Anthropic plans to open its books to a group of institutional investors on Wednesday, disclosing third-quarter financials for the first time as it prepares for what could be the largest initial public offering in history. CNBC, citing two people familiar with the matter, reported that the presentation will take place at the company's San Francisco headquarters; Bloomberg first reported the investor gathering. Anthropic declined to comment.
The timing is not incidental. Private AI labs now move public markets: last week, confusion around OpenAI's updated revenue disclosure — a $50 billion annualized run-rate, roughly $20 billion below the $70 billion figure that had been widely circulated — dragged down major semiconductor names and the broader Nasdaq. According to a person familiar with the matter, OpenAI's updated number strips out revenue from cloud partners and does not reflect changing demand. It is not yet clear whether Anthropic will account for sales the same way, and that accounting choice alone could swing its headline figures by billions.
Investors will be working from thin public data. Figures shared with shareholders earlier this year and reported by the Financial Times put Anthropic's second-quarter revenue above $11.5 billion, with an annualized run rate near $65 billion by the end of July, gross margins above 80 percent before partner revenue shares and training costs, and an adjusted operating profit — a measure that excludes stock-based compensation — that had turned positive. The same reporting put committed spending on computing capacity above $500 billion, most of it with Google, Amazon and Microsoft. All of these are preliminary, unaudited and company-reported numbers, and this week's disclosure will be the first real test of whether the trajectory held in the third quarter.
People familiar with the plans tell CNBC that Anthropic is plowing ahead with a listing that could value it at $2 trillion, based on latest reported revenue that could top $100 billion by the end of this year. That would exceed the valuation of SpaceX, which went public in June and became the reference point for outsized tech debuts. Anthropic filed confidentially with the SEC earlier this year, and Reuters has reported a November listing, though people familiar could not confirm a date.
Not everyone is buying it. Independent research firm New Constructs values Anthropic at roughly $150 billion — about 93 percent below the rumored target — and has called the prospective offering "2026's most ridiculous IPO," arguing the company lacks a viable business model at that price. The circular flow of money in the sector amplifies the concern: Amazon and Alphabet are simultaneously Anthropic's largest backers and its compute landlords, a structure regulators and short-sellers have increasingly put under a microscope.
Macro conditions are hardly cooperating. Climbing interest rates have shaken parts of the AI trade, chip stocks extended an overnight sell-off on Tuesday, and Broadcom CEO Hock Tan told a conference this week that with rates rising, access to capital could become the biggest constraint on AI companies. Market appetite, not technology, may end up setting the timing of the deal.
Either way, the Wednesday session marks a threshold: the moment when the financial mechanics of frontier AI labs — revenue quality, margin structure, partner accounting, customer concentration — stop being private trivia and become the numbers the entire tech trade trades on.
Comments (0)
Log in to join the discussion
Log InNo comments yet