In the past week, Anthropic’s investors have been making their case to anyone who will listen: the maker of the Claude AI models is worth $2 trillion. The Financial Times reported on Aug. 13 that the company plans to begin its initial public offering in October, a debut that would eclipse SpaceX’s June listing and stand as the largest in history. The valuation, investors told the newspaper, is still under discussion inside the company, and some have floated a figure as high as $3 trillion.
The revenue numbers behind the pitch are extraordinary by any standard. The AI lab’s annualized revenue run rate, a measure that infers full-year sales from recent performance, is expected to be between $100 billion and $120 billion by the end of 2026, according to investors. That would put Anthropic’s revenue on track to grow more than tenfold over the course of the year, a pace that has no close precedent in the history of technology companies going public.
On the same day, Reuters reported that Anthropic is in talks to acquire Decart AI, an Israeli startup that makes software to squeeze more performance out of AI chips, in a deal worth about $6 billion. Bloomberg had earlier reported the talks. Decart also builds world models, simulations that help AI systems understand and navigate real-world environments. If completed, the acquisition would be Anthropic’s largest ever.
The two moves belong to the same strategy. An IPO puts Anthropic’s financials under public scrutiny for the first time, and the biggest line item on any AI lab’s income statement is the cost of computing. Decart’s optimization software promises to lower the cost of running and training models, exactly the kind of efficiency gain that would improve the margins investors will examine when the prospectus arrives. Buying the technology ahead of the offering lets Anthropic tell a story of a path to profitability rather than a company spending faster than it earns.
The company’s own positioning may soften the scrutiny. Anthropic has concentrated on enterprise customers, selling Claude to banks, insurers, and software companies that want AI without entangling themselves with its largest rival, and it has signed a string of large corporate deals. Those contracts come with usage commitments and renewals, giving analysts a base from which to project revenue growth, though they also carry lower margins than the consumer subscriptions that drive other labs’ top lines.
Skeptics note that the story rests on numbers the public has not yet seen. Anthropic’s financials remain largely private, and analysts who follow the company say its losses are substantial despite the revenue surge. Some market participants read the investor campaign as an exercise in anchoring: float $3 trillion, and $2 trillion starts to look reasonable. The Financial Times itself quoted one investor floating the higher figure, and its sources were unanimous only on the direction of travel.
Comparisons with earlier records underline how unusual the offering would be. Saudi Aramco raised $29.4 billion in 2019, the largest IPO by proceeds, but that listing was largely a sale of existing shares by the Saudi government. Anthropic’s offering would be a primary sale of new shares, likely smaller in proceeds, but the valuation attached to it would dwarf every technology debut that came before. SpaceX, whose June IPO valued the company at $1.77 trillion, would be surpassed before Anthropic’s stock ever trades.
The timing matters. October gives Anthropic a window before the year’s end, when the run-rate figures its investors are citing become audited history rather than projection.
The offering would also test a market already crowded with AI listings. Chip companies, cloud providers, and model developers have all come to market in the past year, and investors have grown more discriminating, punishing companies that miss expectations and rewarding those that show a clear path to profit. Anthropic’s size would make it the sector’s anchor listing, setting the tone for the AI names that follow it through the pipeline. It also comes as the AI market is absorbing a wave of listings and secondary transactions, from chip makers to model developers, and investors have shown they will pay for AI growth but punish companies that miss expectations.
The Decart talks show how the competitive pressure is reshaping the business. OpenAI and Google are pouring money into models and infrastructure, and Anthropic’s enterprise focus has made it a favorite of corporate customers who want to deploy AI without depending on its largest rival. But enterprise contracts carry lower margins than consumer subscriptions, which makes cost efficiency central to the valuation story.
For the public markets, the offering is a test of whether the private AI boom can survive contact with audited accounts. The company’s revenue growth is real and visible in its customer wins; its spending is a question the prospectus will answer. Investors who have watched AI valuations climb for two years will finally get a balance sheet to argue about.
People familiar with the matter caution that the plan could still change. The valuation is not formally set, the timing depends on market conditions, and the Decart talks could fall through. But the direction is clear: Anthropic intends to reach the public markets at a valuation that would make it the most valuable company ever to list, and it is buying the efficiency tools to defend that number once the books are open.
