Sources detail Anthropic's compute deal spree: 3.5GW+ of expected capacity in 2026, diversifying beyond Nvidia's AI chips to Amazon, Google, and AMD, and more

The cost of large-scale AI model training and inference will fall for companies able to aggregate demand through cloud providers.

Anthropic, a leading AI safety and research company, is reportedly securing significant compute capacity for 2026, with over 3.5 gigawatts expected. This expansion involves a diversification strategy away from a sole reliance on Nvidia's AI chips and infrastructure. The company is in advanced talks to lease large data center spaces from Amazon Web Services, Google Cloud, and Microsoft Azure. Additionally, Anthropic is exploring hardware partnerships, including orders for AMD's upcoming MI300 AI accelerators, and has also engaged with other chipmakers like Intel and Broadcom for networking gear. This aggressive build-out aims to secure the necessary computational power to train and deploy its advanced AI models, such as Claude 3, at scale, signaling a major investment in its future growth and competitive positioning in the AI landscape.

Anthropic's expansive compute deal spree signifies a critical industry trend: the escalating demand for specialized AI hardware and the subsequent logistical challenges of securing it. By diversifying its GPU and infrastructure suppliers beyond the current dominant player, Nvidia, Anthropic is hedging against supply constraints and potential price hikes. This move also suggests a maturing AI market where the economics of large-scale model development necessitate strategic, long-term capacity planning and multi-vendor relationships to ensure operational continuity and cost efficiency.

The broad engagement with cloud providers like AWS, Google Cloud, and Azure, coupled with hardware orders from AMD and interest from Intel, indicates that the era of AI compute being a bottleneck is intensifying. Companies are now treating compute procurement as a strategic imperative, akin to securing raw materials for manufacturing. This competition for resources, however, may lead to more favorable pricing and tailored solutions for large-scale AI players like Anthropic, while potentially exacerbating availability issues for smaller entrants.