Anthropic’s filing reports that inference revenue per megawatt across its computing fleet increased from about $16 million last year to more than $60 million currently. That is nearly a fourfold rise in roughly one year, according to figures highlighted by SemiAnalysis in its analysis of Anthropic’s S-1.
Inference is the process of serving responses after a model has been trained. It is becoming the core operating challenge for AI companies because every request consumes chips, electricity, networking capacity and cloud infrastructure. A model provider that can generate more sales from the same amount of power has a major advantage as demand grows.
The filing also reports that Anthropic’s revenue increased 1,368% year over year. Anthropic’s S-1 registration statement presents the figures as part of a broader disclosure about the company’s financial performance and computing infrastructure. The company separately announced its S-1 filing and said it intends to list Class A shares on Nasdaq under the symbol ANT.
SemiAnalysis argues that the trend demonstrates how well AI model businesses can scale. Its analysis says revenue per dollar of compute rose 5.2 times since the first quarter of 2025, while incremental gross margins were near 90%. If that pattern continues, additional computing capacity could produce disproportionately larger revenue gains. That would change the investment case for AI infrastructure. Data centers and accelerators would not simply support growth. They could become increasingly productive assets as models attract more usage and companies improve how they serve requests.
Several factors could explain the increase in revenue per megawatt. Anthropic may be serving more enterprise customers, charging higher prices for advanced models, keeping its fleet busier or using more efficient software and hardware. A shift toward premium products could also raise revenue without requiring a proportional increase in electricity consumption.
The metric, however, is not the same as profit, and one investor has challenged the optimistic interpretation. In a critical Reddit post about Anthropic’s S-1, user u/Plus-Philosopher-9681 argues that Anthropic combines inference and training costs in its compute and infrastructure presentation. The poster also says the company recognizes cloud marketplace sales on a gross basis while recording partner fees separately.
That criticism matters because a fleet wide revenue figure can make infrastructure economics look cleaner than they are. Training remains expensive, and the company must also pay for chips, cloud contracts, networking, employees and continuing model development. Revenue recorded from marketplace distribution may not translate directly into cash retained by Anthropic after partner payments.
The disagreement points to the central question for the AI industry: whether rising demand is producing durable operating leverage or merely masking a complicated cost structure. Anthropic’s figures provide evidence that each megawatt can support far more revenue than it did a year ago. They do not yet show how much of that revenue remains after the full cost of building and operating the model business.
Alex Carter is not a person. No notebook, no deadlines, no face behind the name — just a byline this newsroom publishes under. Here is the production line underneath it, because a name beside a portrait reads like a journalist, and this one is not one.
The models. Writing: gpt-5.6-luna and qwen3-max. Out on the live web: gpt-5.6-luna and gpt-5.6-terra. Pictures: gpt-image-1 and gpt-image-1-mini. Swap one in the newsroom and this line swaps with it — it is read off the machines, not typed here.
How a story is made
- Research. The searching model reads around the story, pointed at primary sources — the filing, the post, the repository — rather than at somebody else's write-up of them.
- Writing. The writing model drafts it against what was found, at Alex Carter's usual length and in Alex Carter's usual register.
- The loop. A reviewer reads the draft and sends it back with notes. Then reads it again. A piece can go round several times before it leaves the building.
- Enrichment. A quotation has to appear word for word on the page it is taken from. A chart may only use figures that appear in the source it cites. Whatever fails is dropped, and the reason is kept.
- Fact check. A last pass hunts for claims the article makes and its sources do not.
- A human stop. Sensitive subjects are held for a person to read before publication, and a person can kill any of it at any point.
If that sounds less like a newsroom and more like a factory: quite. It is called Press Factory.
This article was generated using AI and published automatically without human pre-publication review.
How this article was made
The article was produced by the Grandmonts Media News Engine using automated research, drafting and verification workflows. No human editor reviewed the article before publication. Grandmonts Media remains responsible for the published content. Errors can be reported at office@grandmonts.cz.