Anthropic's reliance on cloud providers like AWS, Google, and Microsoft for Claude raises risks of service disruptions and liability issues, as revealed in a leaked prospectus.
Capacity issues with cloud providers
An organization purchasing Claude, Anthropic’s AI model, directly or through cloud providers like AWS, Google Cloud, or Microsoft Foundry may encounter situations where these companies do not supply the full computing capacity requested. According to a draft prospectus reported by the Wall Street Journal, these three cloud providers have historically supplied less capacity than Anthropic requested. Suspension terms for service disruptions
The commercial terms for direct access to Claude include provisions for when a supplier stops providing necessary services. In such cases, Anthropic may suspend access to the model, either fully or partially, without being required to compensate the client. The reported document does not mention any actual service disruptions, only a shortfall in capacity compared to requests. Prospectus details and media coverage
The prospectus, which is not yet public, was reported by the Wall Street Journal on October 20, 2026, with Reuters having previously described an earlier version in late September. The content may still change before its official filing. The figures from the WSJ are consistently reported by Reuters, TechCrunch, Investing.com, and investingLive. Anthropic’s silence and changing relationships
Anthropic did not comment on the reports, according to investingLive, and did not respond to inquiries from Reuters and TechCrunch on the same day. TechCrunch notes that AWS, Google, and Microsoft—now seen more as competitors than clients—were previously all investors, clients, suppliers, distributors, and competitors of Anthropic. Their interests may not have fully aligned with Anthropic’s. Financial commitments and infrastructure costs
According to the WSJ, Anthropic’s cloud contracts represent about 450 billion dollars over seven to ten years, mostly non-cancelable. This includes 111 billion with Google Cloud, 110 billion with AWS, and 31 billion with Microsoft. Additional commitments, including chip leases and agreements under negotiation, could push the total close to 750 billion dollars. Third-party data centers and chip capacity
Anthropic is also relying on third-party data centers, though the document from TechCrunch notes their operational history is limited. The WSJ estimates commitments for Nvidia chip capacities with Nscale and Lambda at up to 86 billion dollars. However, Nscale’s own prospectus indicates that its agreements with Anthropic, totaling up to 44.6 billion dollars, lack firm financing. Service suspension and liability terms
The commercial terms for direct access to Claude, effective since June 17, 2025, include provisions for service suspension due to attacks, client violations, legal restrictions, or increased supplier costs. Anthropic commits to "reasonable efforts" to restore access once the issue is resolved, but it excludes liability for damages or lost profits and provides the service "as is," with no guarantee of uninterrupted use. Past service disruptions and legal issues
A legal prohibition previously caused Anthropic to suspend two AI models, Fable 5 and Mythos 5, for all clients after an American export control directive on June 12. These models were restored only partially after the restriction was lifted on June 30, with Fable 5 returning on July 1 and Mythos 5 only available to U.S. organizations. Rising revenue and infrastructure costs
Anthropic reported 16.3 billion dollars in revenue in the first half of 2026, compared to 1.1 billion a year earlier. However, it spent nearly 9 billion on computing and infrastructure in the same period, more than the 7.3 billion for the entire year of 2025. The company’s net loss for the semester reached 80.7 billion dollars, with a significant portion related to accounting adjustments. Client concentration and liability risks
According to the WSJ, two unnamed clients accounted for about a fifth of Anthropic’s revenue in the first half of 2026. The company expects a limited number of clients to continue having a significant impact. The prospectus warns that Anthropic could face liabilities exceeding its insurance coverage for harmful behaviors of its models, including catastrophic or existential risks. IPO plans and regulatory processes
Anthropic aims to conduct an IPO in November, with the SEC requiring a public filing of the prospectus at least fifteen days before the investor roadshow. This will allow investors to review the company’s risk factors and other details as filed.
