I know a traditional SaaS company I worked for that IPO’d years ago and still has no signs that they can be profitable (and many others like it) and nobody seems particularly concerned.
the market cap of a company is computed by the current price of a company's shares, the last price paid; not all the shares of the company were bought at that price, the ones who got shares cheaper are showing paper profits, unrealized. Those who have already cashed out have money in their bank accounts that was transferred from people who wanted to get in. If the company goes bankrupt, their shares will be worthless, but the money they paid for them still remains in the accounts of people who sold their shares: the money was not lost even if some people lost money.
I'm not going to keep going through it but the reason it works to value things the way we do is that the values are comparable and they frequently work out, so snapshots of the economy and the participants are comparable. But "losses" are not like taking gold and feeding it into some deep fold in the earth where it will disappear into the molten middle of earth.
Stock valuations are "expectations for the future". Those expectations weren't money, they were lottery tickes where the lottery consisted of human creativity and human effort. People buying and selling share are moving real money around to trade the expectations. The money didn't go anywhere, it's still there, it's just that expectations for the future have been reduced. It all boils down to humans trading some of their time and potential on a bet that things work out. Some people's effort gets more rewarded than others. Not every team wins the world cup, but people like to play and like to watch.
And I think we passed the threshold for crash down for AI, even if AI companies wont be that profitable. Nvidia/cloud providers will be profitable as long as there is demand for AI.
AI usage seems to have plateaued overall [2], except for niche use cases like coding, that is why companies are forcing it on their employees to justify ROI [3] or creating "products" w/ AI features [4] or embedded addiction.
[1] https://news.ycombinator.com/item?id=48241012
[2] https://news.ycombinator.com/item?id=48179021
https://www.theinformation.com/articles/anthropic-openais-sh...
I sure hope more people think like this, because it's going to leave a lot of money on the table (for me)
The weird thing is that so many people believe that inference is unprofitable. There are large open weights models that companies run at a profit while charging far less than what OpenAI and Anthropic charge. Deepseek V4 just made their 75% off deal permanent and it was already very cheap.
Yes, you have to consider costs of training the models, but as usage grows it’s going to become a smaller and smaller part of the business.
I think we will see some data center businesses and AI companies blow up, but I think the people expecting the entire AI scene to blow up because prices quadruple are going to be disappointed.
You have no idea whether those companies are making a profit.
1. All it takes is one of them operating a loss to gain market share to force the other ones to lower prices to compete.
2. There’s not reason to expect that these relatively small companies are correctly pricing GPU depreciation.
Privatize Profits and Socialize Losses is now Bog-Standard Operating Procedure.
The stock market. Stocks crash, companies go belly-up, tons of people get laid off, unemployment spikes, people die. I don’t give a shit about the companies themselves. I do give a shit about who they employ, both directly and downstream, and the job market that will result from many of them losing their jobs.
In 10 years, we've spent nearly 3x the cost of the entire US interstate highway system on AI.
Some helpful visualizations: https://www.aljazeera.com/news/2026/2/19/visualising-ai-spen...
(Is there a more extreme example so far of this than AI companies, just in terms of raw losses? As far as I know, Netscape's lifetime losses as an independent company "only" total a bit over $100 million dollars, which is a lot, it just doesn't look like all that much when put into perspective...)
We can look at a “success story” like Uber and it is still net negative over its entire existence. This is a business that’s in a literal monopoly/duopoly status in most markets it operates in with vastly reduced regulatory burden compared to the industry disrupted. Literally the ideal scenario for printing money and yet it hasn’t made any. It’s the poster child for the unicorn exit that founders dream of.
The end result is that Uber and companies like it are a financial instruments that transfer dollars away from one set of investors to another set of investors.
If Uber hasn’t yet made its investment back, I struggle to wonder how some of these AI ventures will ever make that money back when their expenditures make Uber look like a small little side project.
Meta has spent almost 4 years worth of its net income for FY2025 on AI going by this website’s data, and counting.
We are decades since Web 2.0 took off, almost 20 years since the iPhone launched, 50 years of Apple Computer. Software isn’t some new industry anymore. There isn’t an industry left that hasn’t completed its digital transformation. These spray and pray economies would have died off years ago if it wasn’t for the fact that software companies have uniquely low cost structures where they don’t need to build factories or distribution networks to get their products to their customers. These low cost structures might just be concealing the fact that it’s not going to be a growth industry forever.
How has the sheer saturation of LLMs not resulted in profit? It has dominated the conversation, center stage, of every news outlet for like 4 years now. It is the most known-about thing currently out there.
And we haven't been able to convert that much captured attention into profitability yet? That seems... bad?
I do wonder why Nvida is included, though. If you include the company that all of the frontier models are pouring money into, of course the net (expenditure - profits) of the collective is going to be closer to zero :-)
If Nvidia is included, does that mean that the money Amazon, Microsoft, and Oracle get for selling compute to the frontier models are included in their revenue?
Because for Amazon in particular, the situation this pages shows is actually much WORSE than I expected. I thought they were making a killing selling compute for model training.
Given how the curves look like in terms of ramping of spend, these are very healthy numbers.
For example: I have index funds which have some of these stocks. So I, by process of revealed-preference, don't think it's a bubble, or I think I will keep my money in through the bubble's pop. I don't have that much else to say!
For the record: I would respect the creator of this site equally or more if he/she said, "I'm shorting these stocks and this is why."
No one really knows how quickly AI hardware investments will become obsolete and thus how long it should be amortized, but 2-3 years would be extremely conservative, and in fact used H100 (discontinued/2 generations old) prices are higher today than they were when the equipment was new several years ago.
I just wanna know how the OpenAI/Anthropic shell game works long-term. So both companies made equity deals with infrastructure providers; OpenAI on Azure, Anthropic on AWS, GCloud, and Colossus. They get a loan of compute credits and then pay for the compute with the credits. So the PaaS are effectively giving them free compute, then book it as revenue; and the AI provider lets them do inference and books that as revenue. So, it's like both types of company have a buffet, and let each other eat there for free. But somebody has to actually buy the pasta salad, with real dollars. Afaict, those real dollars are.... the cash reserves of the PaaS.
How long are they going to eat into that cash? Microsoft and AWS don't really have their own models, whereas Google and SpaceX do. And while Google has tons of cash, SpaceX is perpetually looking for cash. So the only player here that can actually afford to keep doing this, or leave the game entirely, is Google.
The frontier labs have fantastic margin on inference. You do not understand how fantastic. And they have license to change inputs at will based on profitability.
They are not only innovating on models and tooling, they are innovating on cogs (I wrote this btw, and I’m not going to stop writing this way because Claude discovered it’s brilliant).
Speaking of models, the cost of training is not scaling nearly as fast as demand for inference. Training used to be the biggest cost by far, now it’s not.
So margin is increasing, and guess what else is happening? Customers are finding value. And the customers that are finding value are also the ones who happen to have huge enterprise budgets.
And while this is happening, so is implicit collusion (and lock in, and hype, and all that). And so prices are going up.
They’re going to be just fine man, there is no inference bubble.
They can modulate supply. It’s all going to be fine. You should invest.
This. The gross margin on inference is at least 95% if not higher - several open weight models on my tiny consumer DGX Spark easily replace the 15 dollars a day I was paying in tokens for Claw usage with a dollar a day electricity. You add data centre overhead and depreciation, the theoretical net margin will trend lower but depreciation is always far more aggressive than actual product degradation. The old NVIDIA GPU on a 9 year old second hand gaming PC I bought still serves up a small Gemma 4 variant quite reasonably.
Source?
The OpenAi filing will be very interesting indeed.
("trust me bro" statements from sama et al does not count, since I don't trust them)
Edit:
The best argument I have seen look at the price of inference from smaller companies running open models. And assuming they are profitable-ish. Their prices are lower than the OpenAi and Anthropics best models, so maybe they do make money on inference (ignoring all other costs)
Where I can find confirmation of that in public sources?
Everyone’s long term plan is hoping that they build out and survive long enough that, in the end, the market accepts them.
Even your local still-unprofitable restaurant is burning their grandparents’ inheritance money hoping that it works out.
But on the other hand, that’s what Theranos, WeWork, and Pets.com tried too.
When OpenAI goes public it will initially get a tsunami of cash, but it'll also be open to new risk due to the different operating model and transparency. Anthropic might not make it to an S-1 (this year). Even if they got a $30B infusion of cash each, based on their current spending projections, it doesn't cover half of what they need just to break even. In the meantime the PaaS's are holding the bag (and shedding cash).
So where's it going to end? To me, all of this (combined with inflation, degrading of reserve currency, war in middle east) is spookily similar to the railroad panic of 1873. Over-investment in new technologies leveraging too much from the largest financial institutions resulting in prolonged economic crisis. Our only saving grace now are laws ensuring banks have to cover their end; if your money's FDIC/SIPC insured you're safe. But all the businesses and individuals who aren't safe are gonna take a bath, which'll have systemic ripples. Afaict, Google is the only player who can survive all that and come out with profitable AI. (But I'm sure I've missed something because it seems too obvious)
Either the actual revenue of paying customers ramp up or the bubble will pop at some point
I expect the paying customers will actually be companies buying ad, not people buying AI subscriptions
The only way to get consistently rich in any bubble economy.
Oh it doesn't fit the narrative. Never mind then.
You can turn your drive-by dismissal into something really informative if you want to.
Second, even if you take CEOs' words at face value, they didn't distinguish the capex for hardware, electricity, software and salary. You can make up whatever the percentage for hardware and the depreciation rate you believe and fit an arbitrary narrative.
Yet this site suggests that tokens are very unprofitable
Building a datacenter that will produce hundreds of billions of dollars worth of tokens over a multi-decade life shouldn't surprise anyone that it's in the red in year 1 or 2. There's a lot of front loaded capex in this business. If someone built a tractor factory you wouldnt expect 1 year payback.
But the site sort of implies that these companies are selling tokens for less than it takes to inference them. As if this is some sort of COGS ledger. Especially by throwing Nvidia in there. Don't take it too seriously.
Out of all the companies, considering their own silicon etc. I wouldn't be surprised. Though I do wonder in terms of total CapEx and R&D where it would be at...
They're soaking up the investor bonanza into AI - Gemini ain't making them money.
For context Cloud Compute made 20bn in Q1, Other services made 90bn.
Comparing to ad revenue from a company like meta, the story that Gemini tokens are a strong cash drag on Google just doesn't add up. It seems at worst they are losing like 50 cents/1M tokens (including r&d spend, data centers, etc..), and very possible they are actually profitable per token.
Which is much better than anthropic and openai.
Would be weird if they're raising $10 billion after spending only 0.3
https://newsletter.semianalysis.com/p/deepseek-debates
Probably more like 3-4 billion by now?
Sure they're torching money on building consumer LLMs, but they seem to be doing very well optimizing things like ad ranking
https://engineering.fb.com/2025/11/10/ml-applications/metas-...
https://engineering.fb.com/2026/03/31/ml-applications/meta-a...
1/ User targeting is complex - you can charge more for ads if the users you're showing the ads to click
2/ Ads impact user retention - you need to balance making money and keeping users around
3/ AI generated ads - this is a pretty big thing now, where instead of bringing your own media, you just describe your target audience and the AI will A/B test media + CTAs for you
4/ Integrity - you want to vet the ads against laws/site policies
Probably forgetting a few, but there's a reason the ad industry employs so many
1. Outspend and outlast your competition until you have market dominance. Win over and lock in your customers with sweetheart deals.
2. Enshittify and squeeze your customers to pay back your debt.
If you're using AI, you're not paying the true cost right now because we're in phase 1. Be ready for phase 2.
The big labs are actively moving into the application layer, where they’ll have more pricing power. Maybe that layer will end up with a Mac (Anthropic) vs Windows (OpenAI) vs Linux (open-source) dynamic as well if they can create a moat. But so far it’s pretty easy to move between providers.
In that case, AI companies will never get their money back, leading to a huge crash.
Only the later have something to lose if AI bubble gone by tomorrow. Everyone else will just stay with grown capacity and reuse infrastructure for whatever.
Not listing other hardware companies is just dishinest. AI is not a crypto mining where resources are just burned.
AI is exactly like crypto mining in that Nvidia is the one who profited from both
No matter what happen with the AI bubble text, image and video and other generative neural networks are here to stay.
Whatever you like it or not this tech already changed a lot of industries and there is no going back.
Dark fiber, for example, had a much more compelling use case.
Also many of these companies like Amazon, Google, and Meta drive a lot of incremental value due to both AI powered content suggestion and AI powered ad suggestion. Personalized ads has driven a ton of revenue.
https://www.cnbc.com/2026/05/20/anthropic-revenue-explosive-...
The startup blitz-scaling-market-capturing playbook makes makes sense when you spend to scale, not when you spend because you scale, yeah, I understand that step 2 is "and now you squeeze the users", but it will need to be by such a bigger factor...
The core bottlenecks are power and computing capacity, and they actually trace back to the exact same issue. It all comes down to the physical energy it takes to flip or move a single bit inside the ram or disk storage. This concept is subject to fundamental physical barriers.
There are a few ways to tackle this, like improving power efficiency, reducing model size, or pushing hardware further. However, achieving orders-of-magnitude improvement in any of these areas will cost a massive amount of time and money. I wonder if governments, corporations, and investors have the patience to wait for these tech breakthroughs.
This will always be negative for any new business as you are effectively depreciating the assets straight away. Like if you build a hotel and deduct the cost of building it from room income - it would take years before you get the money back but may be quite profitable with GAAP accounting.
GAAP accounting (Generally Accepted Accounting Principles) is what's used for official reporting and tax returns but excludes any increases in IP value or goodwill unless there's a buyout. If you included those the likes of OpenAI or Anthropic would have done pretty well. I'm not sure there's a word for that but basically value of the business less the money that's gone in. It doesn't get reported because 'value of the business' is guesswork and can be prone to BS but is pretty important to real world outcomes. AI is probably doing well on that one. Maybe why
>Is AI Profitable Yet? NO. Everyone's Broke.
doesn't fit with the top companies on the list having many billions in the bank.
Maybe most of them or all of them lose on their bets, but there's potential for a future where revenue grows beyond the immense capex and research investments.
Oracle though... Immensely risky capex to service a startup industry with what will soon be a commodity...
[1] https://www.wired.com/story/spacex-ipo-anthropic-compute-fin...
