AI is going to be a highly-competitive, extremely capital-intensive commodity market that ends up in a race to the bottom competing on cost and efficiency of delivering models that have all reached the same asymptotic performance in the sense of intelligence, reasoning, etc.
The simple evidence for this is that everyone who has invested the same resources in AI has produced roughly the same result. OpenAI, Anthropic, Google, Meta, Deepseek, etc. There's no evidence of a technological moat or a competitive advantage in any of these companies.
The conclusion? AI is a world-changing technology, just like the railroads were, and it is going to soon explode in a huge bubble - just like the railroads did. That doesn't mean AI is going to go away, or that it won't change the world - railroads are still here and they did change the world - but from a venture investment perspective, get ready for a massive downturn.
Something nobody's talking about: OpenAI's losses might actually be attractive to certain investors from a tax perspective. Microsoft and other corporate investors can potentially use their share of OpenAI's operating losses to offset their own taxable income through partnership tax treatment. It's basically a tax-advantaged way to fund R&D - you get the loss deductions now while retaining upside optionality later. This is why the "cash burn = value destruction" framing misses the mark. For the right investor base, $10B in annual losses at OpenAI could be worth $2-3B in tax shields (depending on their bracket and how the structure works). That completely changes the return calculation. The real question isn't "can OpenAI justify its valuation" but rather "what's the blended tax rate of its investor base?" If you're sitting on a pile of profitable cloud revenue like Microsoft, suddenly OpenAI's burn rate starts looking like a pretty efficient way to minimize your tax bill while getting a free option on the AI leader. This also explains why big tech is so eager to invest at nosebleed valuations. They're not just betting on AI upside, they're getting immediate tax benefits that de-risk the whole thing.
> For the right investor base, $10B in annual losses at OpenAI could be worth $2-3B in tax shields (depending on their bracket and how the structure works). That completely changes the return calculation
I know nothing about finances at this level, so asking like a complete newbie: doesn't that just mean that instead of risking $10B they're risking $7-8B? It is a cheaper bet for sure, but doesn't look to me like a game changer when the range of the bet's outcome goes from 0 to 1000% or more.
It all depends on the actual numbers. Consider this simplified example: If you are offered a deal that requires you to lay down 10 billion today and it has a 5% chance to pay out 150 billion tomorrow, your accountants will tell you not to take this deal because your expected return is -2.5 billion. But if you can offset 3 billion in cost to the tax payer, your expected return suddenly becomes $500 million, making it a good deal that you should take every time.
I get that this example is simplified, but doesn’t the maths here change drastically when the 5% changes by even a few percentage points? The error bars on Openais chance of succes are obviously huge, so why would this be attractive to accountants?
If your accountants suggest that you take a single 5% chance deal, they probably skipped maths and statistics and you should fire them.
It's the dumb as rocks MBAs that will go head first into the 5% chance deal.
This applies to any spending Microsoft does. What does it have to do with OpenAI?
Also, classifying business expenses as "cost to the tax payer" seems less than useful, unless you are a proponent of simply taxing gross receipts. Which has its merits, but then the discussion is about taxing gross receipts versus income with at least some deductible expenses, not anything to do with OpenAI.
That just doesn't sound right. This kind of thought process only works if you think you are guaranteed more than that the next year. It only works in crony capitalism where your friends in government put money in your pockets. It's where we are right now, but definitely not something that is sustainable or something to aspire to.
Amazon already has not been paying any sort of income tax to the EU. There was a lawsuit in Belgium but Amazon has won that in late-2024 since they had a separate agreement in/with Luxembourg.
Speaking for EU, all big tech already not paying taxes one way or another, either using Dublin/Ireland (Google, Amazon, Microsoft, Meta, ...) and Luxembourg (Amazon & Microsoft as far as I can tell) to avoid such corporate/income taxes. Simply possible because all the earnings go back to the U.S. entity in terms of "IP rights".
> Amazon already has not been paying any sort of income tax to the EU.
That should be expected, because
https://european-union.europa.eu/priorities-and-actions/acti...
> The EU does not have a direct role in collecting taxes or setting tax rates.
> There was a lawsuit in Belgium but Amazon has won that in late-2024 since they had a separate agreement in/with Luxembourg.
Dec 2023.
> Speaking for EU, all big tech already not paying taxes one way or another, either using Dublin/Ireland (Google, Amazon, Microsoft, Meta, ...) and Luxembourg (Amazon & Microsoft as far as I can tell) to avoid such corporate/income taxes. Simply possible because all the earnings go back to the U.S. entity in terms of "IP rights".
Ireland (due to pressure from EU) closed this in 2020. The amount of tax collected by Ireland quadrupled. See Figure 5 and 6 in link below.
https://budgetmodel.wharton.upenn.edu/issues/2024/10/14/the-...
>> For the right investor base, $10B in annual losses at OpenAI could be worth $2-3B in tax shields
So just a loss for governments, or in other words, socializing the losses.
OpenAIs losses are someone else's (taxed) earnings.
Hi, I'm here to hold the bag?
We really should have thought of this before becoming peasants.
Have you tried not being poor?
It gives you a new opportunity to pull yourself up by the bootstraps. Until mommy and daddy come along with another cash infusion.
You guys are getting bags?
Your pension fund, yes.
This comment makes even less sense than jotras’ comment.
Pension funds buy shares in businesses such as Microsoft. The money going into the pension fund is not typically a function of the tax paid by companies such as Microsoft, but rather from a combination of actuaries’ recommendations, payroll tax receipts, and politicians’ priorities.
Therefore a pension funds’ equity holdings, such as Microsoft, doing well means taxes can be lower.
Private industry loses 10B. Governments mostly affected as they have less free money to extract.
> OpenAI's losses might actually be attractive to certain investors from a tax perspective.
OpenAI is anyways seeking Govt Bailout for "National Security" reasons. Wow, I earlier scoffed at "Privatize Profits, Socialize Losses", but this appears to now be Standard Operating Procedure in the U.S.
https://www.citizen.org/news/openais-request-for-massive-gov...
So the U.S. Taxpayer will effectively pay for it. And not just the U.S. Taxpayer - due to USD reserve currency status, increasing U.S. debt is effectively shared by the world. Make billionaires richer, make the middle class poor. Make the poor destitute. Make the destitute dead. (All USAID cuts)
There's already a lot that the US taxpayer is on the hook for that's a lot less valuable than a best on the next big thing in software, productivity, and warfare.
It shouldn't be the job of the US taxpayer to feed someone that doesn't want to work, study, or pass a drug test, and it absolutely shouldn't be the job of the US taxpayer to feed another country's citizens half a world away.
Hello, I'm British by birth.
That's pretty close to the story other Brits give themselves for why losing the empire was actually a good thing for the UK.
> It shouldn't be the job of the US taxpayer to feed someone that doesn't want to work, study, or pass a drug test
This would make sense if every person was given similar opportunities, like providing quality education to all of our youngest and making higher education a mission rather than a business as a starter.
As a society we move at the speed of the weakest among us, we only move forward when we start lifting and helping the weakest and most vulnerable.
You also need to realize that not doing that work is also cause for other taxpayer money to be spent elsewhere, such as spending an average of 37k $ per incarcerated person, and that ignores all the damage that criminal might've caused, all the additional police staffing and personal security that is needed to be spent outside prisons, etc.
Those are complex systems, are you sure it wouldn't be better to spend the same gargantuan amount of money that's spent on millions of inmates and fighting crime into fighting the causes that make many fall into that?
Again, those are complex, but closed systems and the argument of "we shouldn't spend on X" often ignores the cost of not spending on X.
> It shouldn't be the job of the US taxpayer to feed someone that doesn't want to work, study, or pass a drug test
What about someone who works and still can’t afford enough housing/food?
> shouldn't be the job of the US taxpayer to feed another country's citizens half a world away.
I mean where’s the profit in that, am i right?
The modern welfare state is the compromise reached by capitalist democracies to stave off communist revolutions. If you’re going to kill of the welfare part, be ready for the uprising part.
That's where the surveillance and the militarized police force(s) come in. Especially the former now has reached extraordinary levels, given that almost all communication now is easily trackable.
Compare that to when we still had revolutions, where it was very hard for government to know what is going on, and to find individuals without a huge effort.
I think revolutions have become next to impossible, unless it is lead by significant parts of the elite that controls at least part of the apparatus.
That's not even counting the far more sophisticated propaganda methods, so that many of the affected people won't even begin to target the actual culprits but are lead to chase shadows, or one another.
Whilst that is an option, it wont cover the share price hit from the fallout, which would wipe out more than the debt as when the big domino falls, others will follow as the market panic shifts.
So kinda looking at a bank level run on tech companies if they go broke.
Lmao this is ridiculous. If MSFT really wanted the tax benefits they should’ve just wholly acquired OAI long ago to acquire the financial synergy you speak of.
Can you explain it in another way? What you are saying is that instead of loosing 100% they loose 70% and loosing 70% is somehow good? Or are you saying the risk adjusted returns are then 30% better on the downside than previously thought? Because if you are, I think people here are saying the risk is so high that it is a given they will fail.
It’s hardly a free option, by your numbers it’d be a 20-30% discount.
Sure but if there's no moat would you rather pay 100% or 80% until the credits run out? You reap the 100% spend in the meantime. Not everyone even has the no moat discount.
There is a pretty big moat for Google: extreme amounts of video data on their existing services and absolutely no dependence on Nvidia and it's 90% margin.
Google has several enviable, if not moats, at least redoubts. TPUs, mass infrastructure and own their own cloud services, they own delivery mechanisms on mobile (Android) and every device (Chrome). And Google and Youtube are still #1 and #2 most visited websites in the world.
Not to mention security. I'd trust Google more not to have a data breach than open AI / whomever. Email accounts are hugely valuable but I haven't seen a Google data breach in the 20+ years I've been using them. This matters because I don't want my chats out there in public.
Also integration with other services. I just had Gemini summarize the contents of a Google Drive folder and it was effortless & effective
While I don’t disagree with you, for historical purposes I think it ’s important to highlight why google started its push for 100% wire encryption everywhere all the time:
The NSA and GHCQ and basically every TLA with the ability to tap a fibre cable had figured out the gap in Google’s armour: Google’s datacenter backhaul links were unencrypted. Tap into them, and you get _everything_.
I’ve no idea whether Snowdon’s leaks were a revelation or a confirmation for google themselves; either way, it’s arguably a total breach.
When I worked at PayPal back in 2003/4, one of the things we did (and I think we were the first) was encrypt the datacenter backhaul connections. This was on top of encrypting all the traffic between machines. It added a lot of expense and overhead, but security was important enough to justify it.
And yet Venmo, a Paypal company, publishes transaction data publicly by default, no need to decrypt anything ¯\_(ツ)_/¯
Not that I disagree with your assessment but in the spirit of hn pedantry - google had a very significant breach where gmail was a primary target and that was “only” 16 years ago in mid 2009. So bad that it has its own wikipedia page: https://en.wikipedia.org/wiki/Operation_Aurora
>very significant breach
That page says it was only 2 accounts and none of the messages within the mail was accessed. I wouldn't call that very significant.
Is Google even required to inform you of a data breach?
They're subject to California law, so yeah.
Don't forget the other moat.
While their competitors have to deal with actively hostile attempts to stop scraping training data, in Google's case almost everyone bends over backwards to give them easy access.
‘Actively hostile’ as in objecting to your content getting ripped off without permission?
It's a matter of perspective. In this scenario both sides see the other as hostile, just as one would look at a war happening as an outside observer.
The biggest moat is amount of money. Google has infinite amounts of money the print out of thin air (ads). They don't need complex entangled schemes with circular debts to prop up their operations.
They also have one of the biggest negatives in that they abandon almost everything they build so it’s hard to get invested in thier products.
I agree with the rest though
They don't abandon their money makers. That's the thing people don't get about the Google graveyard meme, they only cut things that obviously aren't working to make them more money.
I have yet to be convinced the broader population has an appetite for AI produced cinematography or videos. Independence from Nvidia is no more of a liability than dependence on electricity rates; it's not as if it's in Nvidia's interest to see one of its large customers fail. And pretty much any of the other Mag7 companies are capable of developing in-house TPUs + are already independently profitable, so Google isn't alone here.
The value of YouTube for AI isn't making AI videos, it's that it's an incredibly rich source for humanity's current knowledge in one place. All of the tutorials, lectures, news reports, etc. are great for training models.
Is that actually a moat? Seems like all model providers managed to scrape the entire textual internet just fine. If video is the next big thing I don’t see why they won’t scrape that too.
Scraping text across the entire internet is orders of magnitudes easier than scraping YouTube. Even ignoring the sheer volume of data (exabytes), you simply will get blocked at an IP and account level before you make a reasonable dent. Even if you controlled the entire IPv4 space I’m not sure you could scrape all of YouTube without getting every single address banned. IPv6 makes address bans harder, true, but then you’re still left with the problem of actually transferring and then storing that much data.
For now, you actually get pretty far with Tor. Just reset your connection when you hit an IP ban by sending SIGHUP to the Tor daemon.
I did that when I was retraining Stable Audio for fun and it really turned out to be trivial enough to pull of as a little evening side project.
IPv6 doesn't make it "harder," as they would typically ban whole /48 prefixes.
And we're probably already starting to see that, given the semirecent escalations in game of cat and also cat of youtube and the likes of youtube-dl.
Reminds me of Reddit's cracking down on API access after realizing that their data was useful. But I'd expect both youtube to be quicker on the gun knowing about AI data collection, and have more time because of the orders of magnitude greater bandwidth required to scrape video.
And reddit turned around and sold it all for a mess of pottage…
Sold being the operative word, rather than giving it away for free.
> Seems like all model providers managed to scrape the entire textual internet just fine
Google, though, has been doing it for literal decades. That could mean that they have something nobody else (except archive.org) has - a history on how the internet/knowledge has evolved.
If you think they are going to catch up with Google's software and hardware ecosystem on their first chip, you may be underestimating how hard this is. Google is on TPU v7. meta has already tried with MTIA v1 and v2. those haven't been deployed at scale for inference.
I don't think many of them will want to, though. I think as long as Nvidia/AMD/other hardware providers offer inference hardware at prices decent enough to not justify building a chip in-house, most companies won't. Some of them will probably experiment, although that will look more like a small team of researchers + a moderate budget rather than a burn-the-ships we're going to use only our own hardware approach.
Well, anthropic just purchased a million TPUs from Google because even with a healthy margin from Google, it's far more cost effective because of Nvidia's insane markup. That speaks for itself. Nvidia will not drop their margin because it will tank their stock price. it's half of the reason for all this circular financing - lowering their effective margin without lowering it on paper.
And, don't forget everyone's buying from TSMC in every case!
It's in Nvidia's interest to charge the absolute maximum they can without their customers failing. Every dollar of Nvidia's margin is your own lost margin. Utilities don't do that. Nvidia is objectively a way bigger liability than electricity rates.
it is in every business’s best interest to charge the maximum…
Utilities and insurance companies are two examples of business regulated to not charge the maximum, for public policy reasons.
we suggesting that nvidia/google/.. be regulated for like utilities?
I think it will be accepted by broader population. But if generation is easy and cheap I wonder if there is demand. And I mean as total demand in the segment. Will there be enough impressions to go around to actually profit from the content. Especially if storage is also considered.
The video data is probably good for training models, including text models.
Given the fact that Apple and Coke but rushed to produce AI slop, and the agreements with Disney, we are going to see a metric fuck-ton of AI-generated cinema in the next decade. The broader population's tastes are absolute harbage when it comes to cinema, so I don't see why you need convincing. 40+ superhero films should be enough.
And yes, all their competitors are making custom chips. Google is on TPU v7. absolutely nobody is going to get this right on the first try among their competitors - Google didn't.
Bigger problem for late starts now is that it will be hard to match the performance and cost of Google/Nvidia. It's an investment that had to have started years ago to be competitive now.
In this case the difference between its and it’s does alter the meaning of the sentence.
On paper, Google should never have allowed the ChatGPT moment to happen ; how did a then non-profit create what was basically a better search engine than Google?
Google suffers from classic Innovator's Dilemma and need competition to refocus on what ought to be basic survival instincts. What is worse is the search users are not the customers. The customers of Google Search are the advertisers and they will always prioritise the needs of the customers and squander their moats as soon as the threat is gone.
Google allowed this to happen because they listened to their compliance department and were afraid of a backslash if LLM says something that could anger people.
Sergey Brin interview: https://x.com/slow_developer/status/1999876970562166968?s=20
This attitude also partially explains the black vikings incident.
Exactly, Google's business isn't search, it's ads. Is ChatGPT a more profitable system for delivering ads? That doesn't appear so, which means there's really no reason for Google to have created it first.
There was a very negative "immune" response from the users when they perceived suggestions from ChatGPT as ads.
This will be hard for them to integrate in a way that won't annoy users / will be better implemented than any other competitor in the same space.
Or perhaps we just deal with all AI across the board serving us ads.... this makes more sense unfortunately.
There’s a very negative immune response to the idea of Netflix running ads.
And yet they’re there, in the form of prominent product placement in all of their original series along with strategic placement in the frame to make sure they appear in cropped clips posted to social media and made into gifs.
Stranger Things alone has had 100-200 brands show up under the warm guise of nostalgia, with Coke alone putting up millions for all the less-than-subtle screen time their products get.
I’m certain AI providers will figure out how to slyly put the highest bidder into a certain proportion of output without necessarily acting out that scene in Wayne’s World.
I suspect google can last much longer in regards to an AI model chat engine that competes with open AI and other companies, without needing a profit from that particular product in a timely manner. I can's say the same for the others. Google is using it's own money to fund this without mch pressure for immediate profit in a time deadline. They can rely on their other services for revenue and profit for the meantime.
Think about it in terms of the research they put out into the ether though. The research grows into something viable, they sit back and watch the response and move when it makes sense.
It's like that old concept of saying something wrong in a forum on purpose to have everyone flame you for being wrong and needing to prove themselves better by each writing more elaborate answers.
You catch more fish with bait.
Agreed. Even xAI's (Grok's) access to live data on x.com and millions of live video inputs from Tesla is a moat not enjoyed by OpenAI.
>Agreed. Even xAI's (Grok's) access to live data on x.com and millions of live video inputs from Tesla is a moat not enjoyed by OpenAI.
Tesla does not have live video feed from (every) Tesla car.
The TAM for video generation isn't as big as the other use cases.
I agree, but isn't the TAM for video generation all of movies, TV, and possibly video games, or all entertainment? That's a pretty big market.
What you’re competing for is people’s attention and the tam for that is biggest there is
YT is also a giant corpus of English via the transcription
Hasn't it all been scraped by other ai companies already?
> AI is a world-changing technology, just like the railroads were
This comparison keeps popping up, and I think it's misleading. The pace of technology uptake is completely different from that of railroads: the user base of ChatGPT alone went from 0 to 200 million in nine months, and it's now- after just three years- around 900 million users on a weekly basis. Even if you think that railroads and AI are equally impactful (I don't, I think AI will be far more impactful) the rapidity with which investments can turn into revenue and profit makes the situation entirely different from an investor's point of view.
> just three years- around 900 million users on a weekly basis.
Well, I rotate about a dozen of free accounts because I don't want to send 1 cent their way, I imagine I'm not the only one. I do the same for gemini, claude and deepseek, so all in all I account for like 50 "unique" weekly users
Apparently they have about 5% of paying customers, the amount of total users is meaningless, it just tells you how much money they burn and isn't an indication of anything else.
> I rotate about a dozen of free accounts .. I do the same for gemini, claude and deepseek
For someone who doesn't like the product and doesn't care about it, you surely make a lot of effort to use it.
Sometime you have to force the trickle down economy a bit, these people are destroying my industry I might as well cost them as much as possible before I have no choice but to move on.
It's also literally 0 effort, click > sign out > click > sign in. It saves me $200 a month, that's not too far from half of my rent
I can understand the spirit, though this reinforces my impression that the product is so good that people jump through hoops to use it, even if they hate it in principle. If they suddenly cut off any free access to it, how much would you be willing to pay per month to keep using it? One dollar? Ten? Twenty?
Also, maybe I'm missing something, but no amount of free accounts on ChatGPT gives you what you get with a paid subscription, especially with a $200 one; and there's paid plans from just $8/month.
I'm going to go out on a limb here and say that users who put that much effort into using this stuff for free, using a dozen different accounts, are very rare.
Railroads carried the goods that everybody used. That’s like almost 100% in a given country.
The pace was slower indeed. It takes time to build the railroads. But at that time advancements also lasted longer. Now it is often cash grabs until the next thing. Not comparable indeed but for other reasons.
It is beside the point, but
> I think AI will be far more impactful
is not correct IMO. Those are two very different areas. The impact of railroads on transport and everything transport-related cannot be understated. By now roads and cars have taken over much of it, and ships and airplanes are doing much more, but you have to look at the context at the time.
Paid user base or free user base? Because free user base on a very expensive product is next to meaningless.
It's meaningful because it shows that people like the product a lot, and for a lot of different reasons. There are only few products that can reach such market penetration, not to mention in only three years. As the quality of AI increases, people will quickly realise that they are willing to pay for it as much as they pay for electricity. And the same goes for businesses.
Railroads enabled people and goods to move from one place to another much easier and faster.
AI enables people to... produce even more useless slop than before?
At this point I'm taking the word "slop" as a sign meaning "I really didn't think this through and I'm just autocompleting based on a gut feeling and the first word that comes to mind".
That's an easy way out, isn't it?
Using thought terminating clichés in general is, and that can include "slop".
The irony of atacking thought terminating clichés while defending slop
Like railroads, internet, electricity, aviation or car industries before: they've all been indeed the future, and they all peaked (in relative terms), at the very early stages of these industries future.
And among them the overwhelming majority of companies in the sectors died. Out of the 2000ish car-related companies that existed in 1925 only 3 survived to today. And none of those 3 ended up a particularly good long term investment.
Anthropic is building moat around theirs models with claude code, Agent SDK, containers, programmatic tool use, tool search, skills and more. Once you fully integrate you will not switch. Also being capital intensive is a form of moat.
I think we will end up with market similar to cloud computing. Few big players with great margins creating cartel.
I thought that, too, but lately I've been using OpenCode with Claude Opus, rather than Claude Code, and have been loving it.
OpenCode has LSPs out of the box (coming to Claude Code, but not there yet), has a more extensive UI (e.g. sidebar showing pending todos), allows me to switch models mid-chat, has a desktop app (Electron-type wrapper, sure, but nevertheless, desktop; and it syncs with the TUI/web versions so you can use both at the same time), and so on.
So far I like it better, so for me that moat isn't that. The technical moat is still the superiority of the model, and others are bound to catch up there. Gemini 3 Preview is already doing better at some tasks (but frequently goes insane, sadly).
>Anthropic is building moat around theirs models with claude code, Agent SDK, containers, programmatic tool use, tool search, skills and more.
I think this is something the other big players could replicate rapidly, even simulating the exact UI, interactions, importing/exporting existing items, etc. that people are used to with claude products. I don't think this is that big of a moat in the long run. Other big players just seem to be carving up the landscape and see where they can can fit in for now, but once resource rich eyes focus on them, Anthropic's "moat" will disappear.
A GPT wrapper isn't a moat.
Most things are wrappers around RDBMSs.
All those features are basically prompts in various formats, not much of a moat.
Google’s moat:
Try “@gmail” in Gemini
Google’s surface area to apply AI is larger than any other company’s. And they have arguably the best multimodal model and indisputably the best flash model?
If the “moat” is not AI technology itself but merely sufficient other lines of business to deploy it well, then that’s further evidence that venture investments in AI startups will yield very poor returns.
It's funny that a decade ago the exit strategy of many of these startups would have been to get acquired by MSFT / META / GOOG. Now, the regulators have made a lot of these acquisitions effectively impossible for antitrust reasons.
Is it better for society for promising startups to die on the open market, or get acquired by a monopoly? The third option -- taking down the established players -- appears increasingly unlikely.
> Now, the regulators have made a lot of these acquisitions effectively impossible for antitrust reasons.
Is there any evidence that this is the case ? For very big merger (like nvdia and Arm tried) sure, but I can't think of a single time regulator stop a big player from buying a start up.
Try “@gmail” in Gemini
I think this is a problem for Google. Most users aren't going to do that unless they're told it's possible. 99% of users are working to a mental model of AI that they learned when they first encountered ChatGPT - the idea that AI is a separate app, that they can talk to and prompt to get outputs, and that's it. They're probably starting to learn that they can select models, and use different modes, but the idea of connecting to other apps isn't something they've grokked yet (and they won't until it's very obvious).
What people see as the featureset of AI is what OpenAI is delivering, not Google. Google are going to struggle to leverage their position as custodians of everyone's data if they can't get users to break out of that way of thinking. And honestly, right now, Google are delivering lots of disparate AI interfaces (Gemini, Opal, Nano Banana, etc) which isn't really teaching users that it's all just facets of the same system.
> I think this is a problem for Google. Most users aren't going to do that unless they're told it's possible.
Google is telling this in about a hundred different popups and inline hints when you use any of its products
I've use the Gemini app on my phone a fair bit recently and I've not seen it. That said, I don't think I've seen any popups either. Maybe I've blocked them...
Users are trained to close those without reading.
That kind of makes it sound like AI is a feature and not a product, which supports avalys' point.
I tried it, but nothing happened. It said that it sent an email but didn't. What is supposed to happen?
Also, Google doesn't have to finance Gemini using venture capital or debt, it can use its own money.
DeepMind also solved the protein folding problem, so they have that going for them.
> AI is going to be a highly-competitive, extremely capital-intensive commodity market
It already is. In terms of competition, I don't think we've seen any groundbreaking new research or architecture since the introduction of inference time compute ("thinking") in late 2024/early 2025 circa GPT-o4.
The majority of the cost/innovation now is training this 1-2 year old technology on increasingly large amounts of content, and developing more hardware capable of running these larger models at more scale. I think it's fair to say the majority of capital is now being dumped into hardware, whether that's HBM and research related to that, or increasingly powerful GPUs and TPUs.
But these components are applicable to a lot of other places other than AI, and I think we'll probably stumble across some manufacturing techniques or physics discoveries that will have a positive impact on other industries.
> that ends up in a race to the bottom competing on cost and efficiency of delivering
One could say that the introduction of the personal computer became a "race to the bottom." But it was only the start of the dot-com bubble era, a bubble that brought about a lot of beneficial market expansion.
> models that have all reached the same asymptotic performance in the sense of intelligence, reasoning, etc.
I definitely agree with the asymptotic performance. But I think the more exciting fact is that we can probably expect LLMs to get a LOT cheaper in the next few years as the current investments in hardware begin to pay off, and I think it's safe to assume that in 5-10 years, most entry-level laptops will be able to manage a local 30B sized model while still being capable of multitasking. As it gets cheaper, more applications for it become more practical.
---
Regarding OpenAI, I think it definitely stands in a somewhat precarious spot, since basically the majority of its valuation is justified by nothing less than expectations of future profit. Unlike Google, which was profitable before the introduction of Gemini, AI startups need to establish profitability still. I think although initial expectations were for B2C models for these AI companies, most of the ones that survive will do so by pivoting to a B2B structure. I think it's fair to say that most businesses are more inclined to spend money chasing AI than individuals, and that'll lead to an increase in AI consulting type firms.
> in 5-10 years, most entry-level laptops will be able to manage a local 30B sized model
I suspect most of the excitement and value will be on edge devices. Models sized 1.7B to 30B have improved incredibly in capability in just the last few months and are unrecognizably better than a year ago. With improved science, new efficiency hacks, and new ideas, I can’t even imagine what a 30B model with effective tooling available could do in a personal device in two years time.
Very interested in this! I'm mainly a ChatGPT user; for me, o3 was the first sign of true "intelligence" (not 'sentience' or anything like that, just actual, genuine usefulness). Are these models at that level yet? Or are they o1? Still GPT4 level?
Not nearly o3 level. Much better than GPT4, though! For instance Qwen 3 30b-a3b 2507 Reasoning gets 46 vs GPT 4's 21 and o3's 60-something on Artificial Analysis's benchmark aggregation score. Small local models ~30b params and below tend to benchmark far better than they actually work, too.
> One could say that the introduction of the personal computer became a "race to the bottom." But it was only the start of the dot-com bubble era, a bubble that brought about a lot of beneficial market expansion.
I think the comparison is only half valid since personal computers were really just a continuation of the innovation that was general purpose computing.
I don't think LLMs have quite as much mileage to offer, so to continue growing, "AI" will need at least a couple step changes in architecture and compute.
I don't think anyone knows for sure how much mileage/scalability LLMs have. Given what we do know, I suspect if you can afford to spend more compute on even longer training runs, you can still get much better results compared to SOTA, even for "simple" domains like text/language.
I think we're pretty much out of "spend more compute on even longer training runs" atp.
> But I think the more exciting fact is that we can probably expect LLMs to get a LOT cheaper in the next few years as the current investments in hardware begin to pay off
Citation needed!
Have you thought about what happens if we get a new improvement in model architecture like transformers that grows the compute needs even further
I, personally, use chatGPT for search more than I do Google these days. It, more often than not, gives me more exact results based on what I'm looking for and it produces links I can visit to get more information. I think this is where their competitive advantage lies if they can figure out how to monetize that.
We don’t need anecdotes. We have data. Google has been announcing quarter after quarter of record revenues and profits and hasn’t seen any decrease in search traffic. Apple also hinted at the fact that it also didn’t see any decreased revenues from the Google Search deal.
AI answers is good enough and there is a long history of companies who couldn’t monetize traffic via ads. The canonical example is Yahoo. Yahoo was one of the most traffic sites for 20 years and couldn’t monetize.
2nd issue: defaults matter. Google is the default search engine for Android devices, iOS devices and Macs whether users are using Safari or Chrome. It’s hard to get people to switch
3rd issue: any money that OpenAI makes off search ads, I’m sure Microsoft is going to want there cut. ChatGPT uses Bing
4th issue: OpenAIs costs are a lot higher than Google and they probably won’t be able to command a premium in ads. Google has its own search engine, its own servers, its own “GPUs” [sic],
5th: see #4. It costs OpenAI a lot more per ChatGPT request to serve a result than it costs Google. LLM search has a higher marginal cost.
I personally know people that used ChatGPT a lot but have recently moved to using Gemini.
There’s a couple of things going on but put simply - when there is no real lock in, humans enjoy variety. Until one firm creates a superior product with lock in, only those who are generating cash flows will survive.
OAI does not fit that description as of today.
I'm genuinely curious. Why do you do this instead of Google Searches which also have an AI Overview / answer at the top, that's basically exactly the same as putting your search query into a chat bot, but it ALSO has all the links from a regular Google search so you can quickly corroborate the info even using sources not from the original AI result (so you also see discordant sources from what the AI answer had)?
The regular google search AI doesn’t do thinky thinky mode. For most buying decisions these days I ask ChatGPT to go off and search and think for a while given certain constraints, while taking particular note of Reddit and YouTube comments, and come back with some recommendations. I’ve been delighted with the results.
I wouldn’t be surprised if ChatGPT was Pareto optimal for buying decisions… but I suspect there are a whole pile of Pareto optimal ways to make buying decisions, including “buy one of the Wirecutter picks” or “buy whatever Costco is selling”.
This will remain the case until we have another transformer-level leap in ML technology. I don’t expect such an advancement to be openly published when it is discovered.
> The simple evidence for this is that everyone who has invested the same resources in AI has produced roughly the same result.
I think this conflates together a lot of different types of AI investment - the application layer vs the model layer vs the cloud layer vs the chip layer.
It's entirely possible that it's hard to generate an economic profit at the model layer, but that doesn't mean that there can't be great returns from the other layers (and a lot of VC money is focused on the application layer).
Whilst those other layers are useful, none of them are particularly hard to build or rebuild when you have many millions of dollars on hand.
One doesn't need tens of billions for them.
Yeah, because making good chips (TPU) and compilers (XLA) is notoriously easy, right?
The railroads provided something of enduring value. They did something materially better than previous competitors (horsecarts and canals) could. Even today, nothing beats freight rail for efficient, cheap modest-speed movement of goods.
If we consider "AI" to be the current LLM and ImageGen bubble, I'm not sure we can say that.
We were all wowed that we could write a brief prompt and get 5,000 lines of React code or an anatomically questionable deepfake of Legally Distinct Chris Hemsworth dancing in a tutu. But once we got past the initial wow, we had to look at the finished product and it's usually not that great. AI as a research tool will spit back complete garbage with a straight face. AI images/video require a lot of manual cleanup to hold up to anything but the most transient scrutiny. AI text has such distinct tones that it's become a joke. AI code isn't better than good human-developed code and is prone to its own unique fault patterns.
It can deliver a lot of mediocrity in a hurry, but how much of that do we really need? I'd hope some of the post-bubble reckoning comes in the form of "if we don't have AI to do it (vendor failures or pricing-to-actual-cost makes it unaffordable), did we really need it in the first place?" I don't need 25 chatbots summarizing things I already read or pleading to "help with my writing" when I know what I want to say.
You're absolutely correct! ( ;) )
The issue is that generation of error-prone content is indeed not very valuable. It can be useful in software engineering, but I'd put it way below the infamous 10x increase in productivity.
Summarizing stuff is probably useful, too, but its usefulness depends on you sitting between many different communication channels and being constantly swamped in input. (Is that why CEOs love it?)
Generally, LLMs are great translators with a (very) lossly compressed knowledge DB attached. I think they're great user Interfaces, and they can help streamline buerocracy (instead of getting rid of it) but they will not help getting down the cost of production of tangible items. They won't solve housing.
My best bet is in medicine. Here, all the areas that LLMs excell at meet. A slightly distopian future cuts the expensive personal doctors and replaces them with (few) nurses and many devices and medicine controlled by a medical agent.
I was really hoping, and with a different administration I think there was a real shot, for a huge influx of cash into clean energy infrastructure.
Imagine a trillion dollars (frankly it might be more, we'll see) shoved into clean energy generation and huge upgrades to our distribution.
With a bubble burst all we'd be left with is a modern grid and so much clean energy we could accelerate our move off fossil fuels.
Plus a lot of extra compute, that's less clear of a long term value.
Alas.
This is different because now the cats out of the bag: AI is big money!
I don't expect AGI or Super intelligence to take that long but I do think it'll happen in private labs now. There's an AI business model (pay per token) that folks can use also.
> don't expect AGI or Super intelligence to take that long
I appreciate the optimism for what would be the biggest achievement (and possibly disaster) in human history. I wish other technologies like curing cancer, Alzheimer's, solving world hunger and peace would have similar timelines.
I think we'll find that that asymptote only holds for cases where the end user is not really an active participant in creating the next model:
- take your data
- make a model
- sell it back to you
Eventually all of the available data will have been squeezed for all it's worth the only way to differentiate oneself as an AI company will be to propel your users to new heights so that there's new stuff to learn. That growth will be slower, but I think it'll bear more meaningful fruit.
I'm not sure if today's investors are patient enough to see us through to that phase in any kind of a controlled manner, so I expect a bumpy ride in the interim.
Yeah except that models don't propel communities towards new heights. They drive towards the averages. They take from the best to give to the worst, so that as much value is destroyed as created. There's no virtuous cycle there...
Is that constraint fundamental to what they are? Or are they just reflecting the behavior of markets when there's low hanging fruit around?
When you look at models that were built for a specific purpose, closely intertwined with experts who care about that purpose, they absolutely propel communities to new heights. Consider the impact of alphafold, it won a Nobel prize, proteomics is forever changed.
The issue is that that's not currently the business model that's aimed at most of us. We have to have a race to the bottom first. We can have nice things later, if we're lucky, once a certain sort of investor goes broke and a different sort takes the helm. It's stupid, but its a stupidity that predates AI by a long shot.
>That doesn't mean AI is going to go away, or that it won't change the world - railroads are still here and they did change the world - but from a venture investment perspective, get ready for a massive downturn.
I don't know why people always imply that "the bubble will burst" means that "literally all Ai will die out and nothing will remain that is of use". The Dotcom bubble didn't kill the internet. But it was a bubble and it burst nonetheless, with ramifications that spanned decades.
All it really means when you believe a bubble will pop is "this asset is over-valued and it will soon, rapidly deflate in value to something more sustainable" . And that's a good thing long term, despite the rampant destruction such a crash will cause for the next few years.
But some people do believe that AI is all hype and it will all go away. It’s hard to find two people who actually mean the same thing when they talk about a “bubble” right now.
Or the airlines. Airlines have created a huge amount of economic value that has mostly been captured by other entities.
The "Railway Bubble" analogy is spot on.
As a loan officer in Japan who remembers the 1989 bubble, I see the same pattern. In the traditional "Shinise" world I work with, Cash is Oxygen. You hoard it to survive the inevitable crash. For OpenAI, Cash is Rocket Fuel. They are burning it all to reach "escape velocity" (AGI) before gravity kicks in.
In 1989, we also bet that land prices would outrun gravity forever. But usually, Physics (and Debt) wins in the end. When the railway bubble bursts, only those with "Oxygen" will survive.
I‘m aware this means leaving the original topic of this thread, but would you mind giving us a rundown of this whole Japan 1989 thing? I would love to read a first-person account.
I am honored to receive a question from a fellow "Craftsman" (I assume from your name).
To be honest, in 1989, I was just a child. I didn't drink the champagne. But as a banker today, I am the one cleaning up the broken glass. So I can tell you about 1989 from the perspective of a "Survivor's Loan Officer."
I see two realities every day.
One is the "Zombie" companies. Many SMEs here still list Golf Club Memberships on their books at 1989 prices. Today, they are worth maybe 1/20th of that value. Technically, these companies are insolvent, but they keep the "Ghost of 1989" on the books, hoping to one day write it off as a tax loss. It is a lie that has lasted 30 years.
But the real estate is even worse. I often visit apartment buildings built during the bubble. They are decaying, and tenants have fled to newer, modern buildings. The owner cannot sell the land because demolition costs hundreds of thousands of dollars—more than the land is worth.
The owner is now 70 years old. His family has drifted apart. He lives alone in one of the empty units, acting as the caretaker of his own ruin.
The bubble isn't just a graph in a history book. It is an old man trapped in a concrete box he built with "easy money." That is why I fear the "Cash Burn" of AI. When the fuel runs out, the wreckage doesn't just disappear. Someone has to live in it.
I nominate Sam Altman to be that Someone.
> Cash is Oxygen. You hoard it to survive the inevitable crash. For OpenAI, Cash is Rocket Fuel. They are burning it all to reach "escape velocity" (AGI) before gravity kicks in.
For OpenAI, cash is oxygen too; they're burning it all to reach escape velocity. They could use it to weather the upcoming storm, but I don't think they will.
Deepseek has invested the same amount as OpenAI?
Your premise is wrong in a very important way.
The cost of entry is far beyond extraordinary. You're acting like anybody can gain entry, when the exact opposite is the case. The door is closing right now. Just try to compete with OpenAI, let's see you calculate the price of attempting it. Scale it to 300, 500, 800 million users.
Why aren't there a dozen more Anthropics, given the valuation in question (and potential IPO)? Because it'll cost you tens of billions of dollars just to try to keep up. Nobody will give you that money. You can't get the GPUs, you can't get the engineers, you can't get the dollars, you can't build the datacenters. Hell, you can't even get the RAM these days, nor can you afford it.
Google & Co are capturing the market and will monetize it with advertising. They will generate trillions of dollars in revenue over the coming 10-15 years by doing so.
The barrier to entry is the same one that exists in search: it'll cost you well over one hundred billion dollars to try to be in the game at the level that Gemini will be at circa 2026-2027, for just five years.
Please, inform me of where you plan to get that one hundred billion dollars just to try to keep up. Even Anthropic is going to struggle to stay in the competition when the music (funding bubble) stops.
There are maybe a dozen or so companies in existence that can realistically try to compete with the likes of Gemini or GPT.
> Just try to compete with OpenAI, let's see you calculate the price of attempting it. Scale it to 300, 500, 800 million users.
Apparently the DeepSeek folks managed that feat. Even with the high initial barriers to entry you're talking about, there will always be ways to compete by specializing in some underserved niche and growing from there. Competition seems to be alive and well.
DeepSeek certainly managed that on the training side but in terms of inference, the actual product was unusably slow and unreliable at launch and for several months after. I have not bothered revisiting it.
Are you talking about the model or their service? There's plenty of options for using their models other than the official DeepSeek API.
Eh, I wouldn't be so sure, chips with brain matter and or light are on its way and or quantum chips, one of those or even a combination will give AI a gigantic boost in performance. Finally replacing a lot more humans and whoever implements it first will rule the world.
> chips with brain matter and or light
The... what now?
You seem to have forgotten that the ruling class requires tax payers to fund their incomes. If we're all out of work, there's nobody to buy their products and keep them rich.
Not sure this equation works out. If demand for labor goes towards zero it really means there is no demand. In other words, when AI and robots fulfil every desire of their owners there really is no need for “tax payers”
Did railroads change the world though?
They only lasted a couple of decades as the main transportation method. I'd say the internal combustion engine was a lot more transformative.
Pretty much every major historical trend of Western societies in the second half of the eighteenth century, from the development of the modern corporation to the advent of total war, was intimately tied to railroad transportation.
Transportation of people, yeah, but it still carries a majority of inter-city freight in North America.
Railroads built America and won multiple large wars.
Umm yes? The metro even if not a big deal in the states is like a small but quiet way it has changed public transport, plus moving freight, plus people over large distances, plus the bullet train that mixed luxury, speed and efficiency onto trains, all of these are quietly disruptive transformations, that I think we all take for granted.
Um meta didn't achieve the same results yet. And does it matter if they can all achieve the same results if they all manage high enough payoffs? I think subscription based income is only the beginning. Next stage is AI-based subcompanies encroaching on other industries (e.g. deepmind's drug company)
Massive upfront costs and second place is just first loser. It’s like building fabs but your product is infinitely copyable. Seems pretty rough.
What exactly is "second" place? No-one really knows what first place looks like. Everyone is certain that it will cost an arm, a leg and most of your organs.
For me, I think that, the possible winners will be close to fully funded up front and the losers will be trying to turn debt into profit and fail.
The rest of us self hoster types are hoping for a massive glut of GPUs and RAM to be dumped in a global fire sale. We are patient and have all those free offerings to play with for now to keep us going and even the subs are so far somewhat reasonable but we will flee in droves as soon as you try to ratchet up the price.
It's a bit unfortunate but we are waiting for a lot of large meme companies to die. Soz!
First place looks a lot like Google…
You and the other hobbyists aren't what's driving valuations. Enterprise subscriptions are.
OpenAI is 80% consumer subs
And do those subs justify their valuation?
They don't, the only thing that can justify it is if they get themselves into every business workflow. That's what the investors are counting on.
for me its clear OpenAI and Anthropic have a lead. I dont buy Gemini 3 being good. it isnt. whatever the benchmark said. same for meta and deepseek.
I’m waiting to get an RTX 5090 on the cheap.
A penny saved is a penny earned
I still don't understand how it's world-changing apart from considerably degrading the internet. It's laughable to compare it to railroads.
Did you try asking chatgpt to explain?
Just in time for a Government guaranteed backstop.
Also that open source models are just months behind
People seem to have the assumption that OpenAI and Anthropic dying would be synonymous with AI dying, and that's not the case. OpenAI and Anthropic spent a lot of capital on important research, and if the shareholders and equity markets cannot learn to value and respect that and instead let these companies die, new companies will be formed with the same tech, possibly by the same general group of people, thrive, and conveniently leave out the said shareholders.
Google was built on the shoulders of a lot of infrastructure tech developed by former search engine giants. Unfortunately the equity markets decided to devalue those giants instead of applaud them for their contributions to society.
You weren’t around pre Google were you? The only thing Google learned from other search engines is what not to do - like rank based on the number of times a keyword appeared and not to use expensive bespoked servers
I was around pre-Google.
Ranking was Google's 5% contribution to it. They stood on the shoulders of people who invented physical server and datacenter infrastructure, Unix/Linux, file systems, databases, error correction, distributed computing, the entire internet infrastructure, modern Ethernet, all kinds of stuff.
And none of that had to do with learning from other search engines…
Eh ... I question that 5% ranking is google's only contribution, even if it was important.
Everyone stood on the shoulders of file systems and databases, ethernet (and firewalls and netscreens, ...) Well, maybe a few stood on the shoulder of PHP.
Google did in fact pretty much figure out how to scale large number of servers (their racking, datacenters, clustering, global file systems etc) before most others did. I believe it was their ability to run the search engine cheap enough that enabled them to grow while largely retaining profitability early on.
Yeah, I remember the moment search engines invented computing, I cannot look at sand the same way anymore /s
Isn't it really the other way around? Not to say OpenAI and Anthropic haven't done important work, but the genesis of this entire market was paper on attention that came out of Google. We have the private messages inside OpenAI saying they needed to get to market ASAP or Google would kill them.
If performance indeed asymptotes, and if we are not at the end of silicon scaling or decreasing cost of compute, then it will eventually be possible to run the very best models at home on reasonably priced hardware.
Eventually the curves cross. Eventually the computer you can get for, say, $2000, becomes able to run the best models in existence.
The only way this doesn’t happen is if models do not asymptote or if computers stop getting cheaper per unit compute and storage.
This wouldn’t mean everyone would actually do this. Only sophisticated or privacy conscious people would. But what it would mean is that AI is cheap and commodity and there is no moat in just making or running models or in owning the best infrastructure for them.
AI is capital intensive because autodiff kinda sucks.
"AI is going to be a highly-competitive" - In what way?
It is not a railroad and the railroads did not explode in a bubble (OK a few early engines did explode but that is engineering). I think LLM driven investments in massive DCs is ill advised.
Yes they did, at least twice in the 19th century. It was the largest financial crisis before 1929
It did. I question the issue of "what problem am I trying to solve" with AI, though. Transportation across a huge swath of land had a clear problem space, and trains offered a very clear solution; created dedicated railing and you can transport 100x the resources at 10x the speed of a horseman (and I'm probably underselling these gains). In times where trekking across a continent took months, the efficiencies in communication and supply lines are immediately clear.
AI feels like a solution looking for a problem. Especially with 90% of consumer facing products. Were people asking for better chatbots, or to quickly deepfake some video scene? I think the bubble popping will re-reveal some incredible backend tools in tech, medical, and (eventually) robotics. But I don't think this is otherwise solving the problems they marketed on.
> AI feels like a solution looking for a problem.
The problem is increasing profits by replacing paid labor with something "good enough".
Doesn't sound like a very profitable problem to solve. At least, not in the long term (which no one orchestrating this is thinking in).
Long term is feudalism, the short term is how we get there.
Well I wish them the worst of luck. Those doing this need to go back to the 1880s and see how that ended long term.
Isn’t this what industrialisation was always about?
This is a use case that hasn't yet been proven out, though. "Good enough" for an executive may not be "good enough" to keep the company solvent, and there's no shortage of private equity morons who have no understanding of their own assets.
I agree, but it's the bet they're making. You don't end up with trillions in investment and valuations with chatbots and meme video generators.
Your view is ahistorical.
This is why I think China will win the AI race. As once it becomes a commodity no other country is capable of bringing down manufacturing and energy costs the way China is today. I am also rooting for them to get on parity with node size for chips for the same reason as they can crash the prices PC hardware.
I disagree based on personal experience. OpenAI is a step above in usefulness. Codex and GPT 5.2 Pro have no peers right now. I'm happy to pay them $200/month.> There's no evidence of a technological moat or a competitive advantage in any of these companies.I don't use my Google Pro subscription much. Gemini 3.0 Pro spends 1/10th of the time thinking compared to GPT 5.2 Thinking and outputs a worse answer or ignores my prompt. Similar story with Deepseek.
The public benchmarks tell a different story which is where I believe the sentiment online comes from, but I am going to trust my experience, because my experience can't be benchmaxxed.
I still find it so fascinating how experiences with these models are so varied.
I find codex & 5.2 Pro next to useless and nothing holds a candle to Opus 4.5 in terms of utility or quality.
There's probably something in how varied human brains and thought processes are. You and I likely think through problems in some fundamentally different way that leads to us favouring different models that more closely align with ourselves.
No one seems to ever talk about that though and instead we get these black and white statements about how our personally preferred model is the only obvious choice and company XYZ is clearly superior to all the competition.
There is always a comment like this in these threads. It’s just 50-50 whether it’s Claude or OpenAI.
We never hear what the actual questions are. I reckon it's Claude being great at coding in general and GPT being good at niche cases. "Spikey intelligence"
I’m not saying that no company will ever have an advantage. But with the pace of advances slowing, even if others are 6-12 months behind OpenAI, the conclusion is the same.
Personally I find GPT 5.2 to be nearly useless for my use case (which is not coding).
Can you give specific examples? I'm super interested to see where it fails.
For me OpenAI is the worst of all. Claude code and Gemini deep research is much much more better in terms of quality while ChatGPT hallucinating and saying “sorry you’re right”.
I use both and ChatGPT will absolutely glaze me. I will intentionally say some BS and ChatGPT will say “you’re so right.” It will hilariously try to make me feel good.
But Gemini will put me in my place. Sometimes I ask my question to Gemini because I don’t trust ChatGPT’s affirmations.
Truthfully I just use both.
I told ChatGPT via my settings that I often make mistakes and to call out my assumptions. So now it
1. Glazes me 2. Lists a variety of assumptions (some can be useful / interesting)
Answers the question
At least this way I don't spend a day pursuing an idea the wrong way because ChatGPT never pointed out something obvious.
Care to share the system prompt?
codex is sooo slow but it is good at planning, opus is good at coding but not at good at seeing the big picture
AI is turning into the worst possible business setup for AI startups. A commodity that requires huge capital investment and ongoing innovation to stay relevant. There’s no room for someone to run a small but profitable gold mine or couple of oil wells on the side. The only path to survival is investing crazy sums just to stay relevant and keep up. Meanwhile customers have virtually zero brand loyalty so if you slip behind just a bit folks will swap API endpoints and leave you in the dust. It’s a terrible setup business wise.
There’s also no real moat with all the major models converging to be “good enough” for nearly all use cases. Far beyond a typical race to the bottom.
Those like Google with other products will just add AI features and everyone else trying to make AI their product will just get completely crushed financially.
There is clearly a very strong moat. OpenAI is close to 1 billion active users on ChatGPT while Claude barely have any non-business users. Even though Anthropic had better models at different times this year, I never stopped using ChatGPT and paying for Plus.
We just don't know who will win in which area yet. It doesn't mean there is no moat.
I don’t think it’s a question of moat. The usage limits on the chat interface with the more advanced Claud models are brutal. I feel like I can barely start a conversation before I get shutdown. However, I switched over to Gemini almost completely and barely ever checkin with ChatGPT these days.
OpenAI has close to 1 billion users which are mostly free users and will switch provider the moment OpenAI start charging them or adding ads. Which they will, as OpenAI themselves said they are losing money even with 200$ subs. So that amount of users is pretty meaningless.
Google search is free to use too. So is Instagram. Facebook. Whatsapp. All of these have ads.
OpenAI says they're very profitable on inference.
> OpenAI says they're very profitable on inference.
Great, but they need to burn billions on advertising, freemeium, and mostly R&D for new models.
So do all their competitors.
All of these have ads. And none of these have an equal value alternative. OpenAI, Claude, Deepseek, Mistral, Gemini, are mostly the same to a regular user.
Bing is mostly the same. Kagi is mostly the same. Yahoo, Yandex, etc. It's 2025. Hardly any difference. There were tens of search engines in the 90s and 2000s that were generally mostly the same. Yet, Google still won and owned nearly all search monetization.
Search was even easier to switch. At least ChatGPT has memory.
Most chat apps are the same as Whatsapp. All of them are free too.
"Ask ChaGPT" is the equivalent to "google it" in 2025.
I did, moved off to Claude in 2024 and many others around me did the same. Do you have data for non-business claim besides anecdotal evidence?
Yes.
Claude has 1% based on this: https://gs.statcounter.com/ai-chatbot-market-share
Consumers overwhelmingly use ChatGPT over Claude. ChatGPT dominance has not wavered.
If you think of it like cloud, where it's a commodity that reaches competitive prices, then you can use it to build products and applications, instead of competing for infrastructure (see also: railroads, optical fiber)
There is tons of money to be made at the application layer, and VCs will start looking at that once the infrastructure layer collapses.
Here's a blog post I wrote about that: https://parsnip.substack.com/p/models-arent-moats
It's a good take. I think both trajectories are occurring simultaneously.
OpenAI challenging Google search is a winner takes all situation, not to mention the vast amounts of user data.
On the other hand, us lesser mortals can leverage AI like a commoditized service to build applications with it.
Not really though. The cloud has some stickiness. It’s pretty hard to move once you’ve settled in. For a lot of AI integrations though it’s just swapping some API endpoints and maybe tweaking the prompting a bit. For probably 95% of AI use cases there almost no barrier to switching.
Not sure why they put so much investment into videoSlop and imageSlop. Anthropic seems to be more focused at least.
Because almost everyone involved in AI race grew up in "winner takes it all" environments, typical for software, and they try really hard to make it reality. This means your model should do everything to just take 90% of market share, or at least 90% of specific niche.
The problem is, they can't find the moat, despite searching very hard, whatever you bake into your AI, your competitors will be able to replicate in few months. This is why OpenAI is striking deal with Disney, because copyright provides such moat.
> your competitors will be able to replicate in few months.
Will they really be able to replicate the quality while spending significantly less in compute investment? If not then the moat is still how much capital you can acquire for burning on training?
There are multiple tech companies with quadrillion-deep pockets.
Is that not what distillation is?
What does moat even mean anymore
> copyright provides such a moat.
Been saying this since the 2016 Alice case. Apple jumped into content production in 2017. They saw the long term value of copyright interests.
https://arstechnica.com/information-technology/2017/08/apple...
Alice changed things such that code monkeys algorithms were not patentable (except in some narrow cases where true runtime novelty can be established.) Since the transformers paper, the potential of self authoring content was obvious to those who can afford to think about things rather than hustle all day.
Apple wants to sell AI in an aluminum box while VCs need to prop up data center agrarianism; they need people to believe their server farms are essential.
Not an Apple fanboy but in this case, am rooting for their "your hardware, your model" aspirations.
Altman, Thiel, the VC model of make the serfs tend their server fields, their control of foundation models, is a gross feeling. It comes with the most religious like sense of fealty to political hierarchy and social structure that only exists as hallucination in the dying generations. The 50+ year old crowd cannot generationally churn fast enough.
OpenAIs opsec must be amazing, I had fully expected some version of ChatGPT to be leaked on torrent sites at some point this year. How do you manage to avoid something that could be exfiltrated on a hard disk from escaping your servers in all cases, forever?
The model size is probably the thing here. I suspect they took the FAANG remote workstation approach, where VScode runs on a remote machine. After all its not that great having a desktop with 8 monster GPUs under your desk. (x100)
Plus moving all that data about is expensive. Keeping things in the datacenter is means its faster and easier to secure.
Totally agree, people love to talk about how hopelessly behind Apple is in terms of AI progress when they’re in a better position to compete directly against Nvidia on hardware than anyone else.
Apple's always had great potential. They've struggled to execute on it.
But really, so has everyone else. There's two "races" for AI - creating models, and finding a consumer use case for them. Apple just isn't competing in creating models similar to the likes of OpenAI or Google. They also haven't really done much with using AI technology to deliver 'revolutionary' general purpose user-facing features using LLMs, but neither has anyone else beyond chat bots.
I'm not convinced ChatGPT as a consumer product can sustain current valuations, and everyone is still clamouring to find another way to present this tech to consumers.
I think a major part of it is the shovel selling. Nvidia is selling shovels to OpenAI. OpenAI is selling shovels to endless B2B, Consulting, Accounting, software firms buying into it...
My goodness, are you really saying, in effect, "I wish people over 50 would just hurry up and die"?!?
Good lord, expressing that kind of sentiment does not make for a useful and engaging conversation here on hacker news.
Clutch pearls
Striking deals without a proper vision is a waste of resources. And that’s the path OAI is on.
It's also why they bought 40% of the world's RAM supply, too
Committed to buying. They dont have the money to actually buy it (at least not yet).
OpenAI is (was?) extremely good at making things that go viral. The successful ones for sure boost subscriber count meaningfully
Studio Ghibli, Sora app. Go viral, juice numbers then turn the knobs down on copyrighted material. Atlas I believe was a less successful than they would've hoped for.
And because of too frequent version bumps that are sometimes released as an answer to Google's launch, rather than a meaningful improvement - I believe they're also having harder time going viral that way
Overall OpenAI throws stuff at the wall and see what sticks. Most of it doesn't and gets (semi) abandoned. But some of it does and it makes for better consumer product than Gemini
It seems to have worked well so far, though I'm sceptical it will be enough for long
Going viral is great when you're a small team or even a million dollar company. That can make or break your business.
Going viral as a billion dollar company spending upward of 1T is still not sustainable. You can't pay off a trillion dollars on "engagement". The entire advertising industry is "only" worth 1T as is: https://www.investors.com/news/advertising-industry-to-hit-1...
Selling a bunch of $20 a month subscriptions isn’t going to make a dent in OpenAI losses. Going viral for a day or two doesn’t help.
Normal people are already getting tired of AI Slop
Because the general idea here is that image and video models, when scaled way up, can generalize like text models did[1], and eventually be treated as "world models"[2]; models that can accurately model real world processes. These "world models" then could be used to train embodied agents with RL in an scalable way[3]. The video-slop and image-slop generators is just a way to take advantage of the current research in world models to get more out of it.
[1] https://arxiv.org/pdf/2509.20328
[2] https://deepmind.google/blog/genie-3-a-new-frontier-for-worl...
Because as with the internet 99% of the usage won’t be for education, work, personal development, what have you. It will be for effing kitten videos and memes.
That’s an unusual way of saying uh…adult entertainment
Are the posters of effing kitten videos a customer base with a significant LTV?
(The obvious well-paying market would be erotic / furry / porn, but it's too toxic to publicly touch, at least in the US.)
Openrouter stats already mention 52% usage is roleplay.
As for photo/video very large number of people use it for friends and family (turn photo into creative/funny video, change photo, etc.).
Also I would think photoshop-like features are coming more and more in chatgpt and alike. For example, “take my poorly-lit photo and make it look professional and suitable for linkedin profile”
Also FWIW I understand that the furry community has a strong culture of commissioning artists for their work, so that's likely to be a headwind against using genAI that isn't explicitly trained only on licensed materials. Sure, there are likely some who would use it regardless, but I expect the use of genAI to generate furry porn to be at least as toxic within that community as the use of genAI to generate furry porn outside of that community.
If only 99% of the Internet was kitten videos and memes
Well, it sure as hell not all 3blue1brown, crr0ww, Feynman, and alike
Because OpenAI stands for AI leader.
If Gemini can create or edit an image, chatgpt needs to be able to do this too. Who wants to copy&paste prompts between ai agents?
Also if you want to have more semantics, you add image, video and audio to your model. It gets smarter because of it.
OpenAI is also relevant bigger than antropic and is known as a generic 'helper'. Antropic probably saw the benefits of being more focused on developer which allows it to succeed longer in the game for the amount of money they have.
> Who wants to copy&paste prompts between ai agents?
An AI!
The specialist vs generalist debate is still open. And for complex problems, sure, having a model that runs on a small galaxy may be worth it. But for most tasks, a fleet of tailor-made smaller models being called on by an agent seems like a solidly-precedented (albeit not singularity-triggering) bet.
not an expert by any means, but wouldn't smaller but highly refined models also output more reproducible results?> But for most tasks, a fleet of tailor-made smaller models being called on by an agent seems like a solidly-precedented (albeit not singularity-triggering) bet.intuitively it sounds akin to the unix model...
But then again the main selling point of using LLMs as part of some code that solves a certain business need is that you don't have to finetune a usecase-specific model (like in the mid 2010s), you just prompt engineer a bit and it often magically works.
>Also if you want to have more semantics, you add image, video and audio to your model. It gets smarter because of it.
I think you are confusing generation with analysis. As far I am aware your model does not need to be good at generating images to be able to decode an image.
It is, to first approximation, the same thing. The generative part of genAI is just running the analysis model in reverse.
Now there are all sorts of tricks to get the output of this to be good, and maybe they shouldn't be spending time and resources on this. But the core capability is shared.
> The generative part of genAI is just running the analysis model in reverse.
I think that hasn't been the case since DeepDream?
I think you're partially right, but I don't think being an AI leader is the main motivation -- that's a side effect.
I think it's important to OpenAI to support as many use-cases as possible. Right now, the experience that most people have with ChatGPT is through small revenue individual accounts. Individual subscriptions with individual needs, but modest budgets.
The bigger money is in enterprise and corporate accounts. To land these accounts, OpenAI will need to provide coverage across as many use-cases as they can so that they can operate as a one-stop AI provider. If a company needs to use OpenAI for chat, Anthropic for coding, and Google for video, what's the point? If Google's chat and coding is "good enough" and you need to have video generation, then that company is going to go with Google for everything. For the end-game I think OpenAI is playing for, they will need to be competitive in all modalities of AI.
> Because OpenAI stands for AI leader.
It'll just end up spreading itself too thin and be second or third best at everything.
The 500lb gorilla in the room is Google. They have endless money and maybe even more importantly they have endless hardware. OpenAI are going to have an increasingly hard time competing with them.
That Gemini 3 is crushing it right now isn't the problem. It's Gemini 4 or 5 that will likely leave them in the dust for the general use case, meanwhile specialist models will eat what remains of their lunch.
Because there is only so much programmers and companies will pay for AI coders. The big prizes is AI-generated TikTok.
The entertainment industry is by far the easiest way to tap into global discretionary income.
When they released their first good image model is when they got a new 100 million users in a week.
Because for all the incessant whining about "slop," multimodal AI i/o is incredibly useful. Being able to take a photo of a home repair issue, have it diagnosed, and return a diagram showing you what to do with it is great, and it's the same algos that power the slop. "Sorry, you'll have to go to Gemini for that use case, people got mad about memes on the internet" is not really a good way for them to be a mass consumer company.
Can Claude not do that? I've sent it pictures for simpler things and got answers, usually Id of bugs and plants.
Yes, Claude is multi-modal.
Because those and world models are the endgame, way way more than text is.
The fact that they do this isn't very bullish for them achieving whatever they define as AGI.
You don't expect AGI to be multi-modal?
What is AGI?
Artificial general intelligence
Because their main use is for advertising/propaganda, which is largely videoSlop & imageSlop even without AI.
Outside of this: https://openai.com/index/disney-sora-agreement/ I don't think there has been much of a win for them even in advertising for image/video slop.
It's like half the poster on here live in some parallel universe. I am making real money using generated image/video advertising content for both B2C and B2B goods. I am using Whisper and LLMs to review customer service call logs at scale and identity development opportunities for staff. I am using GPT/Gemini to help write SQL queries and little python scripts to do data analysis on my customer base. My business's productivity is way up since GenAI become accessible.
that (very vocal) half tried it once and it didn’t work :)
because these are mostly the same players of the 2010's. So when they can't get more investor money and the hard problems are still being cracked, the easiest fallback is the same social media slop they used to become successful 10-15 years prior. Falling back on old ways to maximize engagement and grind out (eventually) ad revenue.
But how much more profitable are they? We see revenue but not profits / spending. Anthropic seems to be growing faster than OpenAI did but that could be the benefit of post-GPT hype.
This article doesn’t add anything to what we know already. It’s still an open question what happens with the labs this coming year, but I personally think Anthropic’s focus on coding represents the clearest path to subscriber-based success (typical SaaS) whereas OpenAI has a clear opportunity with advertising. Both of these paths could be very lucrative. Meanwhile I expect Google will continue to struggle with making products that people actually want to use, irrespective of the quality of its models.
What Google AI products do people not want to use? Gemini is catching up to chatpt from a MAU perspective, ai overviews in search are super popular and staggeringly more used than any other ai-based product out there, a Google ai mode has decent usage, and Google Lens has surprisingly high usage. These products together dwarf everyone else out there by like 10x.
>Google Lens has surprisingly high usage
I use it several times a day just to change text in image form to text form so you can search it and the like.
It's built into chrome but they move the hidden icon about regularly to confuse you. This month you click the url and it appears underneath, helpfully labeled "Ask Google about this page" so as to give you little idea it's Google Lens.
> ai overviews in search are super popular and staggeringly more used than any other ai-based product out there
This really is the critical bit. A year ago, the spin was "ChatGPT AI results are better than search, why would you use Google?", now it's "Search result AI is just as good as ChatGPT, why bother?".
When they were disruptive, it was enough to be different to believe that they'd win. Now they need to actually be better. And... they kinda aren't, really? I mean, lots of people like them! But for Regular Janes at the keyboard, who cares? Just type your search and see what it says.
Is Gemini, as a chatbot, a product that sustains current valuations and investment?
>Gemini is catching up to chatpt from a MAU perspective
It is far behind, and GPT hasn't exactly stopped growing either. Weekly Active Users, Monthly visits...Gemini is nowhere near. They're comfortably second, but second is still well below first.
>ai overviews in search are super popular and staggeringly more used than any other ai-based product out there
Is it ? How would you even know ? It's a forced feature you can not opt out of or not use. I ignore AI overviews, but would still count as a 'user' to you.
we're curious what your source is
Monthly Visits (chatGPT #5, Gemini #26): https://www.similarweb.com/blog/research/market-research/mos...
Search Traffic: https://x.com/Similarweb/status/2003078223135990246
ChatGPT - 5.8b visits - -5.2% MoM
Gemini - 1.4b visits - +14.4% MoM
Yeah, ChatGPT is still more popular, but this does not show Gemini struggling exactly.
doesn't seem accurate since gemini has many entry points
Entry points ? The visits are accurate for the website and app. If you're talking about AI overviews, then that's meaningless for reasons I've already explained.
I do understand why it makes it very hard to compare but it's certainly not meaningless. Google's AI overviews are pretty much the only way that I use AI.
I mean we're all talking about how Google is 'catching up' and 'taking over' Open AI right ? In that case, it genuinely is meaningless. AI Overviews, even if it had the usage OP assumes, is not a threat to Open AI or chatGPT. People use chatGPT for a lot of different things, and AI overviews only handles (rather poorly in my opinion) a small, limited part of the kind of things it gets used for. I use AI mode a lot. It's better than Overviews in every conceivable way, and it's still not a chatGPT replacement.
https://cdn.openai.com/pdf/a253471f-8260-40c6-a2cc-aa93fe9f1...
Where does google struggle to make products people don’t want to use? Is it a personal opinion?
Bart was a flop. Google search is losing market share to other LLM providers. Gemini adoption is low, people around me prefer OpenAI because it is good enough and known.
But on the contrary, Nano Banana is very good, so I don't know. And in the end, I'm pretty confident Google will be the AI race winner, because they got the engineers, they tech background and the money. Unless Google Adsense die, they can continue the race forever.
> Gemini adoption is low, people around me prefer OpenAI because it is good enough and known.
Gemini is built into Android and Google search. People may not be going to gemini.google.com, but that does not mean adoption is low.
If Google is producing very good models and they aren’t gaining much traction, that seems like a pretty bad sign for them, right? If they were failing with bad models, the solution would be easy: math and engineer harder, make better models (I mean, this is obviously very hard but it is a clear path). Failing with good models is… confusing, it indicates there’s some unknown problem.
It’s irrelevant, Google needs to focus on performance enhancements that the enterprise market segment demands - who only operate in the air of objectivity.
If they can achieve that they will cut off a key source of blood supply to MSFT+OAI. There is not much money in the consumer market segment from subscribers and entering the ad-business is going to be a lot tougher than people think.
what are you talking about Gemini adoption has tripled in a few months alone and have around 18% of marketshare and its accelerating.
I’ve heard too many rumors that much of that adoption is from copying ms i.e. bundling gemini into their office suite
gemini is in basically everything from google now, from google docs to firebase to android studio so i wouldn't be surprised...
Gemini adoption via search is legit, though. I had a question, I got an answer, its not forced, fake adoption in that case.
It's hardly legit for everyone. I had a question, I got an incorrect answer I didn't ask for, scrolled past it and got actual results.
Fair
I mean it’s a simple Google search to show that Google isn’t loosing much market share to ChatGPT and most ChatGPT users still use Google.
https://searchengineland.com/nearly-all-chatgpt-users-visit-...
But even more importantly, it obviously isn’t losing money from advertisers to ChatGPT. You can look at their quarterly results.
Anti Gravity is a flop. I mean it uses Gemini under the hood.
But you cannot use it with an API key.
If you're on a workspace account, you can't have normal individual plan.
You have to have the team plan with $100/month or nothing.
Google's product management tier is beyond me.
OK, but Gmail, Google Maps, Google Docs, and Google Search etc are ubiquitous. `Google' has even become a verb. Google might take a shotgun approach, but it certainly does create widely used products.
I will add that there's also Gemini in Chrome. With Chrome being the largest browser by market share, that's a powerful de facto default.
where art thou anti-trust enforcement...> With Chrome being the largest browser by market share, that's a powerful de facto default.Every personal computer user except Chromebook users went out of their way to download Chrome. What exactly do you want “anti trust” to do?
maybe not allow google to bundle gemini with chrome?
So should we also not allow OpenAI to bundle the OpenAI model with the ChatGPT app
Absolutely no one besides ChromeOS users are forced to use Chrome.
That doesn't negate my original point.
What "we" know already is hard to add to, as a forum that has a dozen AI articles a day on every little morsel of news.
>whereas OpenAI has a clear opportunity with advertising.
Personally, having "a clear opportunity with advertising" feels like a last ditch effort for a company that promised the moon in solving all the hard problems in the world.
There are other avenues of income. You can invade other industries which are slow on AI uptake and build an AI-from-ground competitor with large advantages over peers. There are hints of this (not AI-from-ground but with more AI) with deepmind's drug research labs. But this can be a huge source of income. You can kill entire industries which inevitably cannot incorporate AI as fast as AI companies can internally.
I don't. Google has at least a few advantages:
1. Google books, which they legally scanned. No dubious training sets for them. They also regularly scrape the entire internet. And they have YouTube. Easy access to the best training data, all legally.
2. Direct access to the biggest search index. When you ask ChatGPT to search for something it is basically just doing what we do but a bit faster. Google can be much smarter, and because it has direct access it's also faster. Search is a huge use case of these services.
3. They have existing services like Android, Gmail, Google Maps, Photos, Assistant/Home etc. that they can integrate into their AI.
The difference in model capability seems to be marginal at best, or even in Google's favour.
OpenAI has "it's not Google" going for it, and also AI brand recognition (everyone knows what ChatGPT is). Tbh I doubt that will be enough.
And they have hardware as well, and their own cloud platform.
In my view Google is uniquely well positioned because, contrary to the others, it controls most of the raw materials for Ai.
Google's most significant advantage in this space is its organizational experience in providing services at this scale, as well as its mature infrastructure to support them. When the bubble pops, it's not lights-out or permanently degraded performance.
Probably more people use googles AI than anything else. Every search result has an LLM generated summary at the top.
it is a large spinning plate that can only keep spinning with more money, so the plate gets bigger and bigger, with everyone betting that it would carry on spinnning by itself to the stage that it has become too big to fail, due to the fallout, the impact on the stock market upon others companies would wipe out more than the sum of their debts. It's kinda at that stage now as when one domino falls, the impact on others will follow.
Just a case of too many companies have skin in OpenAI's game for it to be allowed to fail now.
The fact is nobody has any idea what OpenAI's cash burn is. Measuring how much they're raising is not an adequate proxy.
For all we know, they could be accumulating capital to weather an AI winter.
It's also worth noting that OpenAI has not trained a new model since gpt4o (all subsequent models are routing systems and prompt chains built on top of 4), so the idea of OpenAI being stuck in some kind of runaway training expense is not real.
The GPT-5 series is a new model, based on the o1/o3 series. It's very much inaccurate to say that it's a routing system and prompt chain built on top of 4o. 4o was not a reasoning model and reasoning prompts are very weak compared to actual RLVR training.
No one knows whether the base model has changed, but 4o was not a base model, and neither is 5.x. Although I would be kind of surprised if the base model hadn't also changed, FWIW: they've significantly advanced their synthetic data generation pipeline (as made obvious via their gpt-oss-120b release, which allegedly was entirely generated from their synthetic data pipelines), which is a little silly if they're not using it to augment pretraining/midtraining for the models they actually make money from. But either way, 5.x isn't just a prompt chain and routing on top of 4o.
Prior to 5.2 you couldn’t expect to get good answers to questions prior to March 2024. It was arguing with me that Bruno Mars did not have two hit songs in the last year. It’s clear that in 2025 OpenAI used the old 4.0 base model and tried to supercharge it using RLVR. That had very mixed results.
That just means their pretraining data set was older. You can train as many models as you want on the same data.
I’m sure all these AI labs have extensive data gathering, cleanup, and validation processes for new data they train the model on.
Or at least I hope they don’t just download the current state of the web on the day they need to start training the new model and cross their fingers.
I think you are messing up things here, and I think your comment is based on the article from semi analysis. [1]
It said: OpenAI’s leading researchers have not completed a successful full-scale pre-training run that was broadly deployed for a new frontier model since GPT-4o in May 2024, highlighting the significant technical hurdle that Google’s TPU fleet has managed to overcome.
However, pre-training run is the initial, from-scratch training of the base model. You say they only added routing and prompts, but that's not what the original article says. They most likely still have done a lot of fine tuning, RLHF, alignment and tool calling improvements. All that stuff is training too. And it is totally fine, just look at the great results they got with Codex-high.
If you got actually got what you said from a different source, please link it. I would like to read it. If you just messed things up, that's fine too.
[1] https://newsletter.semianalysis.com/p/tpuv7-google-takes-a-s...
Didn't they create Sora and other models and literally burned so much money with their AI video app which they wanted to make a social media but what ended up happening was that they burned billions of dollars.
I wonder about what happens to people who make these hilariously bad business decisions? Like the person at Twitter who decided to kill Vine. Do they spin it and get promotoed? Something else?
I'd love a blog or coffee table book of "where are they now" for the director level folks who do dumb shit like this.
Building fiefdoms elsewhere, or whatever. "Have you heard about the new hire from XYZ?"
Why do you think they have not trained a new model since 4o? You think the GPT-5 release is /just/ routing to differently sized 4o models?
they're incorrect about the routing statement but it is not a newly trained model
> It's also worth noting that OpenAI has not trained a new model since gpt4o (all subsequent models are routing systems and prompt chains built on top of 4), so the idea of OpenAI being stuck in some kind of runaway training expense is not real.
This isn't really accurate.
Firstly, GPT4.5 was a new training run, and it is unclear how many other failed training runs they did.
Secondly "all subsequent models are routing systems and prompt chains built on top of 4" is completely wrong. The models after gpt4o were all post-trained differently using reinforcement learning. That is a substantial expense.
Finally, it seems like GPT5.2 is a new training run - or at least the training cut off date is different. Even if they didn't do a full run it must have been a very large run.
>It's also worth noting that OpenAI has not trained a new model since gpt4o (all subsequent models are routing systems and prompt chains built on top of 4)
At the very least they made GPT 4.5, which was pretty clearly trained from scratch. It was possibly what they wanted GPT-5 to be but they made a wrong scaling prediction, people simply weren't ready to pay that much money.
People would have paid it if it was actually significantly better. It was a huge cost increase for a pretty minor performance increase.
> The fact is nobody has any idea what OpenAI's cash burn is.
Their investors surely do (absent outrageous fraud).
> For all we know, they could be accumulating capital to weather an AI winter.
If they were, their investors would be freaking out (or complicit in the resulting fraud). This seems unlikely. In point of fact it seems like they're playing commodities market-cornering games[1] with their excess cash, which implies strongly that they know how to spend it even if they don't have anything useful to spend it on.
[1] Again c.f. fraud
> For all we know, they could be accumulating capital to weather an AI winter.
Right, this is nonsense. Even if investors wanted to be complicit in fraud, it's an insane investment. "Give us money so we can survive the AI winter" is a pitch you might try with the government, but a profit-motivated investor will... probably not actually laugh in your face, but tell you they'll call you and laugh about you later.
They have not successfully trained a new model since 4o. That doesn’t mean they haven’t burned a pile of cash trying.
I know sama says they aren’t trying to train new models, but he’s also a known liar and would definitely try to spin systemic failure.
Gpt-5.2 was new pretrain run i believe
lol the typical AI boosters are down voting you.
Well we do know their consumption of energy is not insignificant and comes at great cost.
How are they updating the data then? Wouldn’t the cutoff date be getting further away from today?
RAG? Even for a "fresh" model, there is no way to keep it up to date, so there has to be a mechanism by which to reference eg last night's football game.
They’re just feeding a little bit of slop in every so often. Fine tuning rather than training a new one.
they're paying million dollar salaries to engineers and building data centers, it's not a huge mystery where their expenses are
> they could be accumulating capital to weather an AI winter
Doubtful. This would be the very antithesis of the Silicon Valley way.
wasn't 4.5 new
Yes it was, op didn't read the reporting closely enough. It said something to the effect of "Didn't pretrain a new broadly released, generally available model"
That's what I meant, though. That's the expensive part.
Wasn't 4.5 before 4o?
How much of this capital is cheap printed credit from the covid area?
A second, less likely bubble?: IP rights enforcement. While the existing content hosters might have a neatly sewn up content agreement with their users such that all their group chats and cat photos can be used for training, I am a lot less confident that OAI came by its training data legitimately.
(Adjacent to this is how crazy it was that Meta were accused of torrenting ebooks. Did they need them for the underlying knowledge? I can’t imagine they needed them for natural langauge examples.)
The best case I can see is they integrate shopping and steal the best high-intent cash cow commercial queries from G. It's not really about AI, it's about who gets to be the next toll road.
Google already puts AI summaries at the top of search. It would be trivial for them to incorporate shopping. And they have infinitely more traffic than OpenAI does. I just don’t see how OpenAI could possibly compete with that. What are you seeing that I’m not?
ChatGPT has already won a lot of people away from Google like my mum, who now defaults to ChatGPT when she has a question. I was just talking to one of their friends last night who is in his 90s and he loves using Perplexity to learn about cooking and gardening.
A lot of people now reach for ChatGPT by default instead of Google, even with the AI summaries. I wonder whether they just prefer the interface of the chat apps to Google that can be a bit cluttered in comparison.
I think I super important aspect that people are overlooking, is that every VC wants to invest in the next "big" AI company, and the probability is in your favor to only give funding to AI companies, bc any one of them could be the next big thing. I think, with a downturn of VC investment, we will see some more investment in companies that arent AI native, but use AI as a tool in the toolbox to deliver insights.
In other words, OpenAI is a freight train without functioning brakes. There I said it.
And thats brave.
Archive/Paywall: <https://archive.is/rHPk3>
thank you!
There is no doubt that OpenAI is taking a lot of risks by betting that AI adoption will translate into revenues in the very short term. And that could really happen imo (with a low probability sure, but worth the risk for VCs? Probably).
It's mathematically impossible what OpenAI is promising. They know it. The goal is to be too big to fail and get bailed out by US taxpayers who have been groomed into viewing AI as a cold war style arms race that America cannot lose.
> The goal is to be too big to fail and get bailed out by US taxpayers
I know this is the latest catastrophizion meme for AI companies, but what is it even supposed to mean? OpenAI failing wouldn’t mean AI disappears and all of their customers go bankrupt, too. It’s not like a bank. If OpenAI became insolvent or declared bankruptcy, their intellectual property wouldn’t disappear or become useless. Someone would purchase it and run it again under a new company. We also have multiple AI companies and switching costs are not that high for customers, although some adjustment is necessary when changing models.
I don’t even know what people think this is supposed to mean. The US government gives them money for something to prevent them from filing for bankruptcy? The analogy to bank bailouts doesn’t hold.
I think what Altman is looking at is becoming so codependent with NVidia and Microsoft that they'll all go down together, meaning the US government would have to deal with the biggest software company and the biggest chip company both imploding together.
If you look at the financial crisis, the US government decided to bail out AIG, after passing on Bear Sterns, because big banks like Goldman Sachs and Morgan Stanley (and even Jack Welch's General Electric) all had huge counterparty risk with AIG.
>I know this is the latest catastrophizion meme for AI companies, but what is it even supposed to mean?
Someone else put it succintly.
"When A million dollar company fails, it's their problem. When a billion dollar company fails, it's our problem"
In essence, there's so much investment in AI that it's a significant part of the US GDP. If AI falters, that is something that the entire stock market will feel, and by effect, all Americans. No matter how detached from tech they are. In other words, the potential for the another great depression.
In that regard, the government wants to avoid that. So they will at least give a small bailout to lessen the crash. But more likely (as seen with the Great Financial Crisis), they will likely supply billions upon billions to prop up companies that by all business logic deserved to fail. Because the alternative would be too politically damaging to tolerate.
----
That's the theory. These all aren't certain and there are arguments to suggest that a crash in AI wouldn't be as bad as any of the aforementioned crashes. But that's what people mean by "become too big to fail and get bailed out".
The closest analogy is the dot-com crash and there really wasn't any bailout for that, despite the short term GDP impact. And billion-dollar companies were involved back in the day too, like Apple, Microsoft, Amazon, Ebay etc. etc.
OpenAI isn't a publicly-traded company though, how will it going to zero affect the stock market?
Its about "animal spirits"
The stock market isn't rational, its a room full of people talking loudly, and moving to various tables.
All it takes is someone outside the room to shout something that triggers panic, and most of the people in the room will run for the exit.
OpenAI collapses and MSFT tanks. Microsoft shareholders aren't quite that dumb.
And that's ignoring the dominoes of other AI firms being pulled out of because OpenAi falters.
> Microsoft shareholders aren't quite that dumb.
If they aren't dumb, why are they investing in MSFT now then if it's a bubble that's doomed to fail? And even in the worst case scenario, a 10-15% decline in the S&P 500 won't trigger the next Great Depression. (Keep in mind that we already had a ~20% drawdown in public equities during the interest rate hikes of 2022/2023 and the economy remained pretty robust throughout.)
Like I said, they aren't "that" dumb. They are playing a risky game, but when they see the number go down rapidly they will pull. Which will make the line go down even faster.
>And even in the worst case scenario, a 10-15% decline in the S&P 500 won't trigger the next Great Depression
Only if you believe the 10% decline won't domino and that the S&P500 is secluded from the rest of the global economy. I wish I shared your optimism.
> and the economy remained pretty robust throughout.
Yeah and we voted the person who orchestrated that out. We don't have the money to pump trillions back in a 2nd time in such a short time. Something's gonna give, and soon.
> Only if you believe the 10% decline won't domino and that the S&P500 is secluded from the rest of the global economy. I wish I shared your optimism.
So your hypothesis is that a 10% decline in the S&P 500 will trigger the next Great Depression, i.e. years of negative GDP growth and unemployment? I agree that it could cause a slight economic slowdown, but I don't think AI and tech stocks are a large enough part of the economy to cause a Great Depression-style catastrophe.
>So your hypothesis is that a 10% decline in the S&P 500 will trigger the next Great Depression, i.e. years of negative GDP growth and unemployment?
Yup. I won't say it's the only factor, nor biggest. But I'm focusing on this topic and not 40+ years of government economic abandonment of the working class. It's the straw that will break the camel's back.
> Someone would purchase it and run it again under a new company.
That happened a long time ago! Microsoft already owns the model weights!
> If OpenAI became insolvent or declared bankruptcy, their intellectual property wouldn’t disappear or become useless
Yes but with all stock growth being in AI companies it would tank the market for one. Secondly, all of those dollars they are using are backed by creditors who would have a default. short of another TARP (likely IMO, the US NEEDS to keep pumping AI to compete with China) .... it could scare investors off too..
Plus with the growth in AI effecting the overall makeup of the stockmarket, something like this hurts every Americans 401k
> It's mathematically impossible what OpenAI is promising
Citation is needed
Don’t that have to make more money in the next 10 years than any company ever has… and that is just to break even.
It’s going to crash, guaranteed
It is the term "mathematically impossible" that caught my attention. Since it is about the future promise of OpenAI, one could debate the likelihood or "statistically improbable", but "mathematically impossible" implies some calculation, proof and certainty. Hence my curiosity.
I've seen some calculation I think from an HSBC analyst that it would take a monthly subscription of $200/mo. from some large portion of the US population for some insane number of years to break even.
> from some large portion of the US population
What a silly calculation.
OpenAI’s customer base is global. Using US population as the customer base is deliberately missing the big picture. The world population is more than 20X larger than the US population.
It’s also obvious that they’re selling heavily to businesses, not consumers. It’s not reasonable to expect consumers to drive demand for these services.
>OpenAI’s customer base is global.
I'd be willing to bet that, like many US websites, OpenAI's users are at lest 60% American. Just because there's 20x more people out there doesn't mean they have the same exposure to American products.
For instance, China is an obvious one. So that's 35%+ of the population already mostly out of consideration.
>It’s also obvious that they’re selling heavily to businesses, not consumers.
I don't think a few thousand companies can outspend 200m users paying $200 a month. I won't call it a "mathematical impossibility", but the math also isn't math-ing here.
Even if you grant that OpenAI might be as successful as Apple at international expansion and support, that’s still only a non-US market about double the size of the US market.
> OpenAI’s customer base is global.
Since when is English everyone's primary language?
ChatGPT is totally fluid in German and French (i.e.), their market size is my no means limited to the anglosphere.
Bailing out OAI would be entirely unnecessary (crowded field) and political suicide (how many hundreds of billions that could have gone to health care instead?)
If it happens in the next 3 years, tho, and Altman promises enough pork to the man, it could happen.
This administration has "committed political suicide" dozens of times this year. What's one more to add to the pile?
>Bailing out OAI would be ... political suicide (how many hundreds of billions that could have gone to health care instead?)
Not that I have an opinion one way or another regarding whether or not they'd be bailed out, but this particular argument doesn't really seem to fit the current political landscape.
Unlikely, Elon bought the presidency and owns a competitor.
on the one hand, i understand you are making a stylized comment, on the other hand, as soon as i started writing something reasonable, i realized this is an "upvote lame catastrophizing takes about" (checking my notes) "some company" thread, which means reasonable stuff will get downvoted... for example, where is there actual scarcity in their product inputs? for example, will they really be paying retail prices to infrastructure providers forever? is that a valid forecast? many reasonable ways to look at this. even if i take your cynical stuff at 100% face value, the thing about bailouts is that they're more complicated than what you are saying, but your instinct is to say they're not complicated, "grooming" this and "cold war" that, because your goal is to concern troll, not advance this site's goal of curiosity...
They've already spent so much money that even if they get any new hardware at a deep discount they will have a very hard time breaking even
Apparently we all have enough money to put it into OpenAI.
Some players have to play, like google, some players want to play like USA vs. China.
Besides that, chatting with an LLM is very very convincing. Normal non technical people can see what 'this thing' can already do and as long as the progress is continuing as fast as it currently is, its still a very easy to sell future.
> Some players have to play, like google
I don't think you have the faintest clue of what you're talking about right now. Google authored the transformer architecture, the basis of every GPT model OpenAI has shipped. They aren't obligated to play any more than OpenAI is, they do it because they get results. The same cannot be said of OpenAI.
Correction: OpenAI investors do take that risk. Some of the investors (e.g. Microsoft, Nvidia) dampen that risk by making such investment conditioned on boosting the investor's own revenue, a stock buyback of sorts.
I don't see a bubble, I see a rapidly growing business case.
MS Office has about 345 million active users. Those are paying subscriptions. IMHO that's roughly the totally addressable market for OpenAI for non coding users. Coding users is another few 20-30 million.
If OpenAI can convert double digit percentages of those onto 20$ and 50$ per month subscriptions by delivering good enough AI that works well, they should be raking in cash by the billions per month adding up to close to the projected 2030 cash burn per year. That would be just subscription revenue. There is also going to be API revenue. And those expensive models used for video and other media creation are going to be indispensable for media and advertising companies which is yet more revenue.
The office market at 20$/month is worth about 82 billion per year in subscription revenue. Add maybe a few premium tiers to that at 50$/month and 100$/month and that 2030 130 billion per year in cash burn suddenly seems quite reasonable.
I've been quite impressed with Codex in the last few months. I only pay 20$/month for that currently. If that goes up, I won't loose sleep over it as it is valuable enough to me. Most programmers I know are on some paid subscription to that, Anthropic's Claude, or similar. Quite a few spend quite a bit more than that. My Chat GPT Plus subscription feels like really good value to me currently.
Agentic tooling for business users is currently severely lacking in capability. Most of the tools are crap. You can get models to generate text. But forget about getting them to format that text correctly in a word processor. I'm constantly fixing bullets, headings and what not in Google docs for my AI assisted writings. Gemini is close to ff-ing useless both with the text and the formatting.
But I've seen enough technology demos of what is possible to know that this is mostly a UX and software development problem, not a model quality problem. It seems companies are holding back from fully integrating things mainly for liability reasons (I suspect). But unlocking AI value like that is where the money is. Something similarly useful as codex for business usage with full access to your mail, drive, spread sheets, slides, word processors, CRMs, and whatever other tools you use running in YOLO mode (which is how I use codex in a virtual machine currently, --yolo). That would replace a shit ton of manual drudgery for me. It would be valuable to me and lots of other users. Valuable as in "please take my money".
Currently doing stuff like this is a very scary thing to do because it might make expensive/embarrassing mistakes. I do it for code because I can contain the risk to the vm. It actually seems to be pretty well behaved. The vm is just there to make me feel good. It could do all sorts of crazy shit. It mostly just does what I ask it to. Clearly the security model around this needs work and instrumentation. That's not a model training problem though.
Something like this for business usage is going to be the next step in agent powered utility that people will pay for at MS office levels of numbers of users and revenue. Google and MS could do it technically but they have huge legal exposure via their existing SAAS contracts and they seem scared shitless of their own lawyers. OpenAI doing something aggressive in this space in the next year or so is what I'm expecting to happen.
Anyway, the bubble predictors seem to be ignoring the revenue potential here. Could it go wrong for OpenAI? Sure. If somebody else shows up and takes most of the revenue. But I think we're past the point where that revenue is not looking very realistic. Five years is a long time for them to get to 130 billion per year in revenue. Chat GPT did not exist five years ago. OpenAI can mess this up by letting somebody else take most of that revenue. The question is who? Google, maybe but I'm underwhelmed so far. MS, seems to want to but unable to. Apple is flailing. Anthropic seems increasingly like an also ran.
There is a hardware cost bubble though. I'm betting OpenAI will get a lot more bang for its buck in terms of hardware by 2030. It won't be NVidia taking most of that revenue. They'll have competition and enter a race to the bottom in terms of hardware cost. If OpenAI burning 130 billion per year, it will probably be getting a lot more compute for it than currently projected. IMHO that's a reasonable cost level given the total addressable market for them. They should be raking in hundreds of billions by then.
Whoever has the most compute will ultimately be the winner. This is why these companies are projecting hundreds of billions in infrastructure spend.There is a hardware cost bubble though. I'm betting OpenAI will get a lot more bang for its buck in terms of hardware by 2030. It won't be NVidia taking most of that revenue.With more compute, you can train better models, serve them to more users, serve them faster. The more users, the more compute you can buy. It's a run away cycle. We're seeing only 3 (4 if you count Meta) frontier LLM providers left in the US market.
Nvidia's margins might come down by 2030. It won't stay in the 70s. But the overall market can expand quicker than Nvidia's profits shrink so that they can be more profitable in 2030 despite lower market share.
Surprised they burn cash advertising on Reddit to “make a mini me” version of yourself where you hold your body in your hand. What a waste of AI lol
I believe in OpenAI but they were running turn your cat into studio ghibili ads on Reddit up until a week or 2 ago.
They need a better marketing strategy.
I would call it the ELIZA bubble. https://en.wikipedia.org/wiki/ELIZA_effect
In a parallel universe, governments invest in the compute/datacenters (read: infra), and let model makers compete on the same playing field.
I’d rather stay far away from this parallel universe.
Why would you want my money to be used to build datacenter that won’t benefit me ? I might use a LLM once a month, many people never use it.
Let the one who use it pay for it.
You are already paying for several national lab HPC centers. These are used for government/university research - no idea if commercial interests can rent time on them. The big ones are running weather, astronomy simulations, nuclear explosions, biological sequencing, and so on.
The biggest run classified nuclear stockpile loads, at least in the US. They cost about half a billion apiece. And are 30 (carefully cooled and cabled) megawatts. https://en.wikipedia.org/wiki/El_Capitan_(supercomputer)
No chance they're going to take risks to share that hardware with anyone given what it does.
The scaled down version of El Capitan is used for non-classified workloads, some of which are proprietary, like drug simulation. It is called Tuolumne. Not long ago, it was nevertheless still a top ten supercomputer.
Like OP, I also don't see why a government supercomputer does it better than hyperscalers, coreweave, neoclouds, et al, who have put in a ton of capital as even compared to government. For loads where institutional continuity is extremely important, like weather -- and maybe one day, a public LLM model or three -- maybe. But we're not there yet, and there's so much competition in LLM infrastructure that it's quite likely some of these entrants will be bag holders, not a world of juicy margins at all...rather, playing chicken with negative gross margins.
>The big ones are running weather, astronomy simulations, nuclear explosions, biological sequencing, and so on.
these things constitute public goods that benefit the individual regardless of participation.
Many more people materially benefit from e.g. good weather forecasts than form video slop generation.
if datacenters are built by the government, then i think it's fair to assume there will be some level of democratic control of what those datacenters will be used for.
What's the democratic control of existing resources? I would make the opposite assumption, it would be captured by the wealthiest interests.
This is literally the current system... adding more democratic controls is a good thing, the alternative is that only rich control these systems and would you look at it only the rich control these systems.
Uncanny really.
Certainly! Your congress representatives would be voting on how to allocate its computing power. (Do you remember who did you vote for last time?)
Sure. Same for healthcare and education right? If you don't have a child or need medical attention, why should you pay for them?
That's like every government initiative. Same as healthcare? School? I mean if you don't have children why do you pay taxes... and roads if you don't drive? I mean the examples are so many... why do you bring this argument that if it doesn't benefit you directly right now today, it shouldn't be done?
There are arguments aplenty that schooling and a minimum amount of healthcare are public goods, as are roads built on public land (the government owns most roads after all).
What is the justification for considering data centers capable of running LLMs to be a public good?
There are many counter examples of things many people use but are still private. Clothing stores, restaurants and grocery stores, farms, home appliance factories, cell phone factories, laundromats and more.
Libraries with books are likely considered public goods right?
Why not an LLM datacenter if it also offers information? You could say it's the public library of the future maybe.
Not all libraries are publicly owned or accessible. Most are run by local municipalities because they wouldn't exist otherwise.
Data centers clearly can exist without being owned by the public.
So can bookstores.
a distinction: the data centers have become the means of production, unlike clothing from a store
How is that distinct from any of my other examples which listed factories? Very few factories in the US are publicly owned; citing data centers as places of production merely furthers the argument that they should remain private.
Healthcare, schools, roads, generative AI. One of these things is not like the others.
We gave incentives to broadband, why not generative AI?
Last-mile services like roads, electricity, water, and telecommunications are natural monopolies. Normal market forces fail somewhat and you want some government involvement to keep it running smoothly.
This is not at all true of generative AI.
I have no idea why you're being downvoted because you're right. The entire point of taxation is to spread the cost among everyone, and since everyone doesn't utilise every government service every tax payer ends up paying for stuff they don't use. That like, the whole point...
If that did happen, how would the government then issue those resources?
OpenAI ask for 1m GPUs for a month, Anthropic ask for 2m, the government data center only has 500,000, and a new startup wants 750,000 as well.
Do you hand them out to the most convincing pitch? Hopefully not to the biggest donor to your campaign.
Now the most successful AI lab is the one that's best at pitching the government for additional resources.
UPDATE: See comment below for the answer to this question: https://news.ycombinator.com/item?id=46438390#46439067
National HPC labs have been over subscribed for decades with extensive queueing/time sharing allocation systems.
It would still likely devolve into most-money-wins, but it is not an insurmountable political obstacle to arrange some sort of sharing.
Edit: I meant to say over subscribed, not over provisioned. There are far more jobs in the queue than can be handled at once
Huh, TIL - thanks for the correction.
https://www.ornl.gov/news/doe-incite-program-seeks-2026-prop...
> The Innovative and Novel Computational Impact on Theory and Experiment, or INCITE, program has announced the 2026 Call for Proposals, inviting researchers to apply for access to some of the world’s most powerful high-performance computing systems.
> The proposal submission window runs from April 11 to June 16, 2025, offering an opportunity for scientific teams to secure substantial computational resources for large-scale research projects in fields such as scientific modeling, simulation, data analytics and artificial intelligence. [...]
> Individual awards typically range from 500,000 to 1,000,000 node-hours on Aurora and Frontier and 100,000 to 250,000 node-hours on Polaris, with the possibility of larger allocations for exceptional proposals. [...]
> The selection process involves a rigorous peer review, assessing both scientific merit and computational readiness. Awards will be announced in November 2025, with access to resources beginning in 2026.
Not sure OpenAI/Anthropic etc would be OK with a six month gap between application and getting access to the resources, but this does indeed demonstrate that government issued super-computing resources is a previously solved problem.
Well, people bid for USA government resources all the time. It's why the Washington DC suburbs have some of the country's most affluent neighborhoods among their ranks.
In theory it makes the process more transparent and fair, although slower. That calculus has been changing as of late, perhaps for both good and bad. See for example the Pentagon's latest support of drone startups run by twenty-year-olds.
The question of public and private distinctions in these various schemes are very interesting and imo, underexplored. Especially when you consider how these private LLMs are trained on public data.
In a completely alternate dimension, a quarter of the capital being invested in AI literally just goes towards making sure everyone has quality food and water.
I'd rather live in a universe where that money is taken out of the military budget.
you'll never win that argument, but I absolutely agree.
people have no idea about how big the military and defense budgets worldwide are next to any other example of a public budget.
throw as many pie charts out as you want; people just can't see the astronomical difference in budgets.
I think it's based on how the thing works; a good defense works until it doesn't -- the other systems/budgets in place have a bit more of a graceful failure. This concept produces an irrationality in people that produces windfalls of cash availability.
As we all know, throwing money at a problem solves it completely. Remember how Live Aid saved Ethiopia from starvation and it never had any problems again?
Without capital invested in the past we wouldn’t have almost anything of modern technology. That has done a lot more for everyone, including food affordability, than actually simply buying food for people to eat once.
Datacenters are not a natural monopoly, you can always build more. Beyond what the public sector itself might need for its own use, there's not much of a case for governments to invest in them.
That could make sense in some steady state regime where there were stable requirements and mature tech (I wouldn’t vote for it but I can see an argument).
I see no argument why the government would jump into a hype cycle and start building infra that speculative startups are interested in. Why would they take on that risk compared to private investors, and how would they decide to back that over mammoth cloning infra or whatever other startups are doing?
Given where we are posting, the motive is obvious: to socialize the riskiest part of AI while the investors retain all the potential upside. These people have no sense of shame so they'll loudly advocate for endless public risk and private rewards.
In a better parallel universe, we found a different innovation without using brute-force computation to train systems that unreliably and inefficiently compute things and still leaves us able to understand what we're building.
> governments invest in the compute/datacenters (read: infra), and let model makers compete on the same playing field
Hmm, what about member-owned coöperatives? Like what we have for stock exchanges.
Socialized losses, private profits
why would they do that? not to mention governments are already doing that indirectly by taking equity stakes in some of the companies.
Same reason they should own access lines: everyone needs rackspace/access, it should be treated like a public service to avoid rent seeing. Having a data center in every city where all of the local lines terminate into could open the doors to a lot of interesting use cases, really help with local resiliency/decentralization efforts, and provide a great alternative to cloud providers that doesn't break the bank.
should the government own all types of "public services"? e.g. search index, video serving infra, etc?
Public ownership of public services hmm?
Smells like socialism. Around here we privatize the profits and only socialize the costs. Like the impending bailout of the most politically connected AI companies.
That sounds like a nightmare.
Do you like this idea?
Prediction: on this thread you'll get a lot of talk about how government would slow things down. But when the AI bubble starts to look shaky, see how fast all the tech bros line up for a "public private partnership."
That seems like a terrible idea. Data centers aren’t a natural monopoly. Regulate the externalities and let it flourish.
Don't forget that the internet exists because of government agencies.
I'm not sure how that's relevant here.
Cool. Time for the government to seize Amazon.
That's malinvestment. Too much overhead, disconnected from long term demand. The government doesn't have expertise, isn't lean and nimble. What if it all just blows over? (It won't? But who knows?)
Everything is happening exactly as it should. If the "bubble" "pops", that's just the economic laws doing what they naturally do.
The government has better things to do. Geopolitics, trade, transportation, resources, public health, consumer safety, jobs, economy, defense, regulatory activities, etc.
Burn rate often gets treated as a hard signal, but it is mostly about expectations. Once people get used to the idea of cheap intelligence, any slowdown feels like failure, even if the technology is still moving forward. That gap is usually where bubbles begin.
why does the article used words like burn and incinerate, implying that OpenAI is somehow making money disappear or something? They’re spending it; someone is profiting here, even if it’s not OpenAI. Is it all Nvidia?
Because that's normal language idioms in financial analysis reporting?
Because typically one expect a return on investment with that level of spending. Not only have they run at a loss for years, their spending is expected to increase, with no path to profitability in sight.
not that I disagree but would it be fair to say though that we have seen this before where it turned out OK? say Uber? Amazon?
IIRC, current estimates are that OpenAI is losing as much money a year as Uber or Amazon lost in their entire lifetime of unprofitability. Also, both Uber and Amazon spent their unprofitable years having a clear roadmap to profitability. OpenAI's roadmap to profitability is "???"
I have lived through Amazon’s rags to riches and there was never a clear plan to profitability. Vast majority of people were questioning sanity of anyone investing in Amazon.
I am not saying OpenAI is Amazon but am saying I have seen this before where masses are going “oh business is bad, losses are huge, where is path to profitability…”
Your recollection is hazy. Bezos chose not to be profitable in order to grow the company, and reap greater rewards in the future. https://www.youtube.com/shorts/wjLs22dNOCE
I don't know either way but every company does this, it's not saying anything meaningful to say that a new company is taking an "investment year" or two or ten.
I do know that in the late aughts, people were writing stories about how Amazon was a charity run on behalf of the American consumer by the finance industry.
To become the next Uber, do I just need to run huge losses?
I wouldn’t but path to success can clearly come from running 10-digit losses for a loooong time, no?
I think you're saying that just running up huge losses is sufficient to create a successful company? But that you personally wouldn't want to run up huge losses? Not sure.
nah, I am saying that many (super) successful businesses ran in red financially for a very long time. I would not run a business that way but I am also (fortunately) not a CEO of a multibillion dollar company
To my knowledge Amazon never debt financed their ops like this
Amazon did borrow money, for a long time.
Where did their financing come from then?
They had a cash-cow called AWS to keep the retail business afloat
Somebody must not be old enough to remember Amazon before AWS. Maybe you also don’t remember that Amazon started selling books before becoming the world’s largest fencing for selling stolen merch. They used to be the butt of many jokes for losing so much money for so many years while they expanded warehouses.
not right away
I suspect most of it is going to utilities for power, water and racking.
That being said, if I was Sam Altman I'd also be stocking up on yachts, mansions and gold plated toilets while the books are still private. If there's $10bn a year in outgoings no one's going to notice a million here and there.
How many gold toilets do you need? I mean, I don't even own one.
Tragically I don't make CEO money so I also don't have one but I presume you'd want to have at least one per mansion and another one in the office. Maybe a separate one for special occasions.
Your burn is the money you spend that exceeds the money you earn, see also "burn rate".
“Burn rate” is a standard financial term for how much money a startup is losing. If you have $1 cash on hand and a burn rate of $2 a year, then you have six months before you either need to get profitable, raise more money, or shut down.
> They’re spending it
That's what the words mean in this context.
OpenAI has #5 traffic levels globally. Their product-market fit is undeniable. The question is monetization.
Their cost to serve each request is roughly 3 orders of magnitude higher than conventional web sites.
While it is clear people see value in the product, we only know they see value at today’s subsidized prices. It is possible that inference prices will continue their rapid decline. Or it is possible that OAI will need to raise prices and consumers will be willing to pay more for the value.
It's easy to get product-market fit when you give away dollars for the price of pennies.
Yes, but that is the standard methodology for startups in their boost phase. Burn vast piles of cash to acquire users, then find out at the end if a profitable business can be made of it.
It’s also the standard methodology for a number of scams.
Scams are our entire economy now. Do whatever you can to own a market, then squeeze your customers miserably once you have their loyalty. Cash out, kick the smoking remains of the company to the curb, use your payout to buy into another company, and repeat.
Kind of like paying more and more to Social Security and Medicare and getting less and less.
And the backstop on asset prices at the expense of the currency's purchasing power.
Most startups have big upfront capital costs and big customer acquisition costs, but small or zero marginal costs and COGS, and eventually the capital costs can slow down. That's why spending big and burning money to get a big customer base is the standard startup methodology. But OpenAI doesn't have tiny COGS: inference is expensive as fuck. And they can't stop capex spending on training because they'll be immediately lapped by the other frontier labs.
The reason people are so skeptical is that OpenAI is applying the standard startup justification for big spending to a business model where it doesn't seem to apply.
>But OpenAI doesn't have tiny COGS: inference is expensive as fuck.
No, inference is really cheap today, and people saying otherwise simply have no idea what they are talking about. Inference is not expensive.
Clearly not cheap enough.
> Even at $200 a month for ChatGPT Pro, the service is struggling to turn a profit, OpenAI CEO Sam Altman lamented on the platform formerly known as Twitter Sunday. "Insane thing: We are currently losing money on OpenAI Pro subscriptions!" he wrote in a post. The problem? Well according to @Sama, "people use it much more than we expected."
So just raise the price or decrease the cost per token internally.
Altman also said 4 months ago:
https://simonwillison.net/2025/Aug/17/sam-altman/Most of what we're building out at this point is the inference [...] We're profitable on inference. If we didn't pay for training, we'd be a very profitable company.
Does that cost to serve multiple stay the same when conventional sites are forced to shovel ai into each request? e.g. the new google search
it's a simple problem really. what is actually scarce?
a spot on the iOS home screen? yes.
infrastructure to serve LLM requests? no.
good LLM answers? no.
the economist can't tell the difference between scarcity and real scarcity.
it is extremely rare to buy a spot on the iOS home screen, and the price for that is only going up - think of the trend of values of tiktok, whatsapp and instagram. that's actually scarce.
that is what openai "owns." you're right, #5 app. you look at someone's home screen, and the things on it are owned by 8 companies, 7 of which are the 7 biggest public companies in the world, and the 8th is openai.
whereas infrastructure does in fact get cheaper. so does energy. they make numerous mistakes - you can't forecast retail prices Azure is "charging" openai for inference. but also, NVIDIA participates in a cartel. GPUs aren't actually scarce, you don't actually need the highest process nodes at TSMC, etc. etc. the law can break up cartels, and people can steal semiconductor process knowledge.
but nobody can just go and "create" more spots on the iOS home screen. do you see?
depends if they can monetize that spot. So either ads or subscription. It is as yet unclear whether ads/subscription can generate sufficient revenue to cover costs and return a profit. Perhaps 'enough ads' will be too much for users to bear, perhaps 'enough subscription' will be too much for users to afford.
right now google pays apple almost $30b a year to be default search in safari. google only has one icon on the home screen (YouTube). just originating google searches could be worth tens of billions. so i don't know. there are a bajillion ways to monetize.
2008: US Banks pump stocks -> market correction -> taxpayer bailout
2026: US AI companies pump stocks -> market correction -> taxpayer bailout
Mark my words. OpenAI will be bailed out by US taxpayers.
In 2008 the US government ended up making more money then they spent though (at least with the tarp), because they invested a ton of money after everything collapsed, and thus was extremely cheap. Once the markets recovered, they made a hefty sum selling all the derivatives they got at the lowest point. Seems like the epitome of buy when low and sell when high tbh.
I'll take that bet.
Banks get bailed out because if confidence in the banking system disappears and everyone tries to withdraw their money at once, the whole economy seizes up. And whoever is Treasury Secretary (usually an ex Wall Street person) is happy to do it.
I don't see OpenAI having the same argument about systemic risk or the same deep ties into government.
Even in a bank bailout, the equity holders typically get wiped out. It's really not that different from a bankruptcy proceeding, there's just a whole lot more focus on keeping the business itself going smoothly. I doubt OpenAI want to be in that kind of situation.
Even if there is a bailout. Will it happen in time? Once the confidence is lost it is lost and valuations have dropped. Bailout would just mean that who ever gave money would end up as bag holder of something now worth lot less.
Banks needed bailout to keep lending money. Auto industry needed one to keep employing lot of people. AI doesn't employ that many.
I just don't believe bailout can happen before it is too late for it to be effective in saving the market.
Not really. It was not about stocks. It was the collapse of insurance companies at the core of 2008 crisis.
The same can happen now on the side of private credit that gradually offloads its junk to insurance companies (again):
As a result, private credit is on the rise as an investment option to compensate for this slowdown in traditional LBO (Figure 2, panel 2), and PE companies are actively growing the private credit side of their business by influencing the companies they control to help finance these operations. Life insurers are among these companies. For instance, KKR’s acquisition of 60 percent of Global Atlantic (a US life insurer) in 2020 cost KKR approximately $3billion.
https://www.imf.org/en/Publications/global-financial-stabili...
I don't think so.
Elon owns a competitor and bought the White House.
For what I use them for, the LLM market has become a two player game, and the players are Anthropic and Google. So I find it quite interesting that OpenAI is still the default assumption of the leader.
Only in HN and some reddit subs I even see the name claude. In many countries AI=ChatGPT.
From what I've seen, 99% of people are using the free version of ChatGPT. Those who are using Claude are on the subscription, very often the $100/month one.
And at one point in the 90s, Internet=Netscape Navigator.
I see Google doing to OpenAI today what Microsoft did to Netscape back then, using their dominant position across multiple channels (browser, search, Android) to leverage their way ahead of the first mover.
That's funny, the way I see it is Microsoft put tens of billions of dollars behind an effort to catch Google on the wrong foot, or at least make Google look bad, but they backed the wrong guy and it isn't quite going to make it to orbit.
ChatGPT dominates the consumer market (though Nano Banana is singlehandedly breathing some life into consumer Gemini).
A small anecdote: when ChatGPT went down a few months ago, a lot of young people (especially students) just waited for it to come back up. They didn't even think about using an alternative.
When ChatGPT starts injecting ads or forcing payment or doing anything else that annoys its userbase then the young people won't have a problem looking for alternatives
This "moat" that OpenAI has is really weak
They took early steps to do so (ads) just recently. User response was as you'd expect.
The consumer market is a loss leader.
That's pretty nuts. With the models changing so much and so often, you have to switch it up sometimes just to see what the other company is offering.
How often do you or people you know use a search engine other than google?
That is different because all of the players I mentioned have credible, near-leading products in the AI model market, whereas nobody other than Google has search results worth a damn. I wouldn't recommend anyone squander their time by checking Kagi or DDG or Bing more than once.
I don't use google. Believe it or not, I get better results via Bing (usually via DDG, which is a frontend for Bing). But I asked the rhetorical question expecting the answer you gave. These people use ChatGPT only for the same reason you exclusively use Google.
What do you use them for?
GPT5.2 Codex is the best coding model right now in benchmarks. I use it exclusively now.
codex cli with gpt-5.2-codex is so reliably good, it earns the default position in my book. I had cancelled my subscription in early 2024 but started back up recently and have been blown away at how terse, smart, and effective it is. Their CLI harness is top-notch and it manages to be extremely efficient with token usage, so the little plan can go for much of the day. I don’t miss Claude’s rambling or Gemini’s random refactorings.
Interestingly Claude is so far down in traffic it's below things like CharacterAI, it's the best model but it's something like 70% ChatGPT, 10% Gemini and Claude is only 1% or so
On the radio they mentioned that the total global chocolate market is ~100B, I googled it when I was home and it seems to be about ~135B. Apparently that is ... all chocolate, everywhere.. OpenAI's valuation is about 500B. Maybe going up to like 835B.
I'd love to see the rationale that OpenAI (not "AI" everywhere) is more valuable than chocolate globally.
... so crash early 2026?
Ignoring that those numbers aren't directly comparable, it did make me wonder, if I had to give up either "AI" or chocolate tomorrow, which would I pick?
Even as an enormous chocolate lover (in all three senses) who eats chocolate several times a week, I'd probably choose AI instead.
OpenAI has alternatives, but also I do spend more money on OpenAI than I do on chocolate currently.
I am just trying to help you write better. Your writing says "if I had to give up either AI or chocolate [...] I would probably choose AI". However, your language and intent seems to be that you would give up chocolate.
It’s a bit of a weird comparison, AI vs a luxury sweet.
Maybe instead of the chocolate market, look at the global washing machine market of $65 billion.
I’d rather give up both AI and chocolate than my washing machine.
If you really wanted to know you could stop eating chocolate or stop using ai and see if you break. Or do both at different times and see how long you last without one or the other.
Wait, aren't you comparing revenue and market cap?
People take old things for granted often. Explains the Coolidge effect, and why plenty of people cheat.
yes. stock v flow error again ("company X cap bigger than country Y GDP" another all too common version).
I love AI, and ChatGPT has been transformative for me. But would I give it up for chocolate? I honestly don't think I could.
I spend a lot more time using AI for work than I do eating chocolate
It’s like watching a train wreck waiting to happen, while everyone praise the madman conduct or who’s operating the locomotor at full steam.
I don't think that's diagnosis (as a clinical term); closer to defamation.
Is it necessary to a point you want to make?
You can just point to behavior of a given entity, such as to conclude it's untrustworthy, without the problematic area of armchair psychoanalysis.
I redacted the comment because you’re right. I need a better form to express the point.
Well they got $40B more to burn lol
The Microsoft effect, everything they touch turns to shit
What does it mean for the AI bubble to pop? Everyone stops using AI en masse and we go back to the old ways? Cloud based AI no longer becomes an available product?
I think it mostly just means a few hundred billion dollars of value wiped from the stock market - all the models that have been trained will still exist, as well as all the datacentres, even if the OpenAI entity itself and some of the other startups shut down and other companies else get their assets for pennies on the dollar.
But it might mean that LLMs don't really improve much from where they are today, since there won't be the billions of dollars to throw at training for small incremental improvements that consumers mostly don't care to pay anything for.
It probably means a price correction on cloud based AI, rather than it disappearing entirely.
It's a good question. Sad that you're downvoted.
paywall, no upvote
Someone posted already the non-paywall version: https://news.ycombinator.com/item?id=46438679
The comparison to railroad bubble economics is apt. OpenAI's infrastructure costs are astronomical - training runs, inference compute, and scaling to meet demand all burn through capital at an incredible rate.
What's interesting is the strategic positioning. They need to maintain leadership while somehow finding a sustainable business model. The API pricing already feels like it's in a race to the bottom as competition intensifies.
For startups building on top of LLM APIs, this should be a wake-up call about vendor lock-in risks. If OpenAI has to dramatically change their pricing or pivot their business model to survive, a lot of downstream products could be impacted. Diversifying across multiple model providers isn't just good engineering - it's business risk management.
Railroad peaked at 6% of GDP in the US.
AI is at 1% of total US GDP right now.
We have 6x more to go.