Anthropic, the artificial intelligence company behind the Claude chatbot, has agreed to a $10 billion deal for computing capacity from a startup that launched just months ago. The agreement, one of the largest of its kind, gives Anthropic access to massive amounts of processing power needed to train and run its AI models.
Why the deal was struck
AI companies are racing to secure computing resources. Training large language models requires thousands of specialized chips, often from Nvidia, and the cloud infrastructure to run them. By locking in capacity early, Anthropic avoids the scramble that has hit other firms. The startup, which has not been named publicly, is building data centers designed specifically for AI workloads. It's a bet that this new player can deliver at a scale that rivals established cloud providers.
What the startup brings
The infrastructure company is only a few months old. That makes the $10 billion commitment a huge vote of confidence. The startup's founders have backgrounds in hardware and data center design, but the company has yet to prove it can operate at the level Anthropic needs. The deal covers multi-year access to computing clusters, though exact terms haven't been disclosed. For a young company, landing a contract this size means it must scale fast — or risk falling short.
Anthropic's move signals that the biggest AI players are willing to pay a premium for guaranteed compute. Microsoft has invested billions in OpenAI, and Google uses its own infrastructure. But a deal with a months-old startup is unusual. It shows that demand is outstripping supply from the usual sources. Other AI companies may now look to similar partnerships with newer infrastructure firms. That could reshape the cloud market, where Amazon, Microsoft, and Google have long dominated.
The $10 billion figure is striking. It's more than many AI startups have raised in total. Anthropic itself has raised billions from investors including Google and Salesforce. But this deal is for capacity, not equity. The startup gets a steady revenue stream; Anthropic gets the chips and data center space it needs to keep its models competitive.
Risks on both sides
For the startup, the challenge is execution. Building data centers takes time and money. Delays could leave Anthropic without the promised compute. For Anthropic, the risk is locking into a single provider. If the startup stumbles, Anthropic may have to scramble for alternatives. The contract likely includes penalties for underperformance, but those details aren't public.
Regulators may also take an interest. A $10 billion deal between an AI company and a new infrastructure firm could raise questions about market concentration and access. No agency has announced a review, but the size of the deal makes it a potential target.
Anthropic hasn't said when the first computing clusters will come online. The startup is racing to deliver. For now, the industry is watching to see if a company that barely exists can handle one of the biggest compute deals ever signed.
