Sources detail Anthropic's compute deal spree: 3.5GW+ in expected capacity in 2026, expanding direct hardware procurement to include Nvidia and AMD, and more

The cost of training large AI models increases significantly, as hardware procurement becomes a direct expense rather than a variable cloud cost.

Anthropic is reportedly securing massive amounts of computing power, with plans for over 3.5 gigawatts of capacity by 2026. This expansion includes a significant shift towards direct procurement of hardware, moving beyond cloud providers. The company is now negotiating deals directly with chip manufacturers like Nvidia and AMD, alongside traditional cloud providers such as Amazon Web Services and Google Cloud. These new agreements are expected to cover Anthropic's substantial growth in AI model training and deployment needs. The scale of this commitment suggests a strategic move to ensure access to cutting-edge AI hardware and control over its infrastructure as the demand for advanced AI capabilities escalates.

Anthropic's aggressive compute procurement strategy signals a maturing AI infrastructure landscape, where leading companies are increasingly bringing hardware acquisition in-house to secure supply and potentially reduce long-term costs. This direct engagement with chip giants like Nvidia and AMD highlights the intense competition for advanced AI processors, suggesting that cloud providers might face greater upstream pressure on supply and pricing. The move also points to a growing trend of AI developers seeking greater control over their compute environment, potentially leading to more specialized hardware configurations and optimized AI workloads.

This development could reshape the economics of AI development, making large-scale hardware ownership a new prerequisite for major AI players. For cloud providers, it signifies a potential shift in market dynamics, with hyperscalers needing to offer more compelling value propositions beyond raw compute to retain demanding AI clients. Investors and competitors will be watching closely to see if Anthropic's direct procurement model proves more cost-effective and scalable than relying solely on cloud infrastructure, potentially setting a new benchmark for AI hardware strategy.