AMD Secures 2.5-Gigawatt AI Data Center Deal with Core Scientific

What happened

AMD entered a 15-year master agreement with Core Scientific to access high-density AI colocation capacity starting in 2027. The partnership begins with more than 500 megawatts of U.S. infrastructure and could expand to 2.5 gigawatts across Core Scientific's data center network. Under the terms, Core Scientific will provide AMD with power, land, and infrastructure to deploy Instinct GPUs, EPYC CPUs, and ROCm software for AI workloads. AMD received market-priced warrants to purchase up to 30 million shares of Core Scientific common stock at $23.47 per share, vesting as hosting capacity comes online. Core Scientific announced the initial commitment alone is expected to generate over $14 billion in contracted revenue over the agreement's term.

Why it matters

The partnership addresses a critical infrastructure bottleneck as AI deployments accelerate. For AMD, securing dedicated data center capacity positions the company to meet growing demand from cloud providers, model developers, and enterprise customers deploying its AI chips. The agreement provides long-term visibility into where AMD's hardware will be deployed and ensures power availability, a key constraint in AI infrastructure. The warrant structure aligns AMD's interests with Core Scientific's execution, limiting immediate dilution while providing upside if the partnership succeeds. For investors, the deal highlights that AI competition is expanding beyond chip design to include power and physical infrastructure as differentiators. The agreement pushes Core Scientific's total leased footprint to approximately 1.1 gigawatts and its lifetime contracted revenue pipeline past $24 billion, demonstrating the scale of infrastructure investment required to support AI workloads.

Bigger picture

The partnership reflects a broader shift in AI infrastructure investment from chips alone to integrated power and colocation solutions. Core Scientific, which has pivoted from Bitcoin mining to AI data centers and now derives 77% of revenue from AI colocation, represents a new class of infrastructure provider essential to scaling AI deployments. The multi-gigawatt scale of the agreement underscores the massive power requirements of modern AI workloads and the premium companies like AMD are willing to pay for guaranteed capacity. The deal follows similar moves by other chip makers to secure long-term infrastructure access as power availability becomes a competitive bottleneck. AMD executive Mathew Hein stated that trusted infrastructure partners are critical as AI deployments accelerate, while Core Scientific CEO Adam Sullivan emphasized the company's execution record in supporting AMD's technology roadmap. The market's divergent reaction-Core Scientific shares rising 10–11% while AMD fell 4–7%-suggests investors view this primarily as infrastructure securing rather than immediate revenue generation for the chip maker.

What to watch

Investors should monitor AMD's ability to fill the secured capacity before 2027 and whether additional customers are signed to justify the infrastructure investment. Key milestones include the sequential vesting of AMD's warrants as Core Scientific brings capacity online, which will signal execution progress. Watch for updates on power costs, utilization rates, and whether the partnership expands toward the full 2.5-gigawatt potential. Broader industry trends around AI infrastructure spending, power availability, and competing colocation deals will provide context for the partnership's strategic value. AMD's ability to deploy its Instinct GPU and EPYC processor roadmap through this dedicated capacity will be critical to converting infrastructure access into competitive advantage in AI markets.