Anthropic co-founder and president Daniela Amodei presented a striking picture last month as CFO Krishna Rao and other executives met with multiple groups of prospective public-market investors. The company's financial data hit record highs. In the spring of 2025, the firm spent about $2.30 in operating costs for every $1 of revenue it generated. Yet by the June quarter of this year, that metric flipped to a slight profit, shattering the stereotype of AI companies burning cash at unsustainable rates. However, some investors left the meetings with a clear message from management: this profitability is likely not here to stay. Executives told investors that Anthropic plans to use its rapid business growth and substantial cash reserves as a window to aggressively sign data center procurement agreements, which will drive up computing infrastructure spending and squeeze operating margins.

These conversations are part of Anthropic's pre-IPO "testing the waters" outreach. The trajectory of this IPO carries outsized weight, influencing hundreds of companies queued up to go public and rippling through the broader macroeconomic landscape. Anthropic has not yet formally kicked off the multi-week IPO process, but once it does, it will ignite preparations that have been kept under tight wraps. Many investors had initially expected a listing as early as this month, but the timeline has now slipped to late October or November. Next up, Anthropic will invite analysts and investors to participate in deep-dive roadshow presentations, a critical pre-IPO step that typically occurs weeks before the public filing of the prospectus. After the filing, it typically takes another three to four weeks for shares to begin trading publicly.

The company's next round of Wall Street roadshows will face far tougher scrutiny than the August sessions. Rival OpenAI has since launched an advanced model that directly matches Anthropic's capabilities and is quietly preparing a new private fundraising round fueled by product momentum. Both OpenAI and Anthropic sit at the center of Silicon Valley and Washington discussions about AI safety and development. At the same time, low-cost open-source weight models pose a risk of squeezing both firms' businesses. Adding to the pressure, rising interest rates are making it more difficult for companies to finance the construction and expansion of data centers. Karen Snow, a former Nasdaq executive who now leads investor relations consultancy Rose & Co, noted that with so many factors converging, investors will examine Anthropic's IPO with a far more cautious eye. She described the offering as having shifted "from a surefire prize to an uncertain asset. Market sentiment is fickle, the winds have changed in the AI sector, and a cloud of uncertainty now hangs over the field."

The battle between public companies and secondary-market investors over revenue expectations and spending plans will be more intense this time than in past cycles. Anthropic's record-breaking growth is matched by massive outlays on power, chips, and data centers, which in turn ripple through a wide swath of related stocks. Company executives are trying to make clear that the firm remains well-funded heading into its IPO. They disclosed to investors that as of early August, cash stood between $120 billion and $130 billion. But they also acknowledged that as computing infrastructure investment ramps up, especially for training next-generation AI models, that war chest will steadily dwindle.

Research and development continues despite AI safety debates. CEO Dario Amodei has called for increased safety evaluations and regulation of frontier models, an informal commitment framework known as "pacing the frontier." But he has also been clear that "pacing doesn't mean halting model training or pausing technical iteration." Going public is crucial to Anthropic's computing plans. An IPO not only raises tens of billions in fresh capital in a single transaction, but also opens the door for the company to tap debt and equity markets repeatedly over the years ahead. Anthropic needs to craft a compelling commercial narrative that balances Wall Street's demand for a clear profitability path. Kevin Xu, founder of Interconnected Capital, a fund focused on AI infrastructure stocks (who has not met formally with Anthropic), said: "Listing means Anthropic formally enters the entire capital market with far more financing channels. But secondary-market investors want assurance that Anthropic can self-fund its capital expenditures, and that debt is a temporary bridge, not an endless cycle of borrowing. This is the shared challenge facing the entire AI industry right now."

At the heart of Anthropic's pitch is persuading the market to accept what it describes in roadshow materials as an unprecedented financial model: revenue grew roughly 1.5 times from the first quarter to the second quarter of this year, climbing from about $4.7 billion to over $11.5 billion. During the same period, computing infrastructure spending rose only about 65%, from $3.4 billion to $5.6 billion. Anthropic is highlighting adjusted operating margin as its key profitability metric. This is a commonly used non-standard financial measure that excludes non-cash costs such as employee stock compensation and the company's large equity donation matching program. In the first quarter alone, these excluded items totaled more than $1 billion. The company posted a net loss exceeding $15 billion in the first quarter, largely driven by an accounting adjustment of roughly $12 billion related to the change in fair value of equity issued to investors. The company told investors its adjusted operating margin swung from negative 13% in the first quarter to positive single digits. The Financial Times reported last week that Anthropic has communicated to a select group of investors that it expects to remain profitable on this basis in the third quarter.

Anthropic's spending plans are enormous. Over the past 11 months, the company has finalized computing procurement agreements totaling $517 billion. Among these, a deal reached in May commits the company to paying $1.25 billion per month to Elon Musk's SpaceX for computing capacity over the coming years. That single contract alone is equivalent to two-thirds of Anthropic's total monthly computing spending for the quarter ending in June, which stood at $1.9 billion. Executives have told investors that the cost per gigawatt of power in future agreements is not expected to stay at current elevated levels. The majority of capital will be directed toward training next-generation models to build future frontier capabilities. The roadshow materials also showcased applications beyond coding, spanning drug discovery, autonomous cybersecurity, supply chain automation, personalized education, and specialized legal and financial consulting. The financial details in the pitch deck are sparse, with no cash flow statement or balance sheet, but customer expansion metrics are prominent. The company says that as of the end of 2025, about 1,500 enterprises had spent at least $100,000 on its AI software over the trailing 12 months. By the end of the June quarter, that customer count had quadrupled to roughly 6,000. More than 100 companies spent over $10 million annually with Anthropic over the past 12 months, and over 1,000 companies spent more than $1 million a year. One prospective investor said Anthropic emphasized that its expansion within large accounts in high-paying industries like finance and insurance is still at an early stage. The pitch deck lists insurance giant AIG as a marquee customer, noting that underwriters have cut the time to produce new policy quotes by half using Anthropic's technology. It also highlights productivity gains at Thomson Reuters law firms and counts music platform Spotify among its clients.

Investor Xu said he expects Anthropic's IPO roadshow to prove that customers are willing to keep paying premium prices for top-tier frontier models rather than migrating entirely to open-source weight models. "It's like the logic of Apple's iOS ecosystem. You're willing to pay extra because the experience justifies the cost. Anthropic has to make that case."