AMD Challenges Nvidia with 12-Gigawatt GPU Deals

AMD Secures 12-Gigawatt GPU Commitment from OpenAI and Meta

Advanced Micro Devices (AMD) is directly challenging Nvidia's market leadership by securing two landmark 6-gigawatt GPU supply agreements with OpenAI and Meta. The deals, announced in October 2025 and February 2026 respectively, establish AMD as a primary infrastructure provider for two of the world's leading AI companies. The first 1-gigawatt deployments for both partners are scheduled to begin in the second half of 2026, marking a significant expansion of AMD's data center footprint.

In a novel "equity for orders" strategy, AMD granted both OpenAI and Meta performance-based warrants to purchase up to 1.6 billion shares of its common stock each. CEO Lisa Su described the structure as a "win-win" that accelerates purchasing and the development of AMD's software ecosystem. The warrants vest based on deployment milestones and performance targets, aligning the long-term interests of the chipmaker with its largest customers. Analyst Ben Bajarin estimates the agreements are worth tens of billions of dollars over at least four years.

Su Visits Korea to Fortify Supply for New MI450 Chip

To support these massive customer commitments, CEO Lisa Su is personally directing a campaign to secure critical supply chain capacity. On March 18, Su is scheduled to visit South Korea for the first time in over a decade to meet with leaders from Samsung and Naver. The primary goal is to lock down the supply of HBM4 memory, a crucial component for AMD's next-generation AI accelerators. The company has already reserved approximately 8% of TSMC's CoWoS advanced packaging capacity for 2026, estimated at 90,000 wafers, to produce its new chips.

This supply chain push coincides with the launch of AMD's most ambitious products. At CES 2026, AMD unveiled its MI450 series, featuring the MI455X chip that delivers a 10x performance increase over its predecessor. These chips will power the new open-standard "Helios" server, a 72-card system providing 2.9 Exaflops of FP8 compute. By securing component supply and manufacturing capacity, AMD aims to ensure it can deliver these high-performance systems at scale.

Data Center Growth Underpins 35% Annual Target

AMD's strategic offensive is built on strong financial performance. The company's data center division revenue grew 39% year-over-year in 2025 to a record $5.4 billion, underscoring the success of its current product lineup. This momentum supports management's ambitious financial targets, which include achieving a compound annual growth rate (CAGR) exceeding 35% over the next three to five years and generating over $20 in earnings per share.

We are just at the beginning of a decade-long AI super cycle.

— Lisa Su, CEO, AMD

The company's strategy focuses on capturing a significant share of the AI infrastructure market, where capital expenditures are projected to reach $600 billion by 2026. By securing cornerstone clients, investing in a competitive product roadmap, and proactively managing its supply chain, AMD is positioning itself as a formidable competitor in the rapidly expanding AI landscape.