The AI chip wars just got a major new chapter. Meta announced on Tuesday a sweeping multiyear deal with Advanced Micro Devices, deploying up to 6 gigawatts of AMD’s graphics processing units across its artificial intelligence data centers. The agreement also includes AI-optimized central processing units and marks one of the most significant challenges yet to Nvidia’s near-total dominance of the AI chip market.
The timing is striking. Just one week earlier, Meta had committed to deploying millions of Nvidia processors to fuel its AI expansion. Now, with a second blockbuster chip deal locked in, it is clear that Meta is not putting all its chips — literally — into one basket.
AMD’s Lisa Su Eyes the Center of Global AI
CEO Lisa Su framed the deal as a defining moment for AMD, saying the company is delivering high-performance, energy-efficient infrastructure built specifically around Meta’s workloads. She described the agreement as placing the chipmaker at the center of the global AI buildout — a bold claim for a firm that has long played second fiddle to Nvidia in this space.
Early shipments of MI450 GPUs inside its Helios rack-scale servers are expected to begin later this year. The Helios platform represents the first large-scale competitor to Nvidia’s Grace Blackwell systems, which have seen explosive demand since their 2024 launch, positioning AMD as a more serious challenger in the AI hardware race.
A Deal Sweetened With 160 Million Shares
Beyond hardware, the agreement carries a significant financial dimension that underscores just how serious both companies are. AMD issued Meta a performance-based warrant to acquire up to 160 million shares — roughly 10% of the company. The first tranche vests once the initial 1 gigawatt of Instinct GPUs ships, with additional tranches tied to purchase milestones, stock price thresholds, and technical benchmarks.
It mirrors a similar arrangement the chipmaker struck with OpenAI in October, where the same 160 million share warrant structure was used. That deal cemented AMD’s growing reputation as a credible second option for AI giants and hyperscalers looking to reduce their heavy dependence on Nvidia.
Meta Goes All-In on AI Infrastructure
This deal fits squarely within Meta’s broader and extraordinarily aggressive AI infrastructure push. In its earnings report last month, the company committed to up to $135 billion in capital expenditures for the year as it races to keep pace with Microsoft, Google, OpenAI, and Anthropic. Meta currently has plans for 30 data centers, with 26 of them located in the United States.
Chip analyst Ben Bajarin of Creative Strategies, who was briefed on the deal, estimates its total value runs into the tens of billions of dollars over at least four years. He also highlighted a key distinction between this deal and Meta’s arrangement with Nvidia — the AMD deployment involves customized GPUs, something Nvidia has not been publicly confirmed to offer in its Meta partnership.
What This Means for AMD and Nvidia’s Rivalry
The numbers tell a sobering story for AMD in today’s hypercompetitive AI market. Nvidia remains the world’s largest publicly traded company with a valuation of $4.66 trillion and controls roughly 90% of the AI chip market. By comparison, AMD sits at a $320 billion valuation. The gap is enormous — but deals like this one are precisely how that gap begins to steadily close.
Nvidia is set to report quarterly earnings on Wednesday, with analysts projecting revenue growth of 68% year over year to $66 billion. AMD, meanwhile, posted fourth-quarter sales growth of 34% to $10.27 billion. The contrast is sharp, but Tuesday’s announcement makes one thing undeniable — the company is no longer just chasing Nvidia. It is competing for the same table0.
