July 28, 2026, (Inside AI) — Core Scientific has signed a deal with AMD to provide up to 2.5 gigawatts of data center capacity, a move that directly supports customer deployments of AMD's AI systems. The agreement, announced on Tuesday, marks a significant pivot for Core Scientific, once known primarily as a bitcoin mining operation, into the high-stakes AI infrastructure market.
This is not a tentative pilot. The scale of 2.5 gigawatts is enormous. For context, the total power capacity of all data centers in Northern Virginia, the world's largest data center market, was around 2.1 gigawatts in 2023. This single deal effectively matches an entire region's infrastructure. It underscores the sheer energy appetite of modern AI workloads and the desperate scramble for ready power.
The deal structure is notable. Core Scientific is not simply leasing space; it is securing capacity specifically for AMD's customers. This suggests a strategic alignment where Core Scientific becomes a preferred infrastructure partner for the AMD ecosystem, potentially offering optimized facilities for AMD Instinct accelerators. It is a direct challenge to Nvidia's dominance, where similar deals have been struck with partners like CoreWeave and Lambda Labs.
Core Scientific's transformation has been dramatic. The company emerged from Chapter 11 bankruptcy in January 2024, shedding its debt-laden bitcoin mining past. By repurposing its existing power infrastructure and data center expertise, it found a lifeline in AI. This deal validates that strategy, proving that the physical assets of crypto mining have a second life in the AI boom.
For AMD, the agreement is a critical piece of its AI strategy. The chipmaker has been aggressively expanding its software ecosystem with ROCm and courting cloud providers, but it needs guaranteed, large-scale deployment venues to compete with Nvidia's entrenched DGX cloud partnerships. Securing 2.5 gigawatts of capacity provides a credible answer to the question of where customers can actually run AMD-powered AI at scale.
The deal also highlights a growing trend: the convergence of energy infrastructure and AI compute. Data center power demands are projected to double by 2030, driven by AI. This has sparked a land grab for sites with access to reliable, often stranded, power. Core Scientific's existing relationships with energy providers and its portfolio of sites with pre-approved high-power connections give it a distinct advantage over new entrants.
However, the announcement leaves key details unanswered. The timeline for bringing this capacity online is unclear. Building or retrofitting data centers for high-density AI workloads, with advanced liquid cooling and specialized networking, is a multi-year endeavor. The specific financial terms, including revenue sharing or take-or-pay commitments, were not disclosed. Investors will be keen to understand the capital expenditure required and the expected return profile.
Industry analysts have noted the potential for conflict with Core Scientific's remaining bitcoin mining operations. The company still operates significant mining capacity, but AI hosting typically offers higher, more stable margins. This deal signals a clear prioritization of AI, which could accelerate the decline of its mining segment. A recent arXiv paper on data center power trends highlights the economic pressures driving such shifts.
The partnership also raises questions about chip allocation. AMD's supply of MI300X and future accelerators is finite. Committing such a large infrastructure block suggests AMD is confident in its supply chain and customer pipeline. It may also indicate that hyperscalers are seeking alternatives to Nvidia for specific inference workloads, where AMD's price-performance ratio is competitive, as detailed in official AMD documentation.
Core Scientific's stock has been volatile since its relisting, but this announcement provides a concrete, long-term revenue anchor. The company is effectively transitioning from a commodity producer to a critical infrastructure provider. The success of this deal will depend on execution: delivering reliable, high-uptime facilities that meet the stringent requirements of enterprise AI customers.
In the broader context, this is more than a single contract. It is a blueprint for how legacy energy-intensive industries can reinvent themselves for the AI era. Other former mining firms, like Hut 8 and Iris Energy, are pursuing similar paths, but none have yet secured a deal of this magnitude. The Core Scientific-AMD partnership sets a new benchmark, forcing competitors to scale up or risk irrelevance.
