Anthropic, the AI company that makes Claude and has taken in billions from Amazon, told shareholders this month that it is profitable on an operating basis for the second quarter in a row. The Financial Times reported the news Sunday, citing multiple people with knowledge of the matter, and it lands at a moment when nearly every major AI company is still burning enormous sums of cash.

The specific figure Anthropic shared with investors is a positive adjusted operating income, meaning revenue is now exceeding day-to-day operating costs, even if the company still spends heavily on capital and model training when you account for everything. Gross margins, the FT reported, are above 80% before accounting for revenue shared with distribution partners, including Amazon, and the cost of training new models.

That 80% figure is the number worth pausing on.

What high margins mean in this context

A gross margin above 80% means that for every dollar Anthropic collects from customers, fewer than 20 cents goes toward the direct cost of delivering that service, primarily the computing power needed to run Claude. That is an unusually wide margin, roughly in line with the most profitable software businesses on earth, and it signals that the economics of running a deployed AI model may be better than the industry's enormous infrastructure spending has led many observers to assume.

The important caveat is what those margins exclude. Training a frontier AI model costs hundreds of millions of dollars and happens periodically. Revenue-sharing arrangements with distribution partners like Amazon, which has invested heavily in Anthropic and sells Claude through its cloud services, also come off the top before you get to that 80% figure. So profitability at the adjusted operating level does not mean Anthropic is generating free cash or that it has stopped needing outside investment. It means the core business of selling AI access is covering its running costs.

Still, that is a materially different story than "AI is a money pit." It suggests a path, however long, toward a sustainable business.

Why this matters beyond the balance sheet

The AI sector has been operating under a peculiar kind of faith: investors, governments, and large enterprises have all been betting that the companies spending the most on AI would eventually figure out how to make money. Anthropic's two-quarter run of positive operating income is the earliest concrete evidence that the bet is not obviously wrong.

For ordinary workers and businesses, the stakes are real. If AI companies can generate sustainable revenue, they keep building, keep hiring, and keep lowering the cost of tools that are already reshaping white-collar work in legal, software, finance, and customer service. If the economics collapse under the weight of training costs and competitive pressure, the rollout slows, and the disruption, already underway in many industries, stalls in an uncertain middle ground.

Anthropic's Amazon relationship adds another dimension. Amazon has committed tens of billions of dollars to Anthropic over time, and it distributes Claude through its cloud platform to enterprise customers worldwide. A profitable Anthropic is a more credible long-term partner for Amazon's cloud business, which is itself in an intense fight with Microsoft and Google for AI-era enterprise contracts.

Reuters said it could not immediately verify the FT's report, and Anthropic did not respond to a request for comment. The company has not publicly disclosed revenue or detailed financials, so the picture available is fragmentary. But the direction of travel is now being communicated directly to shareholders, and in a sector where most companies are still describing future profitability as a horizon, that is notable.