Artificial intelligence pioneer Anthropic is running into severe friction as it attempts to secure the massive cloud computing capacity required to train and run its next-generation Claude models, according to a leaked S-1 filing obtained by The Wall Street Journal.
The confidential regulatory filing reveals that major cloud providers—specifically Amazon Web Services (AWS), Google, and Microsoft—are increasingly viewing Anthropic not merely as a lucrative customer, but as a direct and formidable competitor in the rapidly evolving artificial intelligence marketplace. This shifting dynamic has complicated the company’s efforts to scale its infrastructure through traditional hyperscale partnerships, forcing Anthropic to navigate a complex web of supply constraints, astronomical financial commitments, and high-stakes operational risks just as it prepares to enter the public markets.
The newly surfaced details from the leaked S-1 document expand significantly upon earlier reports published by Reuters and the Financial Times in late September, which first brought Anthropic’s sweeping AI vision and staggering financial burn rate into public view. Among the most striking revelations in the filing is Anthropic’s projection that its cumulative spending on cloud contracts could ultimately exceed $750 billion—a figure that encompasses both existing agreements and massive deals currently under intense negotiation.
Even with this unprecedented capital expenditure plan on the table, the company acknowledges that it is still not securing enough compute capacity to satisfy its surging operational demands. To bridge the widening infrastructure gap, Anthropic has been forced to lease its own proprietary data centers and forge operational partnerships with third-party data center operators. However, these alternative arrangements have introduced a new layer of corporate anxiety. The filing reveals that Anthropic harbors serious concerns regarding these third-party providers, noting that many possess a “limited operating history” which could render them significantly less reliable than established hyperscalers during moments of peak demand or technical stress.
This immense financial and operational pressure comes on the heels of a bruising first half of the year for the AI lab. According to the leaked financial statements, Anthropic posted a staggering $80.7 billion net loss during the first six months of the year. During that same period, the company generated $16.3 billion in revenue, though its client base showed heavy concentration risks, with just two major customers accounting for approximately one-fifth of total sales.
The financial exposure outlined in the prospectus is matched by stark warnings regarding technological and legal liabilities. Building upon previously disclosed catastrophic risk assessments, the S-1 filing explicitly cautions that Anthropic could be held legally liable for damages stemming from harmful or unintended AI behaviors. The company warns that potential liabilities in such scenarios could easily balloon to a scale that exceeds its available insurance coverage, presenting a unique existential risk profile for prospective public shareholders.
Despite these mounting hurdles, commercial momentum for Anthropic’s flagship Claude ecosystem remains undeniably robust. Last week, the company shared an encouraging update with prospective institutional investors, revealing that third-quarter revenue alone had surged to $16.8 billion—representing a dramatic 45% increase compared to the previous quarter. Bolstered by this rapid top-line growth, Anthropic is reportedly pressing forward with its market debut, targeting an initial public offering in November.
Representatives for Anthropic did not immediately respond to requests for comment regarding the leaked filing or the reported timeline for the upcoming public offering.