Anthropic filed its S-1 to go public on the Nasdaq, and the AI company could begin trading as soon as next month at a valuation between $1.5 trillion and $2 trillion, which would place it among the most valuable companies in the world. The filing, published earlier this week, runs to hundreds of pages and gives the first detailed look at the finances of a leading frontier AI lab.
Joey Brookhart, an analyst at the chip and AI research firm SemiAnalysis and co-lead of its Tokenomics model, said the filing showed a company that had scaled efficiently to more than $70 billion in annualized recurring revenue in a short period. He said gross margins, compute margins and free cash flow all looked the way an observer would want, and that he found little to criticize in the document. He attributed a rally in semiconductor, neocloud and compute stocks over the two days since the filing in part to its contents.
Brookhart said the market today remains extremely narrow even at roughly $120 billion of ARR across the two leading labs. The top 1 percent of customers spend $50,000 to $100,000 per employee, while the broader Fortune 500 and Global 2000 do comparatively little spending. Much of Anthropic's revenue is coding-specific, making it almost a single-vertical company, he said.
Reaching the $1 trillion revenue figure Dario Amodei has said he can see requires moving into white-collar work, where AI shifts from augmentation toward labor saving or replacement in some areas, Brookhart said. He pointed to cyber, which can resemble coding through eight-figure-plus contracts with leading enterprises, and to bio as other verticals. He also said enterprises have yet to go through a meaningful Q4 budgeting season with one of these labs because their ARR did not really begin accelerating until the beginning of the year.
Competition has intensified, according to Brookhart. Anthropic accounted for 90 percent of the industry's net new ARR in the first half of the year, while OpenAI was probably the majority of net new ARR in Q3. Open-source token inference costs and ARR grew from about $1 billion in Q1 to more than $10 billion in Q3, he said.
Brookhart also described enterprise interest in running AI on-premises and customizing open weight models, citing law firms running on-prem AI. He noted that DeepMind at Google and Meta Superintelligence Labs have well-funded operations with similar training compute capacity, and he said coding alone will not be enough to triple the business from here.
More AI news from TechManNews.
