Dario Amodei explained the paradox of AI company losses

Dario Amodei from Anthropic revealed a brilliant business model that explains the AI industry paradox. Companies show losses but actually earn billions. Here’s how it works in practice. Imagine each AI model as a separate mini-business. You invest $100 million in training. In a year, this model earns $200 million. 100% profit! But you don’t see this money in reports. Why? Because all $200 million immediately goes into creating the next model, which already costs $1 billion. In accounting, this looks like a loss, though each model pays for itself with interest.

The secret is in exponential value growth. When AI transforms from a smart student to a PhD-level expert, companies are ready to pay tens of times more. Problems get solved that were previously impossible. Plus this isn’t a mass product. Each model is unique, customized for business processes, creates sticky products. Customers don’t want to change providers. You get an endless conveyor belt of growth up to the technological ceiling.

As long as there are new investors ready to invest in the “next model,” the scheme works. But what happens when the money runs out? Or when progress slows down? The main risk is that if growth slows, the entire reinvestment pyramid could collapse.