The Core Scientific agreement
AMD said it has partnered with Core Scientific to secure up to 2.5 gigawatts of AI data center capacity in the United States, with more than 500 megawatts of that footprint planned from 2027. The collaboration is framed around large-scale AI deployments, pairing AMD's Helios rack-scale systems, Instinct GPUs, EPYC CPUs and ROCm software with dedicated physical infrastructure. The arrangement also includes joint work on physical design and ROCm software, positioning AMD as a fuller-stack alternative to Nvidia- or Intel-based deployments.
How the buildout fits AMD's AI stack
The Core Scientific capacity is designed to give AMD a concrete channel to deploy its AI hardware at scale rather than only supply chips. The reference covers facility power, facility cooling, IT space and lifecycle software, with AI clusters sized up to 10.4 MW IT capacity for greenfield deployments. By locating more than 500 megawatts of AI-ready capacity in the U.S. from 2027, the partnership also targets customers with domestic-infrastructure, data-residency or sovereign-AI requirements.
Execution and competitive risks
Filling up to 2.5 gigawatts depends on continued AI spending and on Core Scientific executing its buildout plan, leaving room for under-utilised capacity if demand or funding conditions shift. AMD's analysts have already flagged valuation and execution as key risks, and layering large infrastructure commitments on top could heighten sensitivity if GPU supply, customer ramps or competition from Nvidia and Intel do not line up with expectations.
Outlook into Q2 earnings
Investors will be watching how quickly AMD converts the Core Scientific agreement into signed customer contracts and how often the company references the capacity on upcoming earnings calls. Progress on the initial 500-megawatt tranche from 2027, pricing terms for AI infrastructure and any utilisation disclosures will help show whether the partnership is tracking to plan.
Share this article
