Do you have a proof? Anthropic’s CEO said they Are profitable. Same with OpenAI.
The restaurant next to the mines were profitable up until the moment the mines themselves shut down: one doesn't exist without the other.
You can't ringfence inference as "the profitable bit" and then hand-wave away the training. Without continuous training there is no inference product.
Claude 3 Opus isn't sitting there making revenue in 2026 - the thing is just deprecated. The moment you stop spending billions on the next model, your "profitable" inference business is on borrowed time until someone else makes it obsolete.
Maybe I made a mistake in my analogy... They're not growing a farm and then selling oranges. They're on a treadmill where stopping is death, and the treadmill costs $10bn a year to keep running.
You’re literally describing all companies. Google takes about $270bn/year to run. If they stopped spending that they’d die pretty darn quick. It’s also a description of working - unless you’d built up significant savings, if you stopped working you’re also going to die.
There is no moat. In the end, what we are calling AI today will just be something that is incorporated into an existing programs that people will use to help them accomplish a task. The public will not be paying more for it. It will just be a commodity added to the existing ecosystems we have today. They
No, not quite. It really comes down to opex vs capex and the depreciation schedule for your investment.
Software development is typically categorized as capex, on a 3-5 year depreciation schedule. You assume the software you write today will be generating value for you that long.
If a big, expensive model training project only gives you value for a year or less, that is not like most companies.
But believing that the financials of a project are governed solely by how IRS rules force you to account for headcount is kind of silly.
> If a big, expensive model training project only gives you value for a year or less, that is not like most companies.
The model itself that gets built? Sure (although clearly the timelines are getting longer). However the important bit here is the research that got done along the way and the infrastructure built to make that model building process cheaper, better etc. all of that stuff sticks around but because it’s hard to appreciate externally you discount it to 0 when it’s literally what they actually spent the money on.
But none of that even matters. Google had 270B in opex and their capex has grown from 50B in 2024 to 90B in 2025 and is projected to grow to ~175B for 2026. But even if you discount the “AI” treadmill, you’re still looking at many tens of billions in capex that if they stopped they’d die.
It's the same in other industries. In your example, if labor went into building a whole new assembly line that pays itself off over several years, capex. If it just goes into building products to sell, opex.
https://kpmg.com/kpmg-us/content/dam/kpmg/pdf/2023/tcja-chan...
Unless they are changing the architecture in huge ways. The pre-training done for 3 goes into later models. I am sure the frontier labs are figuring out how to pretrain generic feedstocks that can be fed into downstream training pipelines. DeepSeeks incremental training run cost was what, 5M? Alibaba and DeepSeek have the best most efficient training pipelines, look at the rate at which custom Qwen models are being pumped out.
This claim deserves teasing apart.
Clearly, training is a Red Queen's race today. If a model provider were to unilaterally decide to stop training, they would very quickly lose market share to competitors with better models.
On the other hand, what if market and investment conditions change such that everybody has to stop training?
In that case, the models are still there and still as useful as they were the day before. So why wouldn't there still be an inference product?
The reason they're losing money on paper is because the models keep growing 10x in size every generation but they're not getting 10x returns on model inference (closer to 2x)
Dario's point was the opposite of yours; he used per-model accounting to explain why the company P&L gets worse every year, not better. His own numbers (10x training costs each generation and ~2x revenue return.)…
"It looks like it's getting worse and worse" are his words, not mine.
In 2025, Anthropic's inference costs came in 23% over their own projections. They cut their gross margin forecast from 50% to 40%.
Amazon were losing money, they were losing money because were growing and spent all of their cash flow on growth. It wasn't merely regarded as a hopelessly unprofitable business, if was regarded as potentially fraudulent. The share price collapsed in 2014 because, some thought, the profit would never come, investing in growth was pointless, etc.
Last year Amazon made nearly $100bn in profit. Stock is up 20x from then...this is after AWS was known (everyone also that was a massive fraud, could never be profitable...we know it was printing from day one), after it was the world's biggest retailer, etc.
It is difficult to understate how consistently people make this mistake, not just individually but in aggregate. You see the same thing with restaurants, consumer products, office leasing, so many businesses. This is not to say that the future will happen any particular way but that what Anthropic and co are doing is obviously rational and based upon very real cash flow. Anthropic's growth in revenue is, I believe, unparalleled in modern corporate history. A slight difference in this case is also that the economics of training these models is improving exponentially over time.
In this analogy, model training would be akin to developing better oranges, but your competitors are also developing better oranges so if you stop spending heavily to improve your oranges, consumers are going to buy ~zero oranges from you within a couple years. (Expanding the farm might be analogous to expanding data centers.)
The news they successfully buried was that companies like AirBnB are now running Qwen and open source models. The free oranges are now good enough. There is no future unless the goal is to get to super intelligence and utterly take over the world before anyone else gets one. Anything else and free models are six months behind. The money now is the opposite of what everyone thought a year ago: datacenters. Everyone thought AWS was fucked. Turns out AWS is really good at running Qwen.
I’m struggling to find the quotes.
> Those are the economics of the industry today, or not today but where we're projecting forward in a year or two.
OpenAI published their operating margins today. I can't find a non-paywalled source but Judd Legum reported it to be -122%.