AMD and Rackspace Technology have signed a definitive agreement to deploy 30 megawatts of dedicated AMD-based compute across Rackspace’s global data centers. The multi-year rollout, spanning late 2026 through 2028, finalizes a Memorandum of Understanding announced on May 7, 2026.
The partnership positions AMD as the primary silicon partner for Rackspace’s “governed AI stack”. This is a managed infrastructure model designed specifically for heavily regulated sectors and sovereign environments where data security and strict compliance prevent the use of standard public clouds.
Under the Hood
The infrastructure will combine general-purpose and accelerated hardware, integrating AMD EPYC CPUs alongside AMD Instinct GPUs, including the MI355X and MI350P architectures. Rackspace plans to leverage this mixed environment to route enterprise workloads dynamically, optimizing for either raw inference power or general processing depending on the specific application.
By deploying 30 MW of power, the companies are establishing a sizable footprint aimed squarely at complex enterprise workloads. Rackspace noted that healthcare providers have already signaled early demand for the infrastructure to run clinical AI tools and large-scale data inference.
A Shift to Accountable AI Infrastructure
The strategy behind the deal shifts away from the industry standard of renting unmanaged “bare metal” hardware. Instead, Rackspace and AMD are pitching a fully managed enterprise cloud. The companies argue that as businesses transition from experimental AI chatbots to autonomous “agentic workflows” integrated into core operations, they require a single accountable vendor rather than a fragmented ecosystem of hardware and software providers.
“Enterprises in regulated industries need AI infrastructure that is governed from the ground up, with one operator accountable for business outcomes, not a collection of vendors each owning a piece,” said Rackspace CEO Gajen Kandiah.
The agreement operationalizes four core cloud capabilities initially teased in May:
- Enterprise AI Cloud: The overarching governed architecture.
- Enterprise Inference Engine & Inference as a Service: Managed pipelines to run trained models efficiently.
- Bare Metal AMD Instinct: Direct hardware access for custom enterprise configurations.
To fill the capacity, both companies have committed dedicated sales, marketing, and engineering personnel to jointly source and secure enterprise clients across the healthcare, financial, and public sectors.
A 30-megawatt (MW) capacity power source can power roughly 15,000 to 30,000 homes, depending heavily on the type of energy generation, geographic location, and average household consumption.
