AMD has secured an option on more than 500 megawatts of Core Scientific’s U.S. data-center capacity beginning in 2027, giving the chipmaker a potential route to scale that commitment to 2.5 gigawatts as demand for artificial-intelligence computing infrastructure expands.
The agreement places Core Scientific, historically one of North America’s largest Bitcoin miners, deeper into the market for high-density AI data hosting. Under the partnership announced Tuesday, the companies will work on infrastructure design and deployment plans for systems using AMD Instinct GPUs, EPYC server processors and ROCm software.
The initial 500-megawatt commitment is substantial in a sector where access to available power, suitably zoned land and data-center construction capacity can constrain the rollout of new AI clusters more than chip supply alone. A potential expansion to 2.5 gigawatts would give AMD access to infrastructure on a scale normally associated with the largest cloud-computing operators, although the companies did not disclose the conditions, timetable or financial terms governing any expansion.
Core Scientific did not provide an AMD-specific revenue estimate, and the announcement did not identify the data-center sites that would host the equipment. AMD will receive warrants to buy Core Scientific common shares, traded under the ticker CORZ, at market-based prices, with eligibility linked to specified commercial conditions.
CORZ shares rose about 3% in pre-market trading after the announcement.
A supply agreement focused on AI infrastructure
The partnership extends AMD’s effort to ensure customers can obtain not only accelerators and processors but also the power-intensive physical facilities needed to run them. AI systems built around advanced GPUs require far more electricity and cooling capacity than conventional enterprise servers, making prepared data-center campuses an increasingly valuable part of the computing supply chain.
AMD’s chief strategy officer for corporate development said in the announcement that power, land and infrastructure availability had become critical factors in bringing additional AI computing capacity online. Core Scientific Chief Executive Officer Adam Sullivan said the arrangement centers on the company’s ability to deliver high-density infrastructure at scale.
For Core Scientific, the deal adds a major semiconductor company to a customer pipeline that has been led by CoreWeave, an AI cloud provider. The miner-turned-data-center operator has been converting parts of its existing footprint for colocation, in which a customer places its own computing hardware in a third-party facility that supplies power, cooling, networking and operations.
The model differs from Bitcoin mining, where a company’s revenue is directly exposed to Bitcoin’s price, network difficulty, block rewards and energy costs. Long-term hosting contracts can provide more predictable contracted revenue, but they also demand extensive upfront spending on electrical equipment, cooling systems and building upgrades.
Core Scientific is funding a transition from mining
Core Scientific has accelerated its move toward AI hosting during 2026, using a network of facilities developed originally for large-scale Bitcoin mining. The company operates data centers in Alabama, Georgia, Kentucky, North Carolina, North Dakota, Oklahoma and Texas, and has been evaluating which remaining mining sites can be repurposed for high-density colocation.
In March, Core Scientific said it planned to monetize substantially all of its Bitcoin holdings during 2026 to help fund the buildout, with most of the anticipated sales recorded in the first quarter. That decision illustrates the capital demands of the transition: mining companies can have access to power-connected sites, but turning those sites into AI-ready facilities requires equipment and construction that differ sharply from a warehouse filled with mining machines.
AI customers generally need redundant power systems, advanced cooling and network designs capable of supporting clustered GPU deployments. The required upgrades can be particularly complex at former mining locations, where electrical capacity may already exist but the facility was not designed around the heat density and connectivity requirements of AI servers.
Core Scientific’s existing development pipeline gives it an advantage over companies starting from undeveloped land, yet the AMD agreement does not by itself guarantee that the full 2.5-gigawatt option will be used. Delivery will depend on commercial conditions and Core Scientific’s ability to build and operate the required infrastructure.
AI hosting offers opportunity but not a universal mining solution
Bernstein analysts recently pointed to Core Scientific’s CoreWeave agreements as contracts that could produce a 75% average return on assets over five years. The analysts cautioned that those returns should not be treated as a standard outcome for other Bitcoin miners seeking to enter AI hosting.
That distinction reflects the uneven quality of miners’ existing assets. A facility with low-cost power, transmission access, expandable land and proximity to fiber networks may be attractive for data-center customers. Other mining sites may lack the location, cooling capability or grid reliability needed for AI workloads, even if they once supported profitable mining operations.
AMD’s reservation therefore strengthens Core Scientific’s case that some former mining infrastructure can be adapted for a higher-value computing market. It also gives AMD a possible capacity channel beyond the cloud providers and data-center developers competing for the same constrained U.S. power resources.
The partnership’s financial impact will depend on construction progress, equipment deployment and whether AMD exercises its expansion rights. For now, the agreement gives Core Scientific another route to turn its power-heavy mining footprint into contracted AI capacity while giving AMD an option to secure large-scale U.S. infrastructure before it is absorbed by competing AI deployments.
Explore how AI and trading converge with our guide to AI copy trading and how it works.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
