Anthropic IPO Waiting Game Puts Wall Street on Edge

Anthropic finance chief Krishna Rao and other executives met with groups of prospective public investors last month with his company’s numbers looking spiffier than ever. The company had gone from spending about $2.30 on operations for every $1 of revenue in spring of 2025 to turning a slight profit on that basis in the June quarter—bucking the stereotype of money-losing AI companies that are addicted to spending.

But some investors walked away from meetings with the understanding from executives that the profits would likely be short-lived. Anthropic would take advantage of its sudden growth and deep cash reserves to aggressively sign data center deals, executives told them, which would increase how much they spend on computing infrastructure and could shrink operating profits.

The conversations were part of “testing the waters” meetings ahead of Anthropic’s initial public offering, which will have implications for hundreds of IPO candidates in waiting, not to mention the economy.

Anthropic still hasn’t fired the starting gun for that weekslong IPO process, which would set off a frenzy in what have been tightly guarded deal preparations so far. The IPO, which many investors were expecting to come as early as this month, is now looking like it might not happen until late October or November. The next steps would involve Anthropic hosting research analysts and investors for more in-depth presentations, important pre-IPO moves that usually come a couple of weeks before companies release their IPO prospectuses publicly. It can take another three to four weeks before their stocks actually start trading.

The company’s next tour across Wall Street is likely to come with tougher questions than it faced in August. Rival OpenAI has since launched an advanced model that shows capabilities similar to Anthropic’s, and is now considering a new round of private fundraising on the momentum.

Both OpenAI and Anthropic have been at the center of a debate across Silicon Valley and Washington about the safety of AI development, and face the prospect of their businesses being undercut by cheaper, open-weight models. Meanwhile, rising interest rates will make it more difficult for companies to finance the building and expansion of data centers.

The confluence of factors is likely to make investors look harder at the Anthropic IPO, said Karen Snow, a former Nasdaq executive who runs an investor relations consultancy, Rose & Co. The Anthropic IPO “went from being a no-brainer to not being a no-brainer,” she said. “The markets can be finicky and sentiment has shifted a bit on AI. The issue is there’s a bit of a gray cloud over AI right now.”

The typical dance between public investors and newly listed companies about revenue forecasts and spending plans is expected to play out on a more dramatic stage than usual. Anthropic has grown its business at a record pace, with its moves to spend on power, chips and data centers affecting scores of stocks.

Anthropic executives have tried to make clear they still have plenty of financial firepower ahead of an IPO. They told investors the company had between $120 billion and $130 billion of cash as of early August. But they were planning to start going through that as Anthropic invests more heavily in computing infrastructure, particularly to focus on training new AI models, they said.

AI Development Continuing Despite Safety Concerns

While Chief Executive Dario Amodei has called for more safety evaluation and regulation for advanced AI models—under an informal set of commitments called “pacing the frontier”—he has clarified that “pacing does not mean halting model training or technical progress.”

Going public will be particularly important for Anthropic’s compute plans, freeing up Anthropic to not only raise tens of billions of dollars of additional cash in the IPO itself but also to tap debt and equity markets for years to come. The story Anthropic tells to investors will have to balance Wall Street’s desire to see a path to profits.

“Anthropic going public means it is elevating itself to the entire financial market, making it more palatable for them to get different options of financing,” said Kevin Xu, founder of Interconnected Capital, an investment fund focused on AI infrastructure stocks, who hasn’t formally met with Anthropic. “But the public market community still wants to trust the fact that Anthropic can fund its own capital expenditure, and that debt is a bridge to that—not that the borrowing has no end in sight. That’s the problem the entire AI complex has right now.”

Anthropic’s pitch hinges on investors buying into what the company called in its presentation an “unprecedented financial model,” including revenue that grew by roughly 2.5 times between the first and second quarter this year, from about $4.7 billion to more than $11.5 billion. Its spending on computing infrastructure, meanwhile, only grew by about 65% during the same period, from $3.4 billion to $5.6 billion.

Anthropic has focused investors on an “adjusted operating margin” figure as a measure of profitability, a fairly common non-standard figure that excludes non-cash costs such as stock-based compensation for employees and an expansive share-based donation-matching program that Anthropic runs. Those excluded expenses totaled more than $1 billion in the first quarter of this year. Its net loss was more than $15 billion in the first quarter, but that was skewed by an accounting charge of roughly $12 billion, reflecting a change in fair value on equity issued to investors.

The adjusted operating margin improved from negative 13% in the first quarter to a positive, single-digit percentage in the second quarter, it told investors. The Financial Times reported last week that Anthropic told a small group of investors it would turn a profit on that basis again in the third quarter.

Big Spending Plans

Anthropic has been spending more heavily on data center deals, including a May agreement to pay Elon Musk’s SpaceX $1.25 billion a month for computing infrastructure over the next few years. That single deal represented the equivalent of two-thirds of the $1.9 billion Anthropic spent monthly across all its computing infrastructure in the quarter that ended in June.

Anthropic executives told investors they don’t expect to continue paying that much per gigawatt of power in future deals. They said its spending would be tilted heavily toward training its next models so that it could develop the “future frontier.” The company teased investors with potential applications of its tech beyond coding, saying in its presentation those would include drug discovery, autonomous cybersecurity, supply chain automation, individually tailored education and expert legal and financial advice.

Anthropic’s presentation was light on financial details. It didn’t include cash flow statements or balance sheets. But it laid out how quickly it was adding customers. The company said about 1,500 companies had spent $100,000 or more on its AI software over the previous 12 months at the end of 2025. By the end of the June quarter this year, that number of clients had increased fourfold to about 6,000.

It said more than 100 firms spent more than $10 million each with Anthropic in the 12-month period. More than 1,000 firms spent more than $1 million each. One of the prospective investors said Anthropic emphasized that it was in the early stages of expansion inside of big-spending customers like financial firms and insurance companies.

In its presentation, Anthropic highlighted customers like insurance giant AIG, which it said its technology helped cut the time in half for underwriters to get a quote for a new policy. It also pointed to improved productivity for legal firm Thomson Reuters as well as music app Spotify.

Xu, the Interconnected investor, said he would look for Anthropic to make the case in its IPO roadshow that its customers will continue to pay handsomely for the most intelligent models, rather than switching entirely to open-weight models. “In the same way [Apple’s] iOS ecosystem stuck—you pay extra but you find it fine and you can afford it and the experience improvement is worth the money,” he said. “I think Anthropic has to get there.”