Just now, Anthropic's annualized revenue has exceeded 65 billion US dollars, and it is sprinting for the largest IPO in history.

If you look back just a year ago, it would be hard to imagine how steep Anthropic's revenue curve would become.

Earlier this morning, Bloomberg reported that Anthropic disclosed to investors that by the end of July this year, the company's annualized revenue run rate (ARR) had reached 650 billion U.S. dollars. Calculated based on the annualized level of over 9 billion U.S. dollars at the end of 2025, this figure has grown to more than 7 times the original in just a few months.

Meanwhile, Anthropic's preliminary revenue for its most recently completed quarter (Q2) also exceeded 11.5 billion U.S. dollars. In the same period last year, this figure was only 787 million U.S. dollars, meaning the latest quarterly revenue is about 14.6 times that of the same period last year.

For an AI company that was widely regarded as a challenger to OpenAI just a few years ago, this is probably the most dramatic acceleration in its commercialization process to date.

From 470 billion USD to 650 billion USD in two months

First of all, it should be clarified that the 650 billion U.S. dollars does not mean that Anthropic has actually earned this much revenue within a year.

The metric used here is "annualized revenue run rate". Simply put, it estimates how much revenue the company can generate for the whole year if its performance at the current revenue level for a period of time continues for a full year. This metric is especially suitable for observing companies with rapid growth, as it can reflect current business changes faster than the revenue of a completed fiscal year.

And Anthropic's current growth rate is very prominent.

At the end of 2025, its annualized revenue run rate just exceeded 90 billion U.S. dollars; by May this year, the figure had crossed 470 billion U.S. dollars; by the end of July, it further reached 650 billion U.S. dollars.

In other words, in just two months, Anthropic's annualized revenue run rate has increased by about 38% again.

From a longer time scale, the change is even more obvious. According to the data disclosed by Bloomberg, Anthropic's total revenue in 2025 is about 100 billion U.S. dollars. Now, its annualized scale calculated at the latest revenue speed has reached more than 6 times that figure.

Revenue growth has also begun to be reflected at the profit level.

Anthropic's preliminary revenue for its most recently completed quarter exceeded 11.5 billion U.S. dollars, a substantial increase compared with 787 million U.S. dollars in the same period of 2025. At the same time, the company recorded positive adjusted operating profit for the quarter. For frontier model companies that are still in the stage of large-scale computing power investment, this point is particularly noteworthy.

Over the past few years, the growth story of large model companies has usually been accompanied by the other side: training costs, inference costs, data center investments and talent costs have expanded simultaneously. Rapid revenue growth does not mean that losses will disappear quickly.

Anthropic's emergence of positive adjusted operating profit at least indicates that its scale expansion is gradually demonstrating operating leverage.

Claude Code transformed Anthropic's commercialization speed

Behind Anthropic's skyrocketing revenue, an unavoidable variable is the enterprise market, especially programming and complex knowledge work scenarios.

Claude was initially seen more as a direct competitor to ChatGPT. But over the past year, an obvious change in Anthropic's commercialization is that it has begun to go deeper and deeper into the real workflows inside enterprises. Claude Code is the most typical product among them.

Developers can directly let Claude read code repositories, modify files, execute commands, troubleshoot problems, and continuously complete long-chain software engineering tasks. As the usage scale of Coding Agents in enterprises expands, model invocations have changed from occasional one-off conversations in the past to continuously running workflows that consume a large amount of Tokens.

For model companies, the revenue brought by these two usage modes is completely different.

A single question and answer may only invoke the model for a few seconds; a software engineering Agent that lasts for tens of minutes or even hours may repeatedly read context, invoke tools, modify code and re-verify results. This also explains why enterprise-level Agents are increasingly rapidly becoming the revenue engine for frontier model companies.

Judging from the data disclosed so far, Anthropic is already one of the most obvious beneficiaries of this trend.

Bloomberg previously reported that OpenAI's recent annualized revenue run rate has exceeded 400 billion U.S. dollars. If we simply compare this metric, the newly disclosed 650 billion U.S. dollars of Anthropic is significantly higher.

However, these two figures still need to be compared with caution. The statistical calibers of annualized revenue run rate of different companies may not be completely consistent, and differences may arise from factors such as revenue recognition methods and the calculation methods of enterprise contracts.

The IPO is getting closer, and 650 billion USD is the most direct bargaining chip

The timing of this figure is also very delicate. Anthropic is on the eve of going public.

According to previous reports from the Financial Times, Anthropic's investors expect that the company may land on the public market in October this year, with an IPO valuation of 2 trillion U.S. dollars or even higher.

If it goes public at this valuation in the end, Anthropic will surpass SpaceX, which went public earlier this year, to become the highest-valued IPO in history.

For Anthropic, which is about to enter the capital market, the annualized revenue run rate of 650 billion U.S. dollars is obviously an extremely important asset.

AI companies have had extremely high valuations in recent years, but the capital market will eventually return to several very traditional questions: how much revenue there is, how long the growth can be sustained, when profits will appear, and whether huge computing power investment can be converted into sustainable cash flow.

Anthropic was long seen as a follower behind OpenAI. Now, it has proven to potential investors that changes have taken place.

Model capabilities are still important, and enterprise market penetration, revenue growth rate and profitability are gaining more and more weight.

The current environment Anthropic is facing is not all smooth sailing. In June this year, the company temporarily shut down access to two advanced models, Claude Fable 5 and Mythos 5, due to government regulatory requirements, and restored access after about two weeks of negotiation. Even so, the company's latest revenue figures are still growing rapidly, which makes this revenue curve even more noteworthy.

Reference Links:

https://www.bloomberg.com/news/articles/2026-08-17/anthropic-revenue-run-rate-surpasses-65-billion-ahead-of-ipo?srnd=phx-technology

https://www.ft.com/content/840ac156-af1c-4a82-b260-ae791072fcfa?syn-25a6b1a6=1

This article is from the WeChat official account "Machine Heart" (ID: almosthuman2014), written by a contributor focused on large AI models, and published with authorization from 36Kr.