Anthropic’s cloud partnerships are getting messy

Hyperscalers may limit AI startup access to the compute they desperately need.

Anthropic faces significant challenges securing sufficient cloud computing capacity from major providers AWS, Google, and Microsoft. These cloud giants increasingly view Anthropic, the maker of the Claude AI model, as a competitor rather than a primary customer, complicating their partnership dynamics. A leaked S-1 filing reveals Anthropic's concerns about the reliability of third-party data center operators it is now engaging, alongside leasing its own facilities, to supplement hyperscaler capacity. The company anticipates potentially spending over $750 billion on cloud contracts, with a substantial portion still under negotiation. This scramble for resources occurs amidst Anthropic's substantial financial performance, reporting an $80.7 billion net loss in the first half of the year against $16.3 billion in revenue, with two customers contributing approximately 20% of sales. The filing also highlights potential liability risks for harmful AI behavior exceeding insurance coverage, even as Anthropic reported $16.8 billion in third-quarter revenue, a 45% increase from the prior quarter, and targets a November IPO.

Anthropic's struggle for cloud resources underscores a broader market shift where infrastructure providers are increasingly integrating AI capabilities, leading them to prioritize their own AI development over serving potential competitors. This strategic pivot by AWS, Google, and Microsoft suggests a maturing AI landscape where access to essential computing power could become a significant bottleneck for independent AI developers. The company's aggressive spending and reliance on less proven data center partners highlight the immense capital expenditure required to compete at the forefront of AI, a burden that could deter future entrants or force consolidation.

The situation implies that AI development may become increasingly concentrated among a few vertically integrated players who control both the AI models and the underlying infrastructure. For other AI companies, this necessitates developing more capital-efficient strategies, exploring alternative hardware solutions, or forging partnerships with entities less directly competitive with their core offerings. Anthropic's potential IPO may be a crucial test of whether it can execute its ambitious plans despite these infrastructure constraints and competitive pressures.