When the Suppliers Become Competitors
For most of the generative AI boom, the relationship between frontier model developers and the cloud giants looked simple. The labs needed enormous amounts of compute, the hyperscalers had it, and both sides profited from the arrangement. That tidy picture is starting to fray. According to a leaked S-1 filing reported by The Wall Street Journal, Anthropic is struggling to secure enough capacity from Amazon Web Services, Google, and Microsoft, all of which increasingly view the Claude maker as a competitor rather than a reliable customer.
The scale of what Anthropic is trying to buy is striking. The company believes it could commit more than $750 billion to cloud contracts, and some of those agreements are still under negotiation. Even that figure apparently is not enough. Anthropic is leasing its own data centers and working with third-party operators, but it has flagged concern that those newer providers have a “limited operating history,” which could make them less dependable than the established hyperscalers during periods of peak demand.
The Economics Behind the Race for Capacity
The financial picture helps explain why this matters. The filing reflects an $80.7 billion net loss in the first half of the year, set against $16.3 billion in revenue over the same period. Only two customers accounted for roughly a fifth of sales, a concentration that would unsettle any investor. Meanwhile, the company told prospective backers that third-quarter revenue alone reached $16.8 billion, up 45 percent from the prior quarter. Growth at that pace is impressive, but it also raises the stakes for every gigawatt of capacity that fails to arrive on schedule.
The filing also warns that Anthropic could face liability for damages caused by harmful AI behavior that exceed its available insurance coverage. That disclosure signals how the industry is beginning to price catastrophic risk alongside compute costs. Insurers, lenders, and enterprise customers will all be watching how such exposures are managed as the company reportedly eyes a November IPO.
What This Means for the Broader AI Market
The tension between Anthropic and its cloud partners is a preview of a wider structural question. If the largest providers of AI infrastructure also build competing models, smaller developers may find it harder to access the processing power they need, and enterprise buyers may face uneven service levels depending on which platform hosts their models. Customers who have standardized on a single cloud could find that the AI tools they depend on are constrained by a rival’s priorities.
For businesses evaluating AI tools, the practical lesson is to look beyond model benchmarks and ask hard questions about capacity guarantees, uptime commitments, and the financial durability of the vendors involved. For consumers, the effects will show up more slowly in pricing, feature rollouts, and the reliability of AI-powered apps. Adoption depends on confidence that the tools will be there tomorrow, and right now the infrastructure behind them is anything but settled. Buyers who pay attention to these supply dynamics today will be better positioned to choose AI products that keep working as demand keeps climbing.
