Core Scientific and AMD have formed an infrastructure partnership under which AMD will secure up to 2.5 gigawatts of data center capacity for customers deploying its artificial intelligence technologies.
The companies will collaborate on physical infrastructure design and the deployment of AMD Instinct graphics processing units, EPYC processors, and ROCm software.
The agreement will initially provide the AMD ecosystem with access to more than 500 megawatts of U.S. infrastructure beginning in 2027. The relationship has the potential to expand to up to 2.5 gigawatts over time.
Access to electrical power, land, cooling systems and purpose-built data centers has become a major constraint on the deployment of new AI computing capacity.
AMD develops processors and software used by model developers, cloud providers and enterprises, but customers also need facilities capable of supporting high-density hardware installations.
Core Scientific specializes in designing, constructing and operating data centers for high-density colocation workloads. The company’s facilities are designed to support computing systems that require substantial power and cooling capacity.
The partnership is intended to give AMD customers a more direct path to infrastructure optimized for the company’s processors and software.
AMD Instinct GPUs are designed for AI training, inference, and high-performance computing workloads. EPYC CPUs support general-purpose data center processing, while ROCm provides the software environment for operating and developing applications on AMD accelerators.
Coordinating the physical infrastructure with these technologies could help customers reduce the time required to prepare sites, install hardware and bring computing clusters online.
The agreement also positions Core Scientific as an important infrastructure provider within AMD’s broader effort to compete in the expanding AI accelerator market.
Core Scientific currently operates facilities in Alabama, Georgia, Kentucky, North Carolina, North Dakota, Oklahoma and Texas. The company generates most of its revenue from high-density colocation services and is repurposing remaining cryptocurrency mining facilities for AI and other advanced computing workloads when feasible.
Converting mining facilities can provide a faster route to AI capacity because those locations may already have access to substantial electrical power and supporting infrastructure. However, AI deployments can require different networking, cooling and reliability standards than cryptocurrency mining.
Core Scientific’s work with AMD will involve preparing facilities for dense processor configurations and supporting the requirements of model builders, cloud platforms and enterprise customers.
AMD will also receive warrants to purchase Core Scientific common stock at market terms. The warrants are subject to specified commercial conditions, and the companies did not disclose the potential number of shares or the full financial terms.
The equity component could give AMD an additional interest in Core Scientific’s growth if the infrastructure relationship expands.
For Core Scientific, the agreement creates an opportunity to further diversify away from cryptocurrency-related revenue and to secure additional demand for its high-density data center portfolio.
The planned scale is significant. One gigawatt equals 1,000 megawatts, so the full 2.5-gigawatt opportunity could support either several large AI campuses or a distributed network of facilities.
Actual deployment will depend on customer demand, site development, electrical availability, equipment supply, and the commercial conditions contained in the agreement.
KEY QUOTES:
“Core Scientific’s extensive portfolio of AI-ready data centers expands access to the infrastructure our customers need.”
Mathew Hein, Senior Vice President and Chief Strategy Officer of Corporate Development at AMD
“Our ability to deliver high-density infrastructure at scale positions us to support AMD’s technology roadmap.”
Adam Sullivan, CEO of Core Scientific
