In routine investor briefings in late July, executives at Anthropic shared a number that would have seemed implausible two years ago: the company’s annualized revenue had crossed $65 billion, according to people familiar with the matter. The figure marks a gain of roughly 38% from the $47 billion pace recorded in May and a more than sixfold increase from the approximately $9 billion annualized rate at the end of 2025.
Anthropic disclosed the figure during its regular updates to investors, one of the people said. The surge in sales strengthens the case for the public listing the company has been preparing. Anthropic and OpenAI have both filed confidentially for initial public offerings, and Anthropic is expected to reach Wall Street as soon as this fall, ahead of OpenAI, its larger and better-known rival.
The revenue picture explains the rush. Anthropic’s growth over the past year has come from two engines: large enterprise contracts that put its Claude models into corporate software budgets, and a fast-growing developer business in which companies pay for API access to build their own AI products. Its coding tools, in particular, have become fixtures in software teams that once wrote everything by hand, and usage there has compounded quarter after quarter, according to analysts who track the company.
The climb from about $9 billion to $65 billion in annualized revenue in roughly seven months is the steepest ramp of any private software company in recent memory. Part of the increase reflects price and usage growth on existing customers rather than new names alone, people familiar with the matter said, which is a double-edged signal: it proves the product’s value, but it also concentrates revenue in a small number of very large buyers who could in principle negotiate harder or consolidate vendors.
The public markets will test the story. At $65 billion in annualized revenue, Anthropic would price on a multiple that makes even today’s stretched AI valuations look tame, and investors will want to see the gross margins, compute costs, and customer concentration behind the headline number. AI companies carry unusual cost structures: every dollar of revenue leans on expensive chips, data centers, and electricity, and the accounting for those costs will be scrutinized line by line in the prospectus.
The company’s dependence on its cloud partners adds another layer of disclosure. Anthropic trains and runs its models on infrastructure rented from hyperscale cloud providers, which means its largest suppliers are also, in some cases, investors and potential rivals. Investors will look for how much of the company’s compute is contracted for years ahead and how exposed the model is to price increases in power and hardware.
The timing also matters. Anthropic is preparing to sell shares into a market that has embraced AI companies but grown more discriminating about valuations after a year of volatility in the sector. A successful listing this fall would give the market its first pure-play look at the economics of a frontier AI lab, with all the numbers out in the open for the first time. OpenAI, which has filed confidentially as well, would follow, and the two prospectuses together would define how Wall Street prices the AI industry for years.
Regulators are watching the same documents. The European Union’s AI Act imposes obligations on the most powerful models, and safety questions have dogged the industry as deployment outruns oversight. People familiar with the company said Anthropic has been expanding its safety and trust teams as it prepares for the scrutiny that comes with a public listing, and that the prospectus will describe its risk framework in detail.
For Anthropic, the $65 billion number is a fact, not a forecast. The question for the fall is whether the markets see the same growth ahead, or whether the multiple gets set by the risks in the footnotes.
The growth has a distinctive shape. Anthropic’s customers are increasingly large enterprises that have standardized on Claude for everything from customer service to internal knowledge systems, and the company has built its sales force around landing a small number of very large contracts rather than chasing volume, according to people familiar with its operations. The approach produces lumpy quarters but high retention, and it has made the company’s revenue more predictable than the consumer-driven numbers at some of its rivals.
The competitive context sharpens the story. OpenAI has its own growth narrative and its own confidential filing, and Google’s Gemini business operates at a scale that neither lab can match, which means Anthropic’s $65 billion run-rate will be measured against two very different baselines when it goes public. The IPO will also test how much of the company’s value rests on models that could be commoditized by cheaper competitors, a question every frontier lab faces and none has fully answered.
The roadshow will be watched as closely as the filing. Anthropic’s founders have been public about their view that the industry’s safety questions are as important as its commercial ones, and investors will want to see how that conviction survives the pressure to hit quarterly numbers. The fall listing, if it happens on schedule, would make Anthropic the first pure-play frontier AI lab on public markets, and the price the market puts on its $65 billion run-rate will be the industry’s first real valuation benchmark.
