AMD, Anthropic strike major AI chip deal

Advanced Micro Devices (AMD) has agreed a significant artificial intelligence chip arrangement with Anthropic centered on AMD’s upcoming Instinct MI450 accelerator family. The structure of the agreement links a substantial hardware commitment to a reported investment component of up to $5 billion from AMD into Anthropic.

Multiple reports describe the deal as focused on high-performance AI servers, with Anthropic positioned as a major customer for AMD’s next-generation accelerators. The partnership is framed around building out large-scale infrastructure for training and running advanced AI models.

Anthropic is expected to acquire up to 2 gigawatts of AMD accelerator capacity under the agreement. This figure underscores the planned scale of MI450-based infrastructure dedicated to Anthropic’s AI workloads.

Shipments of Instinct MI450 chips tied to the deal are anticipated to begin in the first half of 2027. The schedule sets a multi-year deployment path that would bring AMD’s next-generation accelerators into Anthropic’s production environment as that hardware becomes available.

The reported investment of up to $5 billion from AMD into Anthropic is described as contingent on deployment milestones. This structure links AMD’s financial commitment to the pace and scale of the accelerator roll-out.

By tying funding to clearly defined deployment thresholds, the deal aligns incentives for both companies around timely installation and utilization of MI450 capacity. It also provides a framework for scaling the partnership in stages rather than as a single, upfront commitment.

The MI450-centered agreement positions AMD to expand its role in high-performance AI training and inference markets that have been heavily focused on competing accelerator platforms. Securing Anthropic as a large, forward-committed customer supports AMD’s efforts to grow share in advanced data center AI workloads.

For Anthropic, the arrangement lays out a path to substantial dedicated accelerator capacity over several years, underpinned by a structured investment and deployment plan. The scale and timing of the deal reflect a long-term build-out of infrastructure for increasingly demanding AI applications.