Anthropic Locks In Billions in AI Infrastructure Contracts
Claude Maker Ramps Up Massive Computing Commitments as Demand for AI Services Drives Expansion
Anthropic is rapidly expanding its access to computing infrastructure, committing billions of dollars to cloud and data-centre agreements as demand for its Claude artificial intelligence models continues to grow. Recent deals include a reported $35 billion agreement with Nvidia-backed cloud provider Lambda, adding to several other major infrastructure commitments.
The Lambda agreement is tied to a Texas data-centre project being developed by Hut 8 and is expected to provide Anthropic with substantial computing capacity. The facility is planned around hundreds of megawatts of power, highlighting the enormous amount of electricity and specialised hardware now required to operate advanced AI systems at scale.
Anthropic has also committed to other large infrastructure arrangements, including a reported $45 billion agreement with Nscale and a separate deal involving Microsoft and Nvidia through Azure. The company is increasingly securing capacity from multiple providers rather than depending on a single cloud partner, giving it greater flexibility as it scales Claude and its AI coding products.
The spending spree reflects a broader shift in the AI industry, where access to computing power has become one of the most important factors in developing and operating increasingly capable models. Anthropic had previously been cautious about making very large long-term commitments, but rapid growth in demand has pushed the company toward securing infrastructure much more aggressively.
Anthropic’s infrastructure push also comes as the company prepares for further expansion and reportedly considers a potential public listing. By locking in computing capacity ahead of time, the Claude maker is positioning itself to handle rising demand while competing with OpenAI, Google and other major AI companies. The scale of these agreements demonstrates how the AI race is increasingly becoming a race not only between models, but also between companies capable of securing enough chips, data centres and electricity to run them.
