know who it is.
I think the market for Oracle Cloud is the same for early GCP: companies with large enough needs and strong enough engineering teams that they can leverage "X runs on Oracle Cloud" into deep discounts. And then cover the gaps with engineering.
We host more than 200 customers in OCI (because we have to). Its terrible. The service is terrible and they are breaking stuff all the time. Amsterdam down for a few hours today alone. We spend millions with them and can’t even get someone to join a bridge. It’s baaad
Microsoft pretends they do this with Azure where they'll let you signup with a credit card but anyone who has used Azure knows that the web admin is absolutely terrible and almost always requires speaking to a disinterested human to set it up properly.
I'm skeptical Oracle fits into the wider SaaS business category vs enterprise sales.
Not if the goal is to get you into a sales funnel with a sales exec to juice the contract value. Oracle doesn’t give a single shit about someone spending $500/month
I signed up for Oracle Cloud. I couldn’t get any of the free trial options to work due to capacity limits. I couldn’t get my payment method added so I could pay for real servers.
Then they terminated my free trial early with no explanation. I tried to add a payment method again and it didn’t work.
It turned into a bigger joke when Oracle sales people started emailing me to ask how my trial was going. They must have been given a list of email addresses and no information about the accounts. I would ask them for help getting my account unlocked or adding a payment method, they would send me emails for a couple weeks saying they were looking into it, then they’d ghost me.
Then a month later a new sales rep would email me and start the process over.
I checked Reddit and there were dozens of stories with the same experience.
Every couple of years I try Oracle Cloud and nothing ever really works. It feels like it was built by people who literally cannot see the final result of what they work on. Signing up, adding cards, clicking links in e-mails -- nothing worked! Nothing has ever reliably worked!
I hate Oracle with a passion for everything they've done throughout the years, I hope they burn in hell. Of course I don't want that for most of the people working there, but those making the decisions? Kindly go the way of your cloud and vanish from relevance.
Market signals on an impending AI bust are broader than just Oracle’s woes.
For example, Amazon just had a challenging bond offering where the market is clearly starting to seriously question the ROI on all this money being pumped into AI buildout. That does not bode well at all for AI-only companies without broader cash flow from other businesses. And when the cash dries up this whole thing comes crashing down like a house of cards.
I was at the ophthalmologist for the second time in two weeks - my new prescription wasn't quite right, new lenses should be here this week.
All that to say: I had to move my focus around a bit and re-read "...pumped into AI buildout." several times, because I thought I was reading Ed Zitron :D
A bunch of press on this today you can look up. Demand on the offering was much lower than expected and what materialized in prior rounds. Amazon had to sweeten the deal to get the money loaned.
Low bid to cover ratio - it's rare for bond auctions to out and out fail (that would be fairly disastrous), but you can have an auction where they successfully sell all the bonds they were trying to sell but with much less demand than they were hoping for.
That's not a good sign and it's a blatant red flag for the market
> Market signals on an impending AI bust are broader than just Oracle’s woes.
It's worse than that - I believe that Oracle is one of the (many) companies right now that, if their AI experimentation fails, will stop the music, and everyone will be running for a chair.
Oracle is one of a few foundational components in the circular-investing group of AI companies. If they fail to make their commitments they're the first domino to fall.
Wouldn't it be wiser to get out of the market into fixed rate assets like government bonds? Maybe have some into puts on SPY (or QQQ since tech would probably have bigger losses) too, but mainly getting out of long positions on what seems a really overvalued stock market
I just hope the companies aren't considered too big to fail. Bailing them out would be a bad idea.
They will be. When the SHTF, you'll see Rubio in the room^H^H^H^H circus tent, sitting right next to Bessent, arguing that propping up OpenAI is as much a national security interest as bailing out GM was.
#1: Great question, and I would love to hear the answers (And am learning from the ones posted)
#2: What I've done so far: Haven't bought stock in a year. Have moderate short positions on Palantir, SpaceX, and Tesla. Have big short positions in the most popular Quantum computing companies. (Scams IMO). I have sold most of my positions ("profit taking"?) in stocks which have gone up a lot in the past year. (Nvidia, Broadcom etc), and am no longer using margin; about 1/3 of my brokerage value is now "cash", generating ~3% interest.
Retail investors do not have access to systems that calculate risk, margins, pnl, etc... and generally also don't have the necessary knowledge and market data to price such instruments correctly.
Most ppl are better off KISSing and lowering risk by selling equity for fixed income.
You almost always lose a lot of money if you're seeking safety. Protection from downside risk on your S&P500 investments may cost 20-30% of your investment at which point you're better off just selling the investment and hoping it doesn't go up by that much.
It's worth adding that conventional wisdom says, you can't time the market. On average, people shifting between cash and stocks to time shocks lose out over just holding a fixed portfolio.
Sometimes conventional wisdom stops being wise. Also 90% of the people in charge of conventional wisdom have their personal wealth depend on retail investors not selling.
At the same time, one can make financial decisions based on risk rather than longterm expected returns. For instance, I'm happy with fixed income yields rn.
What would scare me is losing a big chunk of my portfolio in a downturn, exactly when I'm also most likely to lose my job.
I moved 80% of my money out of Vanguard's Target Date Retirement funds and into a money market on June 1st. In the 1.5 months since, the remaining Target Date Retirement fund has fluctuated up and down by about 0.1%. It has basically plateaued. I don't think I am losing out on potential short term gains. I like the idea that I have cash available to buy in on the day of the crash.
My boss has already done this several times over the past couple years because of some impeding market crash. Then he goes back and buys a week or so later.
Good luck dude! This kind of move can pay off big or not, clearly. I’ve personally talked to fable about this a lot, suggest everyone does.
There are a lot of failure modes. The dot-com bubble looked obvious in 1997; it popped in 2000. Anyone shorting in '97-'98 was carried out on a stretcher before being vindicated. In fact 2000-2002 fell in three brutal legs over two years, and anyone who leveraged up after the first 25% leg was destroyed by the next two.
> I moved 80% of my money out of Vanguard's Target Date Retirement funds
which target date fund exactly? You can increase risk/reward buy choosing a target date fund far in the future or you can reduce risk/reward by choosing a target date fund closer to the present. The point of those funds is to gradually reduce your risk as you get closer to your planned retirement date. I moved my 401k into a target date fund about +10 years from my planned retirement (I'm 50). So a little bit on the risk++ side but not much.
I don't know, but they aren't really in the same category either. The pandemic didn't shut down everything. It didn't really shut down much, people worked from home and got deliveries instead of doing things in person. There were sectors that were hit bad, but certainly not everything.
The AI crash is about stock market indicator ratios matching those that preceded other major crashes. That's what got me spooked. I don't want to be heavily invested in those companies when/if something bad happens.
Reminder: Serious people have been predicting a market crash "within the next 3 months" for 3 years now. In that time, the "market" has gone up around 70% (66%-86% depending on the what part you are looking at).
A friend of mine and I go out to lunch every 3 months and talk about, among other things, investing. We've made a trope of it, calling out the people who are predicting an imminent market crash every time we have lunch.
I'm not saying that it doesn't look like it's going to crash, but I'll also say that there's also a very sizeable downside potential for getting out of the market.
Just sell all your ETFs and buy them again when the market goes up or down. You're very likely to lose money with options and you will definitely lose a lot of money if you buy enough options to hedge your full exposure.
> What's the best way to hedge against this, considering many of us have significant savings in the market?
honestly, if you're >= 10 years away from needing that money (retirement or whatever) then the best hedge is to ignore the news and just keep contributing to your investment as always. I got caught up in a couple moments (tarif drama April before last was one) where i panicked and sold and then it only took a few months to get back to even meanwhile 18% of my capital gains were now due to the taxman. I wrote a check to the IRS for 10's of thousands for no reason except over reacting and ignoring every financial advisor's advice.
if you're going to need your investment money within 10 years then you need to get advice on how to start reducing risk (and therefore reward) because you don't have time to survive and repair from a crash.
I knew guys who panicked in Feb 2020, at the start of covid. They moved everything to cash, never got back into the market. Things recovered faster than they thought. The unfortunate truth is they would've more than doubled their money if they stayed invested.
Assuming you are the average person, and not a financial professional, using actual financial hedging instruments properly is unlikely, and far more likely to just increase risk and lower expected return.
A realistic way for an American citizen to reduce risk in the current market is to have a globally diversified portfolio that under-allocates to the US.
Everyone in the tech and media world is dead set on this being a bubble.
Yet, even now, Fable is able to do the work of 4-5 engineers when used by a single senior engineer. Teams can and will shrink.
Look at all the production and advertising companies switching over to Seedance. I know ad firms bidding 1/4th their typical contract price (pharma, P&G, etc.) and winning contract after contract.
This isn't dotcom "dark fiber" before demand. The demand is here now, big legacy firms are just struggling with deploying it. Nimble small teams are making a killing.
> Everyone in the tech and media world is dead set on this being a bubble.
is completely orthogonal to this:
> Yet, even now, Fable is able to do the work of 4-5 engineers when used by a single senior engineer.
The industry being in a bubble or not is irrelevant to the tech being good or bad. The dot-com bubble popped (and was a bubble) even while the tech was fit for purpose.
I think the "bubble" is more about return on investment and not usefulness of the technology. So much money has been invested on the assumption that so much return is going to materialize. The more money going in the bigger the expectation of return, that's the bubble.
> Yet, even now, Fable is able to do the work of 4-5 engineers when used by a single senior engineer. Teams can and will shrink.
If that's true or not, it's a bit irrelevant. Maybe teams won't shrink because of Jevon's paradox, or maybe tech debt will catch up.
But it doesn't matter because the people calling this a bubble mostly believe that the companies burning money cannot have the return on investment needed. This can be for a variety of reasons, but my favourite one is just that open source AI models are good enough, cost a fraction of what the frontier ones do (with predictable costs), can be fine tuned, and can be relied upon (no orange tweet banning your acces to the model you've been using). So for me OpenAI and Anthropic will really struggle to merit their valuations.
And then companies like Oracle are just a dumpster on fire. GPU hosting is a commodity business; expensive one, for sure, but there's no way in hell they'll make actual returns on the money burned with zero moat. And things are even worse when you consider the political involvement of the CEO and his nepo baby, which can easily burn good will.
Yes, but all bubbles (except the tulips...) have a real, valuable, new technology at their core. That it's amazing technology doesn't stop the financial side of it being a bubble. In fact it all but ensures it is.
The qemu-img(1) installed on my system claims that it supports every disk format supported by Virtualbox [0], so I guess the only thing left would be to be able to handle the "machine definition" file.
qemu definitely won't do that out of the box, so, yeah, VirtualBox is better than qemu there. But I bet there's a fancy-pants GUI out there that has an import wizard that will handle that for you.
My understanding of the ai circular financing racket is that not everyone will be running for a chair.
Nvidia owns all the chairs, and they’re letting other companies pretend to for a while, but if it all falls apart the backstop to the collapse will be nvidia.
Doesn't Nvidia's success depend hugely on AI money pumping up demand for their products? If/when AI companies run out of money to keep investing in data centres, the bottom will fall out of the market and hopefully we can go back to buying reasonably priced graphics cards.
Takes a lot of IaaS to support the GPUs and workflows, all of that kit is immediately re-useable as general purpose compute to exit the commercial DCs they operate out of today.
Much of their current debt fuelled expansion isn't singular to AI. The circular narrative ignores this.
Oracle has been a toxic tech for a long time (along with IBM). I don't think anyone is putting it in the same bucket of modern tech companies, let alone AI companies.
Frankly, their forays into dubious financial engineering and investments are expected at this point.
If you don't count the autosummary/gen answer at the top of googling an answer I would say so. Outside of the more technically inclined crowd I think the sentiment is if you aren't at the forefront (opus/fable/chatgpt) then your last or at least indistinguishable from all the rest of the lesser models.
If you're selling deterministic output, just use traditional code. If you product is inference, it has to be the best inference. This becomes more apparent when you bounce between powerful models and smaller cheaper ones, the cheaper ones _feel_ worse to use.
Pretty sure Google fits any definition of major model maker that SpaceX does, and had their IPO long before SpaceX.
Meta and Microsoft both are also significant makers of GenAI models that are public, though neither has a big tentpole LLM line that they sell access.to commercially like OpenAI, Anthropic. Google, SpaceX, which I infer might be what you mean by major model maker.
Meta had Llama, which set a lot of things on fire in a good way, and then disappeared from the scene as tech advanced.
What does Microsoft have?
Not sure SpaceX counts. Nobody sane uses Grok. It's untrustable due to reality-distorting political bias training, and it's strongly associated with CSAM production. Not what you want in a reliable corporate utility.
Nothing says “full of shit” like someone saying “market is signaling an impending X”. Why not make a huge levered bet and get wildly rich if you think so?
This is a pretty Oracle-specific situation, isn't it? They bet the company on an AI infrastructure buildout and levered hard to do it. Google, Amazon, and Microsoft aren't in comparable situations. Oracle is transforming itself into a value-added CoreWeave (not just in terms of product packaging but also the financial structure of the company), in a way the other hyperscalers aren't.
This story has been playing out for years now, and reads to me like the market simply recognizing that Oracle is not in the same business as it once was. It could succeed, wildly, at this new thing, but its risk isn't going to be valued based on the business it was 10 years ago.
I think he is wrong when he says AI is useless. I agree that it's far from being as useful as assholes like Amodei claim it is, but it clearly has use cases.
Then again, that is the weak part of Zitron's argument, and laughably the only thing his critics like to engage.
When he brings up the bonkers economics of AI, and how absolutely unsustainable it is, his arguments are very solid. I am still to see one good counter to him there. It's not like he lacks haters who could try.
well, I've been working on the Zig compiler lately, there you have to swear off using AI in any capacity. even your tab-complete, yes. I find that yes, I'm a little slower with regexes instead of GPT mini/Luna models as a find/replace engine, but I am quickly getting grasp of it, and I see the future where I'm faster with Neovim and ripgrep than Codex. the structure of the standard library is super-logical and the code is eye-wateringly readable. I don't even feel the need in, or will to, use Codex in any capacity when working on such a great project. DVUI is also great to work with.
that said, I also used to work in environments where even Claude Opus tripped over itself when trying to untangle its own slop. that made me looking forward to AI going bust and folding back into tab-complete. that bubble revealed a lot of projects that always were slop, even before AI. the LLM policy became a great proxy of the quality standard the codebase and the community upholds itself to
Levered free cash flow (LFCF) is the cash remaining after a company has paid its debts and operational costs. Oracle has 167.43B debt. $43 billion in last fiscal year.
Google, Meta, Microsoft, Amazon will be fine if AI bubble bursts. Oracle will be among first to go down in flames after OpenAI.
Is just wild to me people thinking ai is tulip fever or a massive bubble when every part of my life ai is entering. Even these forums 35 percent of posts are ai or vibe code related. At work (medical field) ai is replacing scribes and it can read an ecg better than your average doc. TSMC and chip companies are using in their pipelines. Pharm and bio companies are using. Archeologists are using to decode scrolls and find new petroglyphs. Education and tutoring will never be the same ... kids got lucky having YouTube but now you basically have your private tutor. Vfx is being infiltrated. Computer security. I look around and robots are delivering my food and waymo is picking me up. I turn on the news and in the last couple months Ukraine is now using ai targeting on their drones in addition to the machine vision. My apartment complex recently had a renovations and paint job and my landlord showed me how they designed the color scheme and renovations with chat gpt before getting a crew to do the work. I made an app for my family photography contest for the first time something I never dreamed of at 40 years old with no programming knowledge. I updated my framer website faster than I ever have with Ai.
So please explain to me how this is a bubble especially considering that most of these feature are based on llm and not even on how we primarily interact with the world ...visually. the bubble will happen after I can turn on a webcam and the program watches me draw or do a golf swing and gives me realtime tips or i put on some ar glasses and it coaches me at work .
The amount of compute needed for graphics real time info is astronomical compared to llm . We are so far from the top of a bubble. The problem with ai in my opinion invest with the mindset that what goes up must come down and if it went up big it must come down hard soon. That's not a rule of nature or anything somethings are bedrock and keep going up. I'm sure when electricity was invented and reached every house maybe some people thought the bubble was over but we keep needing more and more. There is zero evidence now that we will need less ai compute.
I think it's logical to be skeptical of chatgpt IPO etc but the sector as a whole is crushing and maybe because of fear will have some hiccups but will certainly prevail for a long time imo
People said similar things leading up to the dot com crash. The commercialization of the internet was indeed a watershed moment. That didn't mean it wasn't a bubble. Both can be true at the same time.
Should(or When) said 'bubble' 'pop', AI isn't going away or expected to lose relevance.
If you were around about the time of the Dot-Com bubble, you can better make sense of the saying.
The web never stopped being useful, it was the ridiculous and speculative valuations of companies, and outlandish claims that couldn't sustain themselves and eventually 'popped'.
203 comments
[ 4.2 ms ] story [ 66.9 ms ] threadI use their services, but I frankly don't care who provides it. I'll chase the chepest/best and have no issue switching from one to another.
The only moat I can see is Microsoft providing its services to companies in its Azure system. Nervous IT departments probably like that it's not leaving their control if Bob in the SAP team spins up some AI crap.
For the hyperscalers, there is an ease of remaining in the Azure/AWS/GCP fabric from a data provenance perspective, particularly for regulated industries or large, risk-averse enterprises. There's also, of course, a certain network egress tax in most cases.
Nvidia's entire business is dependent on Google not being able to make TPUs fast enough.
I am about to spend $20M, if I buy anything other than Nvidia, and things go wrong, I am going to get blamed, and if things go right I will get no credit. This is why AMD is making no progress outside of very narrow cases and supercomputing.
This is the leap, nobody really wants to front a model for someone else. If i build an agent, or a service that requires a model, I'd prefer to push the model onto someone else, preferably at no cost. This is a leap as I'm sure right now, most people / businesses are thinking actually i do want to own / front the model.
However, if you accept the leap the easiest way to do this is to make the model the users problem.
From a business point of view that makes things really easy, from a customer point of view, they simply have to accept whatever their vendor of choice is pushing down their throats.
So as a business I build for whatever model Google makes available to android, and whatever model windows bundles, and whatever model Apple bundles, and, excluding the long tail of Chinese vendors and Linux (sorry, its always left out) and that's it, problem solved, and the customer picks up the tab for the tokens
https://www.youtube.com/watch?v=2J2Fb1bBufA
Vendor lock in cannot happen, or you're bankrupt.
That's a losing proposition for any token provider - it's expensive and slow, and when you're done everyone with money to rent a last-gen H100 is going to distill your "closed" model anyway.
The specialized models for targeted verticals being discussed may well not be sold by tokens, but instead be behind the scenes powering dedicated packaged solutions where the customers don't have raw access to the model. Token providers still won’t have a moat, but AI isn't just selling tokens.
But I switched from ChatGPT to Claude 3 months ago because my account was down for like 6 hours. I haven’t used it since. It’s too easy to switch away from chatbots on a whim. There is no moat for that.
But... Anthropic doesn't have a moat. It's clear at this point that SOTA models are not a moat, and Opus 4.6-level (or GLM 5.2) is sufficient.
Google, though... they own the entire vertical, from the semiconductors to the end-user software. They may have a moat.
It’s not much of a moat, but it’s more than a lot of orgs have.
But as they have repeatedly pointed out, creating software is almost zero-cost now, so software cannot be a moat.
After all, all of the Claude software can be vibe-coded by any competitor; that's the dream that Anthropic has been selling anyway...
https://www.youtube.com/watch?v=2J2Fb1bBufA
There are competing definitions of what intelligence even is, and the one that I find most striking is from Francois Chollet which is that intelligence can be boiled down to skill acquisition efficiency. This type of definition makes intelligence more akin to polishing a ball than growing a watermelon.
The superintelligence doomers warn that the watermelon is going to start growing exponentially and crush everyone. But what might actually be happening is that we are not growing a watermelon but rather polishing the ball until its really smooth and shiny. There's a point where you can get it to micron levels of polish but for most tasks (white collar text domains tasks), it's smooth enough! You will be able to go to the ball store and buy a low cost made in china ball for most tasks.
The real challenge is actually branching out domains and modalities to tackle things like blue collar labor. Over time, white collar work automatable or able to be made hyperefficient by LLMs will see LLM commoditization.
https://www.youtube.com/watch?v=2J2Fb1bBufA
and they STILL went out of business because they over-estimated the demand for their shitty rails they built to the middle of nowhere. Same with "AI."
https://www.youtube.com/watch?v=2J2Fb1bBufA
When you are already trusting 100% of your data, and computing on that data, to someone like AWS, it doesn't meaningfully increase risk to use an additional service, even if it is an AI service.
Wouldn't want to be negative at a time like this.
I don't know what possessed Ellison to ruin a functioning company, but it will be interesting if he gets a margin call for ORCL's other debt exposures, which are Ellison's massive loans against his ORCL stock.
They are just rich. And also assholes.
Same thing that drives all these execs of large companies - naked greed!
"If only we can fire all workers, imagine how profitable we'll be!"
They are attempting to set civilisation on fire with the intention of being on top when they no longer need humans.
Given that Oracle and Microsoft are major counterparties of OpenAI, it seems odd that their stocks have been performing so poorly recently. Can anyone square this circle for me?
It’s like saying a new apartment building is “profitable” because the monthly income covers the monthly running costs, but ignoring the giant mortgage that covers the cost of building the building. That thinking is a good way to go bankrupt in real estate and a good way to go bankrupt in AI.
Or that it’s all hearsay and no one has released financials yet?
That's the best possible interpretation of them.
The other possible interpretation is that they are manipulating the numbers (that they have to show to investors) and inference isn't actually profitable either. If they are not manipulating the numbers right now, both companies have a serious case of uncontrolled operational costs that they have to solve too.
Correct. Because a less charitable interpretation will squint at their marketing spend and wonder if they are just sweeping a lot of their expenses on categories they don't belong to make their business look less bonkers to those that prefer to not ask the ugly questions.
I'm personally on the team that think it's honest. But then they have to explain why their marketing numbers are so weird.
To put it in perspective, that’s about the combined campaign budget of the Trump and Harris campaigns.
I would expect to be seeing a lot more openAI ads than I’m seeing right now. And they would be everywhere.
My less than charitable guess is that the leaked financials is a piece of fiction intended to make inference not look unprofitable.
What about retention though? Or upselling to higher tiers?
I would question where they decided to draw the line… because it’s a grey area.
That's why it's so hard to get a residential mortgage, for example. It's more of a partnership, with more mutual vulnerability, than most people think. Same thing seems to be true here.
Given what happened with xAI’s excess capacity lease to Anthropic, and Meta’s noises about doing the same, seems likely that the demand for inference will continue to slope upwards for a while. If I’m Oracle, I’m not worried about being able to utilize the data centers I’ve built for some price, almost certainly a profitable one.
I’m guessing, though, that Oracle made their capital investments on assumptions of a higher price & return. Possibly because it wasn’t clear when these decisions were made how much competition OpenAI would have at the frontier.
I don’t think this math is all that hard. Capital markets have everything they need to start to figure it out, most especially a year or two of history to project forward.
When dealing with sophists you have to pin them down on specific points and prevent them from shifting the conversation.
What does that tell me? Just one of many things about the prospects of the deal still happening. That one in particular says to me they won’t be deterred. Bond rating may suggest the opposite. Lots more complexity than those two things but “fun” to speculate.
https://www.forbes.com/sites/jonmarkman/2026/04/06/oracles-m...
In addition to that the form basically only worked in Edge. We emailed support, they changed something on the backend. It still did not work. We gave up.
In retrospective that was a very clear warning sign that their priorities were misguided. I'm glad we did not waste any further time and effort on them.
I'm guessing they don't care if actual business gets caught up in that because from their POV actual business comes from an account manager, and self-serve is just them cargo culting AWS/GCP
Uh, good luck guys.
CrowdStrike and Uber
> Hetzner
I don't know of any upper market EMEA customers on Hetzner. I've met Scaleway, OVHCloud, and even STACKIT users but never Hetzner.
Any hosted service can be bent into the shape of a cloud. Large parts of Oracle Cloud balance sheet is probably just hosted PeopleSoft and similar.
They have this in common with IBM which, at least on paper, have a large cloud business.
I'm skeptical Oracle fits into the wider SaaS business category vs enterprise sales.
Then they terminated my free trial early with no explanation. I tried to add a payment method again and it didn’t work.
It turned into a bigger joke when Oracle sales people started emailing me to ask how my trial was going. They must have been given a list of email addresses and no information about the accounts. I would ask them for help getting my account unlocked or adding a payment method, they would send me emails for a couple weeks saying they were looking into it, then they’d ghost me.
Then a month later a new sales rep would email me and start the process over.
I checked Reddit and there were dozens of stories with the same experience.
I hate Oracle with a passion for everything they've done throughout the years, I hope they burn in hell. Of course I don't want that for most of the people working there, but those making the decisions? Kindly go the way of your cloud and vanish from relevance.
For example, Amazon just had a challenging bond offering where the market is clearly starting to seriously question the ROI on all this money being pumped into AI buildout. That does not bode well at all for AI-only companies without broader cash flow from other businesses. And when the cash dries up this whole thing comes crashing down like a house of cards.
All that to say: I had to move my focus around a bit and re-read "...pumped into AI buildout." several times, because I thought I was reading Ed Zitron :D
That's not a good sign and it's a blatant red flag for the market
It's worse than that - I believe that Oracle is one of the (many) companies right now that, if their AI experimentation fails, will stop the music, and everyone will be running for a chair.
Oracle is one of a few foundational components in the circular-investing group of AI companies. If they fail to make their commitments they're the first domino to fall.
A few puts on SPY dated a year or two out?
I just hope the companies aren't considered too big to fail. Bailing them out would be a bad idea.
https://www.openmarketsinstitute.org/publications/no-bailout...
They will be. When the SHTF, you'll see Rubio in the room^H^H^H^H circus tent, sitting right next to Bessent, arguing that propping up OpenAI is as much a national security interest as bailing out GM was.
#2: What I've done so far: Haven't bought stock in a year. Have moderate short positions on Palantir, SpaceX, and Tesla. Have big short positions in the most popular Quantum computing companies. (Scams IMO). I have sold most of my positions ("profit taking"?) in stocks which have gone up a lot in the past year. (Nvidia, Broadcom etc), and am no longer using margin; about 1/3 of my brokerage value is now "cash", generating ~3% interest.
Never buy derivatives as a non institutional investor.
Most ppl are better off KISSing and lowering risk by selling equity for fixed income.
What? Absolutely not.
However! If you don’t want to learn and want to get rich quick instead, stay away.
At the same time, one can make financial decisions based on risk rather than longterm expected returns. For instance, I'm happy with fixed income yields rn.
What would scare me is losing a big chunk of my portfolio in a downturn, exactly when I'm also most likely to lose my job.
There are a lot of failure modes. The dot-com bubble looked obvious in 1997; it popped in 2000. Anyone shorting in '97-'98 was carried out on a stretcher before being vindicated. In fact 2000-2002 fell in three brutal legs over two years, and anyone who leveraged up after the first 25% leg was destroyed by the next two.
which target date fund exactly? You can increase risk/reward buy choosing a target date fund far in the future or you can reduce risk/reward by choosing a target date fund closer to the present. The point of those funds is to gradually reduce your risk as you get closer to your planned retirement date. I moved my 401k into a target date fund about +10 years from my planned retirement (I'm 50). So a little bit on the risk++ side but not much.
https://workplace.vanguard.com/investments/product-details/f...
You want to search for the chart at "Allocation to underlying funds (actual)"
The AI crash is about stock market indicator ratios matching those that preceded other major crashes. That's what got me spooked. I don't want to be heavily invested in those companies when/if something bad happens.
but it is possible to do safely. i’m a few decades in now
A friend of mine and I go out to lunch every 3 months and talk about, among other things, investing. We've made a trope of it, calling out the people who are predicting an imminent market crash every time we have lunch.
I'm not saying that it doesn't look like it's going to crash, but I'll also say that there's also a very sizeable downside potential for getting out of the market.
honestly, if you're >= 10 years away from needing that money (retirement or whatever) then the best hedge is to ignore the news and just keep contributing to your investment as always. I got caught up in a couple moments (tarif drama April before last was one) where i panicked and sold and then it only took a few months to get back to even meanwhile 18% of my capital gains were now due to the taxman. I wrote a check to the IRS for 10's of thousands for no reason except over reacting and ignoring every financial advisor's advice.
if you're going to need your investment money within 10 years then you need to get advice on how to start reducing risk (and therefore reward) because you don't have time to survive and repair from a crash.
Assuming you are the average person, and not a financial professional, using actual financial hedging instruments properly is unlikely, and far more likely to just increase risk and lower expected return.
A realistic way for an American citizen to reduce risk in the current market is to have a globally diversified portfolio that under-allocates to the US.
> market
These are two different things.
Because there are instruments that make market exposure easier, doesn’t make market exposure correct 100% of the time.
Yet, even now, Fable is able to do the work of 4-5 engineers when used by a single senior engineer. Teams can and will shrink.
Look at all the production and advertising companies switching over to Seedance. I know ad firms bidding 1/4th their typical contract price (pharma, P&G, etc.) and winning contract after contract.
This isn't dotcom "dark fiber" before demand. The demand is here now, big legacy firms are just struggling with deploying it. Nimble small teams are making a killing.
> Everyone in the tech and media world is dead set on this being a bubble.
is completely orthogonal to this:
> Yet, even now, Fable is able to do the work of 4-5 engineers when used by a single senior engineer.
The industry being in a bubble or not is irrelevant to the tech being good or bad. The dot-com bubble popped (and was a bubble) even while the tech was fit for purpose.
If that's true or not, it's a bit irrelevant. Maybe teams won't shrink because of Jevon's paradox, or maybe tech debt will catch up.
But it doesn't matter because the people calling this a bubble mostly believe that the companies burning money cannot have the return on investment needed. This can be for a variety of reasons, but my favourite one is just that open source AI models are good enough, cost a fraction of what the frontier ones do (with predictable costs), can be fine tuned, and can be relied upon (no orange tweet banning your acces to the model you've been using). So for me OpenAI and Anthropic will really struggle to merit their valuations.
And then companies like Oracle are just a dumpster on fire. GPU hosting is a commodity business; expensive one, for sure, but there's no way in hell they'll make actual returns on the money burned with zero moat. And things are even worse when you consider the political involvement of the CEO and his nepo baby, which can easily burn good will.
qemu definitely won't do that out of the box, so, yeah, VirtualBox is better than qemu there. But I bet there's a fancy-pants GUI out there that has an import wizard that will handle that for you.
[0] <https://www.virtualbox.org/manual/topics/storage.html#vdidet...>
Nvidia owns all the chairs, and they’re letting other companies pretend to for a while, but if it all falls apart the backstop to the collapse will be nvidia.
Much of their current debt fuelled expansion isn't singular to AI. The circular narrative ignores this.
Frankly, their forays into dubious financial engineering and investments are expected at this point.
Is Meta even in this race anymore?
If you're selling deterministic output, just use traditional code. If you product is inference, it has to be the best inference. This becomes more apparent when you bounce between powerful models and smaller cheaper ones, the cheaper ones _feel_ worse to use.
Meta and Microsoft both are also significant makers of GenAI models that are public, though neither has a big tentpole LLM line that they sell access.to commercially like OpenAI, Anthropic. Google, SpaceX, which I infer might be what you mean by major model maker.
What does Microsoft have?
Not sure SpaceX counts. Nobody sane uses Grok. It's untrustable due to reality-distorting political bias training, and it's strongly associated with CSAM production. Not what you want in a reliable corporate utility.
This story has been playing out for years now, and reads to me like the market simply recognizing that Oracle is not in the same business as it once was. It could succeed, wildly, at this new thing, but its risk isn't going to be valued based on the business it was 10 years ago.
The problem in this market is that too many players are trying to play a winner-takes-all angle.
For the companies that pull it off, it could be very lucrative.
In a real market we’ll get a couple of big winners rather than one, but there isn’t enough room for all of these moonshot efforts to land.
I don’t see the whole thing coming crashing down, but I do see a consolidation coming that leaves some companies in a very bad state.
[1] https://en.wikipedia.org/wiki/Heise_Group
That's quite the coincidence.
I think he is wrong when he says AI is useless. I agree that it's far from being as useful as assholes like Amodei claim it is, but it clearly has use cases.
Then again, that is the weak part of Zitron's argument, and laughably the only thing his critics like to engage.
When he brings up the bonkers economics of AI, and how absolutely unsustainable it is, his arguments are very solid. I am still to see one good counter to him there. It's not like he lacks haters who could try.
that said, I also used to work in environments where even Claude Opus tripped over itself when trying to untangle its own slop. that made me looking forward to AI going bust and folding back into tab-complete. that bubble revealed a lot of projects that always were slop, even before AI. the LLM policy became a great proxy of the quality standard the codebase and the community upholds itself to
Anyway...
Levered free cash flow (LFCF) is the cash remaining after a company has paid its debts and operational costs. Oracle has 167.43B debt. $43 billion in last fiscal year.
Google, Meta, Microsoft, Amazon will be fine if AI bubble bursts. Oracle will be among first to go down in flames after OpenAI.
So please explain to me how this is a bubble especially considering that most of these feature are based on llm and not even on how we primarily interact with the world ...visually. the bubble will happen after I can turn on a webcam and the program watches me draw or do a golf swing and gives me realtime tips or i put on some ar glasses and it coaches me at work .
The amount of compute needed for graphics real time info is astronomical compared to llm . We are so far from the top of a bubble. The problem with ai in my opinion invest with the mindset that what goes up must come down and if it went up big it must come down hard soon. That's not a rule of nature or anything somethings are bedrock and keep going up. I'm sure when electricity was invented and reached every house maybe some people thought the bubble was over but we keep needing more and more. There is zero evidence now that we will need less ai compute.
I think it's logical to be skeptical of chatgpt IPO etc but the sector as a whole is crushing and maybe because of fear will have some hiccups but will certainly prevail for a long time imo
LBMs will eat robotics, and that alone will eat a double digit percentage of labor.
If you were around about the time of the Dot-Com bubble, you can better make sense of the saying.
The web never stopped being useful, it was the ridiculous and speculative valuations of companies, and outlandish claims that couldn't sustain themselves and eventually 'popped'.