Meta partners with AMD for AI chips after planning Nvidia GPU deployment

• The AMD deal involves customized GPUs in Helios rack-scale servers starting later this year, a key differentiator from Meta's Nvidia agreement which does not include customization
• Meta is investing up to $135 billion in capital expenditures this year and plans 30 data centers (26 in the U.S.) as it competes in the global AI race
• AMD's 6-gigawatt deployment represents the first large-scale competition to Nvidia's Grace Blackwell systems, though Nvidia still controls roughly 90% of the AI chip market with a $4.66 trillion valuation

Meta Expands AI Chip Strategy with Major AMD Deal

Key Development:

Meta has signed a multiyear agreement with AMD to deploy up to 6 gigawatts of GPUs for AI data centers, just one week after committing to using millions of Nvidia processors. This dual-supplier strategy represents one of the industry's largest AI deployments.

Deal Specifics:

• Early shipments of AMD's MI450 GPUs in Helios rack-scale servers begin later this year
• Agreement includes AI-optimized CPUs alongside GPUs
• Estimated value: tens of billions of dollars over at least four years, according to analyst Ben Bajarin of Creative Strategies
• Critical differentiator: AMD is providing customized GPUs for Meta, unlike the Nvidia arrangement

Financial Context:

• Meta committed up to $135 billion in capital expenditures this year for AI infrastructure
• Plans include 30 data centers (26 in the U.S.)
• AMD valued at $320 billion versus Nvidia's $4.66 trillion market cap
• Nvidia controls approximately 90% of the AI chip market

Market Implications:

This deal marks a significant breakthrough for AMD in the AI chip sector, with AMD's Helios systems representing the first large-scale competition to Nvidia's dominant Grace Blackwell platforms. The partnership validates AMD's position as a viable alternative supplier in a compute-constrained market.

Meta's diversified chip strategy demonstrates the company's leverage in controlling its full technology stack while mitigating supply chain risks. The timing—shortly before Nvidia's Wednesday earnings report (expected $66 billion revenue, up 68% year-over-year)—underscores the intense competition in AI infrastructure.

AMD CEO Lisa Su emphasized delivering "high-performance, energy-efficient infrastructure" optimized for Meta's specific workloads, positioning AMD centrally in the global AI buildout.