If you have been following the AI arms race, you know that computing power is the new gold rush. AMD just made a massive move to secure its stake. That initial block is just the start — the arrangement can be expanded to as much as 2.5 gigawatts across five facilities. This AMD data center capacity deal is a clear signal that the company is serious about competing in the AI infrastructure space at scale.
This AMD Core Scientific partnership specifically covers the deployment of AMD Instinct graphics processing units, EPYC server processors, and ROCm software. For you, the practical takeaway is that this is not just about raw square footage — it is about high-density colocation designed for AI workloads. As data centers become the factories of the digital age, this kind of AI data center expansion means more competition, more choices, and potentially better pricing for cloud services down the line.
1. Massive Capacity: 529 MW to 2.5 GW Across Five U.S. Sites
This deal moves from general trends to a concrete plan. Starting in 2027, AMD secures an initial 529 megawatts of critical IT capacity across five Core Scientific facilities. These sites are located in Texas, Oklahoma, Alabama, and Georgia. Critical IT capacity refers to the power dedicated to running servers and cooling, which is the backbone of any AI-ready infrastructure. For you, this means more computing power available for AI services and applications down the road.
Beyond the initial commitment, AMD has the option to reserve an additional 1,925 megawatts through December 28, 2028. That brings the potential total to approximately 2.5 gigawatts. This expansion of AMD data center capacity is massive, allowing the company to scale its operations significantly. By securing this AI-ready infrastructure now, AMD is ensuring it has the physical resources to support growing demand for AI processing, which could lead to more robust and accessible cloud offerings for consumers and businesses alike.
2. $14 Billion Deal Structure: Leases, Warrants, and Revenue
This Amd data center capacity expansion is backed by a substantial financial framework. Core Scientific estimates that the initial 15-year agreements represent more than $14 billion in potential base contracted revenue. Under this structure, AMD directly leases 377 megawatts at facilities in Pecos and Hunt County, Texas, and Muskogee, Oklahoma. An unnamed neocloud has leased another 152 megawatts at sites in Auburn, Alabama, and Dalton, Georgia. These leases come with initial terms of 15 years and options for three five-year extensions, providing long-term stability for both parties.
In addition to the lease agreements, Core Scientific has issued AMD a warrant covering up to 30 million shares at an exercise price of $23.47 per share. Approximately 6.5 million of those shares vested when the initial agreements were signed. This data center lease agreement structure includes Core Scientific stock warrants that give AMD a stake in the company, aligning incentives over the long haul. The combination of base contracted revenue and equity positions makes this a significant commitment for both companies as they work to meet growing demand for AI processing capacity.
3. Technology Stack: AMD Instinct GPUs, EPYC CPUs, and Helios Architecture
Beyond the financial structure, the specific technology driving this AMD data center capacity expansion is worth a closer look. The partnership centers on deploying AMD Instinct GPUs, EPYC server processors, and the ROCm software platform. This combination is designed to handle demanding AI workloads, from training large models to running inference at scale. You might wonder how these components fit together. AMD’s Helios reference architecture provides the blueprint. It brings together Instinct GPUs, EPYC CPUs, Pensando networking, and ROCm software, all supported by liquid cooling and rack-level power distribution. This integrated approach aims to simplify deployment and improve efficiency for large-scale AI clusters.
For now, the companies have not disclosed the specific processor models, rack configurations, networking systems, power architecture, or cooling technology planned for the five locations. That means you won’t find exact specs like which Instinct GPU generation or EPYC CPU series will be used. However, the reliance on the Helios framework suggests a focus on a standardized, scalable design. This makes it easier to replicate across multiple sites, which is crucial for a project of this scope. The emphasis on liquid cooling and rack-level power distribution also hints at the high density and performance these facilities will need to deliver.
4. Operational Collaboration: Core Scientific Develops, AMD Co-Designs
Turning that kind of high-density infrastructure into reality requires a close partnership between the builder and the user. Under this data center development partnership, Core Scientific will develop and operate the sites, handling the heavy lifting of construction and day-to-day management. AMD, in turn, collaborates on the physical infrastructure design and system deployment, ensuring the facilities are perfectly tuned for its hardware from the ground up. This infrastructure co-design model is a practical way to guarantee performance without AMD taking on the full burden of facility ownership. Crucially, the agreement does not transfer ownership of the facilities to AMD, keeping the assets on Core Scientific’s books while AMD secures the massive Amd data center capacity it needs.
Also worth a read: Avatar Aang’s Messy Release Continues at Comic-Con.
The timeline for this ambitious project is equally measured. The first phase is expected to become available from 2027, giving both companies ample time to execute the complex buildout. Core Scientific’s track record provides strong confidence in the plan. The company’s high-density colocation business generated $136.7 million of its $164.2 million in revenue during the second quarter of 2026. This high-density colocation revenue demonstrates a proven ability to manage the extreme power and cooling demands that next-generation AI hardware requires. For you, it means a reliable path for AMD to bring its future silicon to life at an unprecedented scale.
5. Strategic Implications: AMD vs. Nvidia in AI Infrastructure
This deal does more than just secure power for AMD. It reshapes how you might view the AMD Nvidia AI competition over the next few years. By locking down this massive 2.5 GW of capacity, AMD ensures it has the physical space to deploy its next-generation AI accelerators at scale. That is a direct challenge to Nvidia’s current dominance in the AI infrastructure market. Without this kind of capacity, even the best chip design is just a paperweight. Now, AMD can actually bring its hardware to life in data centers ready for extreme power loads.
There is another strategic layer here. The unnamed neocloud data center leasing 152 MW at Core Scientific sites in Alabama and Georgia suggests a multi-tenant model. This neocloud could host AMD’s chips for third-party customers, making the Amd data center capacity available to businesses that cannot build their own facilities. Core Scientific’s regulatory filing also gives AMD protections over equipment at those sites, showing a coordinated plan. For you, this means AMD is not just building its own cloud — it is creating a flexible infrastructure network that can compete with Nvidia’s established ecosystem, all without needing to own every brick and cable.
Frequently Asked Questions
How will AMD’s data center capacity from Core Scientific become available to customers?
AMD plans to roll out the capacity in phases, with the first phase expected to come online in 2027. You can expect a gradual deployment that allows Core Scientific to prepare its high-density colocation infrastructure. This step-by-step approach ensures reliability and performance for your workloads.
How does this agreement affect AMD’s competitive position against Nvidia in AI infrastructure?
By securing a large block of dedicated data center capacity, AMD strengthens its ability to offer scalable AI solutions. This move directly challenges Nvidia’s dominance in the AI hardware market. For you, it means more choices and potentially better pricing as competition intensifies.
What benefits can AMD customers expect from this increased data center capacity?
With more amd data center capacity available, you gain access to optimized environments for running AMD-powered AI and HPC workloads. The partnership with Core Scientific means you can leverage their existing colocation expertise for faster deployment. Expect improved performance and reliability for your most demanding applications.
