AMD supplies Anthropic with 2GW of Instinct GPUs

AMD Powers Up AI Giant with Massive Compute Deal

The race for cutting-edge artificial intelligence is increasingly a battle waged in the realm of silicon and sheer computational power. In a move that signals a deep commitment to building the future of AI infrastructure, AMD has teamed up with Anthropic, embarking on an ambitious project designed to power one of the world’s leading language model developers.

The collaboration is not just about chips; it’s a strategic pact spanning hardware deployment and financial investment. AMD has committed to deploying up to 2 gigawatts of its Instinct MI450 Series GPUs for Anthropic, signaling an unprecedented level of trust in AMD’s architecture for handling massive AI workloads.

Adding further weight to this partnership, AMD is also making a strategic equity investment of up to $5 billion into the Claude developer, solidifying their role as both a supplier and a shareholder in the AI revolution.

This joint venture involves deploying cutting-edge hardware within AMD’s Helios rack-scale systems. The initial deployment of one gigawatt of capacity is slated to come online in the first half of 2027, setting an aggressive timeline for realizing this massive compute potential.

The technical synergy behind this deal is impressive. Anthropic’s infrastructure will leverage Instinct MI455X GPUs paired with EPYC “Venice” CPUs and Pensando networking. These systems are packed with high-bandwidth memory, including 31 terabytes of HBM4, delivering up to 2.9 FP4 exaFLOPS of inference compute per accelerator.

Beyond the physical hardware, the partnership emphasizes collaborative innovation. The two companies have launched a multi-year engineering collaboration aimed at optimizing workloads for Instinct GPUs and accelerating ROCm development. This teamwork ensures that the specialized hardware is utilized to its fullest potential in training and running complex AI models like Claude.

As Tom Brown, co-founder and chief compute officer at Anthropic, noted, this diversified approach allows them to “map the right workloads to the right hardware,” underscoring the importance of tailoring infrastructure to demanding computational needs.

This ambitious AMD deployment places the company firmly at the center of the global AI supply chain. It contributes significantly to a broader compute portfolio that already includes partnerships with Nvidia, Amazon’s Trainium chips, Google’s TPUs, and massive capacity rented from SpaceX’s Colossus 1 data center.

The impact of this strategic deployment is already visible across the industry. While Anthropic secures its own power, other major players are also aggressively deploying similar infrastructure; for instance, OpenAI is slated to receive its first gigawatt of AMD capacity in the second half of 2026, and Oracle is moving quickly to deploy 50,000 MI450 GPUs this quarter.

This monumental deal demonstrates how strategic hardware partnerships are reshaping the landscape of high-performance computing, pushing the boundaries of what large language models can achieve through innovative silicon design and collaborative engineering.