Which is where your analogy breaks down and why you think you’re taking crazy pills. Inference is growing and selling the oranges in your analogy. Model building is growing the farm to sell larger, juicier more addicting oranges.
The same mistake was made with Amazon, and a million other tech companies in the early 2010s.
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.
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.
> Inference is growing and selling the oranges in your analogy. Model building is growing the farm to sell larger, juicier more addicting oranges.
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.)
In this particular case, inference and training are intertwined. It might be one thing if Anthropic could get away with training a new model every five years and control costs that way. But they can't. Put another way, their inference has no value without continuous, very expensive training. Because consumers aren't purchasing based on price but capability, otherwise the Chinese models on OpenRouter would have buried OpenAI and Anthropic already.
Last month Anthropic tried to control the narrative by drumming up the “super scary AI” trope.
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.
My reading on anthropic is that he strongly infers that they're profitable, realizes what he has said and immediately walks it back as explicitly not the case today and reframes it as a guess about some indeterminate point in the future.
> Those are the economics of the industry today, or not today but where we're projecting forward in a year or two.
Profitable for inference if you completely ignore training costs and that you absolutely must continuously train new models.
Which is where your analogy breaks down and why you think you’re taking crazy pills. Inference is growing and selling the oranges in your analogy. Model building is growing the farm to sell larger, juicier more addicting oranges.
The same mistake was made with Amazon, and a million other tech companies in the early 2010s.
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.
Are ya fuckin' serious mate?
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.
13 replies →
> Inference is growing and selling the oranges in your analogy. Model building is growing the farm to sell larger, juicier more addicting oranges.
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.)
In this particular case, inference and training are intertwined. It might be one thing if Anthropic could get away with training a new model every five years and control costs that way. But they can't. Put another way, their inference has no value without continuous, very expensive training. Because consumers aren't purchasing based on price but capability, otherwise the Chinese models on OpenRouter would have buried OpenAI and Anthropic already.
Last month Anthropic tried to control the narrative by drumming up the “super scary AI” trope.
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.
And ignore capital costs, depreciation, user churn etc
Anthropic says they are losing money but hope to be profitable soon, if they can double their revenue. https://techcrunch.com/2026/05/20/anthropic-says-its-about-t...
OpenAI published their operating margins today. I can't find a non-paywalled source but Judd Legum reported it to be -122%.
AI CEOs are known to say many things telling the truth, probably isn’t one of them.
Do you mind sharing source links to that profitability claim?
I’m struggling to find the quotes.
Open AI: https://simonwillison.net/2025/Aug/17/sam-altman/
Anthropic: https://x.com/jaminball/status/2052112309364162874
My reading on anthropic is that he strongly infers that they're profitable, realizes what he has said and immediately walks it back as explicitly not the case today and reframes it as a guess about some indeterminate point in the future.
> Those are the economics of the industry today, or not today but where we're projecting forward in a year or two.
If only they had their books open to do more than just "say"
do you have proof? Taking these guys at face value is not wise