Core Scientific Doubles AI Data Center Capacity to 1.1 GW in $14 Billion AMD Deal

Core Scientific has significantly expanded its role in the artificial intelligence infrastructure race by entering into a 15-year agreement with AMD. The deal, announced alongside the company’s second-quarter results, doubles its leased AI data center capacity to approximately 1.1 gigawatts (GW) and represents a major strategic pivot toward high-performance computing. The agreement is valued at more than $14 billion in potential base contracted revenue over its term.

The partnership covers 530 megawatts (MW) of capacity across five campuses and grants AMD exclusive rights to reserve up to an additional 2 GW of future capacity. Construction is already underway at the company’s Pecos, Texas, facility, with initial deployments scheduled to begin in the first half of 2027. Additional development is progressing in sites located in Texas, Oklahoma, Alabama, and Georgia, with full deployment expected to continue through 2028. Under the terms of the deal, Core Scientific will host infrastructure built around AMD Instinct GPUs, EPYC processors, and the ROCm software platform, targeting cloud providers, AI model developers, and enterprise customers.

Mathew Hein, AMD’s chief strategy officer for corporate development, said in a statement that Core Scientific’s extensive portfolio of AI-ready data centers is key to helping customers deploy AMD’s AI solutions at scale. Matt Kimball, vice president and principal analyst at Moor Insights & Strategy, noted that the agreement is less about AMD becoming a landlord and more about removing a critical barrier to AI accelerator adoption. “It’s not about selling accelerators at this point, it’s about selling a deployment path,” Kimball said. He explained that AMD appears to be leveraging its balance sheet to mitigate deployment risk by ensuring the availability of high-density power, liquid cooling, and specialized facilities, which have become as crucial as the chips themselves.

The headline figures, however, require careful interpretation. The $14 billion represents base contracted revenue spread over 15 years, not a single upfront booking. Similarly, the additional 2 GW of capacity represents reservation rights rather than firm customer commitments. Despite this, the deal solidifies Core Scientific’s position in the AI infrastructure market. Its total contracted AI portfolio now stands at 1.1 GW of leased power capacity, representing over $24 billion in potential contracted revenue. As of mid-July, the company was already billing customers for 437 MW, equivalent to an annualized colocation revenue run rate of approximately $635 million.

Core Scientific executives emphasized that execution will now be the primary differentiator. “Value is not announced. It is delivered,” CEO Adam Sullivan said during the company’s earnings call. Sullivan noted that Core Scientific now has two AI customers, each committing to more than 500 MW across five campuses, which provides greater diversification and validates its strategy of developing campuses before securing long-term tenants. Chief Operating Officer Matt Brown also provided one of the clearest cost benchmarks in the industry, stating that fully delivered AI data center capacity is expected to cost approximately $11 million to $12 million per megawatt. At that rate, the initial AMD deployment alone represents roughly $6 billion in infrastructure investment.

The expansion underscores Core Scientific’s rapid transformation from a bitcoin mining company to a leading AI infrastructure provider. Colocation revenue surged to $136.7 million in the second quarter from $10.6 million a year earlier, while its digital asset self-mining revenue fell to $21.5 million as the company continues shifting power from bitcoin mining to AI workloads. To support future growth, Sullivan said the company has identified more than 2 GW of additional development opportunities that could come online between late 2028 and 2030. For the broader industry, this agreement signals that competitive advantage is no longer solely defined by AI accelerators; it is increasingly about a vendor’s ability to provide a full deployment path that includes power, liquid cooling, and energized capacity years before it is needed.