Anthropic Warns of Operator Reliability Risk as It Leases Data Centers

Anthropic is leasing data centers and working with outside operators to expand Claude’s computing capacity, adding a reliability risk as it struggles to secure enough computing capacity from major cloud partners.

Some of those operators have limited operating histories and may be less reliable than major cloud providers. The concern adds a new dimension to Anthropic’s expansion as it prepares a proposed initial public offering.

Amazon remains Anthropic’s primary cloud provider and training partner. Anthropic has also committed $30 billion to Microsoft Azure capacity, while Google is another major capacity provider. All three companies compete with Anthropic in AI.

Anthropic has explored a potential lease of up to 1 gigawatt of data-center capacity in the United States.

The company has also signed a multiyear AI cloud services deal with IREN, a former Bitcoin (BTC) miner that has shifted into AI data centers. The agreement gives Anthropic access to GPU computing capacity beyond its major cloud partners.

An updated confidential IPO draft lists about $450 billion in cloud agreements over seven to 10 years. Broader computing commitments exceed $600 billion, while potential spending could reach $750 billion if arrangements still under negotiation are included.