Anthropic's Revenue Jumps, but Its Computing Bills Take a Bigger Bite

Anthropic reported $17 billion in revenue for its latest quarter, up from $11.5 billion in the quarter before. The figures come from an updated S-1 filing, the document a company submits when it prepares to sell shares to the public. This version covers results through September 30. The earlier filing stopped at June 30. TechCrunch reported the figures on November 3, 2026.
The costs in the filing have drawn the most attention.
"Compute" means the computer chips and data centers that run Anthropic's AI models. It made up 58% of revenue in the latest quarter, up from 48% in the quarter before. The quarterly compute bill was $10 billion. Of that, $5.0 billion was for compute bought on short-term contracts.
Short-term contracts cost much more than multi-year ones. Buying them is a bit like booking a hotel room at the front desk after you arrive. You pay the highest rate because you could not plan ahead.
Revenue per dollar of compute spend fell to 1.72x. That means each dollar spent on compute brought in $1.72 in revenue. In the quarter before, the figure peaked at just over $2.
Anthropic said it expects compute and infrastructure expenses to rise in dollar terms. It also expects them to remain its largest operating expense for the foreseeable future. The company is also building its own computing base under a project it calls "Anthropic Direct." That effort is at an early stage, and the filing gives no timeline.
Bloomberg reported on November 2 that Anthropic's revenue reached nearly $17 billion ahead of an IPO. It also reported that the company disclosed a twelve-fold increase in quarterly revenue. Bloomberg further reported that Anthropic intends to spend more than $895 billion in the years ahead. That is a stated plan, not a set of payments already committed.
On profit, Yahoo Finance reported an operating loss of $567 million for the third quarter of 2026. An operating loss measures the business's day-to-day results. A separate Reuters report, citing the Wall Street Journal, put Anthropic's net loss for the first half of 2026 at $80.7 billion. That larger number grew mainly because the fair value of bonds rose, which changed how the bonds are listed on the company's books. A figure shaped by that kind of accounting says little about whether each sale makes money. That is why the operating line matters more to anyone studying the business.
In my view, the central question in these numbers is not demand. Revenue has grown faster than costs. Yet computing costs are taking a larger share of each revenue dollar, and the drop in revenue per compute dollar points the same way. Short-term capacity is the most expensive way to buy the power to run AI at scale. Longer contracts and owned equipment could lower that cost, but both carry risk. They commit money and capacity before anyone knows how fully they will be used.
The broader context here is that the market is paying a premium for Anthropic's growth, and the filing shows that premium being paid in real time. If the company moves more of its work onto longer contracts or its own equipment, the same revenue could bring better margins. If it cannot, a bigger slice of every revenue dollar will go to the data center companies and chip makers that supply it. The next set of disclosures should show whether the 58% figure is a peak or a trend.
For the wider industry, the filing offers something rare. It shows what a leading AI company pays for the capacity to run its models, and what that capacity costs when it is bought on short notice. Whatever the IPO outcome, the cost structure in this filing is likely to shape how investors judge every rival's computing strategy.


