AMD will invest up to $5 billion in Anthropic as part of a sweeping deal that includes tens of billions of dollars worth of AI server purchases, the companies announced.

The agreement centers on Anthropic purchasing up to 2 gigawatts of AMD's Instinct MI450 chips starting in the first half of 2027. The investment will be released in tranches as certain deployment milestones are met.

The partnership represents AMD's most aggressive push into the AI infrastructure market, directly challenging Nvidia's dominance in training and inference hardware. AMD's MI450 series targets the same high-performance computing workloads that currently rely on Nvidia's H100 and upcoming Blackwell chips.

Why This Deal Matters

For Anthropic, the arrangement secures dedicated compute capacity for training future Claude models and scaling inference operations. The AI safety lab has been expanding its enterprise offerings, recently launching Claude-driven "managed projects" for persistent task management.

AMD gains a marquee customer to validate its AI chip roadmap. The 2-gigawatt commitment would represent one of the largest single AI infrastructure deployments outside of hyperscale cloud providers.

The companies are jointly identifying data center locations for the chip deployments. Industry sources suggest the facilities will likely be distributed across multiple regions to support Anthropic's global expansion plans.

Meanwhile, OpenAI separately announced its infrastructure spending plans have grown to $750 billion through 2030, with a $20 billion data center campus in Georgia as the first major project.

The AMD-Anthropic deal comes as AI companies face mounting pressure to secure dedicated compute resources amid ongoing chip shortages and geopolitical tensions affecting supply chains.

Both companies expect the first MI450 deliveries to begin in the second quarter of 2027, with full deployment scaling through 2029.