Core Scientific’s Muskogee Bet: Can Crypto-Era Infrastructure Fuel AI?

extract

Core Scientific’s Muskogee Bet: Can Crypto-Era Infrastructure Fuel AI? The operator’s data center buildout in Oklahoma, paired with a new 1.5 GW plan in Texas, positions the company to deliver roughly 3 GW of AI and high-density colocation capacity. Core Scientific plans to scale its Muskogee, Okla., campus to roughly 1.5 GW of gross power capacity, the latest step in its pivot from bitcoin mining to large-scale AI data center infrastructure. The move signals how crypto-era operators are repurposing energy-intensive assets and grid relationships to meet the growing demand for AI-ready data center capacity. The company said the site could ultimately support about 1 GW of leasable power through a combination of acquisitions, new construction, grid expansion studies, and behind-the-meter generation strategies. To accelerate that plan, Core Scientific agreed to acquire Polaris DS LLC, which controls 440 MW of contracted power through an agreement with Oklahoma Gas & Electric. The transaction is expected to close in the third quarter of 2026, pending regulatory approvals. Core Scientific also said it has secured roughly 250 acres to support future development and is already constructing a second unleased 82.5 MW building at the Muskogee campus, with initial delivery expected in the fourth quarter of 2027. “Our multi-tiered approach in Muskogee demonstrates how we can combine acquisitions, development expertise, and innovative power solutions to unlock gigawatt-scale capacity,” CEO Adam Sullivan said. Crypto Infrastructure Pivots to AI The Oklahoma project follows Core Scientific’s April 2026 announcement to scale its Pecos, Texas, site to roughly 1.5 GW of gross power capacity while repurposing existing bitcoin mining infrastructure for AI workloads. Those moves underscore intensifying competition across the emerging neocloud market, where AI infrastructure providers are racing to secure power, land, and transmission capacity before utility bottlenecks slow deployment. Former bitcoin miners, including Core Scientific, IREN, Cipher, and Hut 8, increasingly pitch themselves as faster-moving alternatives to traditional hyperscaler buildouts by repurposing existing energy-heavy infrastructure for GPU clusters and AI colocation. Core Scientific is executing “a masterclass in asset arbitrage,” said Dave McCarthy, research vice president, cloud and edge infrastructure services, at IDC, as it shifts infrastructure built for bitcoin’s volatile economics toward higher-margin, long-term AI demand. “The most valuable legacy of the crypto mining era isn’t the coins themselves, but the hard-to-replicate power permits and grid relationships that now form the backbone of the AI era,” McCarthy said. Core Scientific said its existing 70 MW leased facility in Muskogee, designed for Nvidia GB300 systems, is still in final testing and commissioning and on track for delivery in the second quarter of 2026. The company is also pursuing additional grid-connected capacity through ongoing load studies while developing behind-the-meter generation plans intended to supplement utility-delivered power. That strategy increasingly mirrors broader trends in AI infrastructure development as hyperscalers and colocation providers confront tightening utility markets and lengthy interconnection queues. Rather than waiting solely on grid upgrades, operators increasingly pair campuses with on-site or adjacent generation assets to accelerate deployment schedules. GW Campuses Redefine AI Infrastructure The Muskogee expansion highlights how AI infrastructure developers now frame projects in gigawatt-scale terms, once associated primarily with hyperscale cloud regions or heavy industrial facilities. Together, the Pecos and Muskogee announcements represent roughly 3 GW of gross power capacity tied to campuses that Core Scientific aims to position for AI and high-density colocation workloads. Core Scientific, which emerged from bankruptcy in 2024, has accelerated efforts to transition away from dependence on bitcoin mining revenue toward long-term AI infrastructure contracts and high-density colocation services. The company has increasingly emphasized converting legacy mining assets into AI-ready campuses capable of supporting dense GPU deployments.