Core Scientific’s AMD Pact Reaches 2.5GW — Finding That Power Is the Next Question
Core Scientific and AMD unveiled a major AI infrastructure partnership on July 28, and the scale of it is hard to overstate.
The agreement hands AMD more than 500 megawatts of U.S.-based compute capacity right away — with a contractual path that stretches all the way to 2.5 gigawatts.
Also Read: GPT-5.6 Hits Its Highest ARC-AGI-3 Score — the Gains Came from Two Toggles
That’s a step change, not an incremental expansion.
The deal was disclosed through an SEC 8-K filing on Business Wire. A Stock Titan report confirmed it as well, citing Core Scientific’s Q2 2026 colocation revenue of $136.7 million.
Core Scientific And AMD Partnership Reshapes U.S.
AI Compute Access
The agreement was filed with the SEC as an 8-K on Business Wire, and a Stock Titan report confirmed Core Scientific’s (CORZ) Q2 2026 colocation revenue of $136.7 million alongside the announcement. The deal makes Core Scientific and AMD (AMD) a central pairing in U.S.
AI infrastructure, with Core Scientific serving as the primary infrastructure backbone for AMD’s growing ecosystem of AI customers. AMD, the chipmaker competing directly with Nvidia in the AI accelerator market, gains guaranteed access to power-dense, AI-ready colocation space without building its own facilities.
Colocation, in this context, means Core Scientific owns and operates the physical data center buildings, power connections, and cooling systems, while AMD’s customers deploy their own servers inside.
The Core Scientific and AMD arrangement lets AMD scale its compute footprint rapidly without the multi-year lead times that greenfield data center construction requires.
The 500 megawatts secured at signing is already significant. For scale, a single modern AI training cluster running thousands of next-generation GPUs typically consumes between 50 and 200 megawatts.
The 2.5-gigawatt ceiling, if fully built out, would represent one of the largest single-vendor AI compute reservations in the United States.
From Bitcoin Miner To AI Landlord
Core Scientific began as a cryptocurrency mining company, operating large fleets of application-specific integrated circuit machines to earn Bitcoin (BTC) block rewards. The business model shared infrastructure DNA with AI data centers: both require enormous power supplies, high-density rack space, and precision cooling.
That overlap became a strategic asset as AI demand exploded and Bitcoin (BTC) mining margins compressed.
Core Scientific began converting mining halls into high-power-density AI colocation facilities, signing long-term contracts with hyperscalers and AI firms that needed compute fast and had little interest in building their own campuses.
The company filed for Chapter 11 bankruptcy protection in December 2022 as cryptocurrency prices collapsed and energy costs spiked, but emerged from restructuring in January 2024 with its assets intact and its pivot to AI colocation accelerating. The $136.7 million in Q2 colocation revenue reported alongside the Core Scientific and AMD announcement marks the clearest financial proof point yet that the transition is generating durable income.
Why The 2.5GW Number Matters
The headline figure of 2.5 gigawatts deserves scrutiny.
Announced capacity ceilings in data center deals are frequently aspirational, contingent on power procurement approvals, grid interconnection timelines, and construction financing. The immediate 500-megawatt commitment is the operative number for near-term planning purposes.
That said, 2.5 gigawatts is a credible long-term target given the land and power pipeline Core Scientific has assembled across Texas and other power-rich states.
The company has historically located campuses near low-cost, abundant electricity, a legacy of its mining roots that now functions as a competitive moat against data center operators entering from the real estate or telecom sectors.
For AMD, the Core Scientific and AMD partnership is also a customer acquisition tool. Offering prospective AI chip buyers a guaranteed path to U.S. compute capacity reduces a key friction point in enterprise sales cycles, where the inability to find adequate data center space has become a bottleneck as acute as chip supply itself.
The announcement lands as the broader AI infrastructure market faces fresh scrutiny.
Google’s parent Alphabet disclosed this week that AI capital expenditure reached $44.9 billion in a single quarter, pushing the company into negative free cash flow for the first time. That backdrop makes Core Scientific and AMD’s asset-heavy, long-term-contract model look comparatively stable: the revenue is locked in, and the power is already paid for.
Read Next: AI Data Center Campus In Texas Poised to Dominate AI Infrastructure
