AI models cost massive sums of money to be trained, and this means that leading AI companies like OpenAI and Anthropic have yet to turn a profit, but things could be different on a per-model basis.
Anthropic CEO Dario Amodei has indicated that the individual models that the company creates are profitable, even if one includes the cost of training them. But given how Anthropic keeps training new models, it reports overall losses at a company level.
“There’s two different ways you could describe what’s happening in the model business right now,” Amodei said on the Cheeky Pint podcast. “So, let’s say in 2023, you train a model that costs $100 million, and then you deploy it in 2024, and it makes $200 million of revenue. Meanwhile, because of the scaling laws, in 2024, you also train a model that costs $1 billion. And then in 2025, you get $2 billion of revenue from that $1 billion, and you’ve spent $10 billion to train the model,” he added.
“So, if you look in a conventional way at the profit and loss of the company, you’ve lost $100 million the first year, you’ve lost $800 million the second year, and you’ve lost $8 billion in the third year — it looks like it’s getting worse and worse. (But) if you consider each model to be a company, the model that was trained in 2023 was profitable. You paid $100 million, and then it made $200 million of revenue. There’s some cost to inference with the model, but let’s just assume, in this cartoonish cartoon example, that even if you add those two up, you’re kind of in a good state. So, if every model was a company, the model, in this example, profitable,” he added.
“What’s going on is that while you’re reaping the benefits from one company, you’re founding another company that’s much more expensive and requires much more upfront R&D investment. The way this is going to shake out is that it’s going to keep going up until the numbers get very large, and the models can’t get larger, and then there will be a large, very profitable business. Or at some point, the models will stop getting better, and there will perhaps be some overhang — we spent some money, and we didn’t get anything for it — and then the business returns to whatever scale it’s at,” he said.
Amodei makes a compelling point, but this analysis might not necessarily work out for all model companies. For starters, there’s plenty of competition in the models space, with Google, OpenAI, Anthropic and xAI closely tied at the frontier, with open-source companies right behind them. This competition could result in price wars, which would mean that companies might not necessarily be able to reap profits from the models they’re building. Also, an unusual aspect of the AI model space seems to be that companies cannibalize their own products — even before they might’ve extracted all the value from a model, they release a new one, and the older model can’t make much money after that. These are interesting dynamics, but Amodei isn’t the only AI CEO that’s hinted at profitability — OpenAI CEO Sam Altman has said that OpenAI is profitable if one excludes training costs, which indicates that the company is making money on inference of its models. It remains to be seen if these early indications of profitability end up coming true at the company level, but given the investments in these companies, there’s lots of money riding on the fact that they do.