AMD and Rackspace sign agreement for planned 30 MW AI compute rollout

AMD and Rackspace Technology have signed a definitive agreement to deploy an initial 30 MW of AMD-based compute across Rackspace’s global data centres. The phased rollout is expected to begin in late 2026 and continue through 2028.

The agreement puts a commercial framework behind the memorandum of understanding announced on 7 May. It makes AMD the technology partner for the chip layer of Rackspace’s governed AI stack, which is being developed for regulated and sovereign environments.

The rollout centres on managed AI infrastructure

Rackspace plans to use AMD Instinct GPUs, including MI355X, MI350P and future successor products, alongside AMD EPYC CPUs inside its Enterprise AI Cloud architecture. The company intends to use that mix to direct workloads to the appropriate type of compute, depending on whether they need graphics processors for AI acceleration or central processors for general-purpose tasks.

At full deployment, AMD and Rackspace expect the 30 MW footprint to support regulated enterprise workloads. The release points to healthcare providers as one area of early interest, particularly for clinical AI and large-scale inference, where trained AI systems process live requests.

Gajen Kandiah, CEO of Rackspace Technology, framed the agreement around the need for clearer responsibility in enterprise AI infrastructure.

“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,” Kandiah said. “This collaboration combines the right compute with the right operating model and delivers something the market hasn’t offered before: a governed AI stack with one accountable partner from silicon to outcomes.”

Rackspace is building a wider AI services stack

The agreement is expected to support four capabilities first announced with the MOU: Enterprise AI Cloud, Enterprise Inference Engine, Inference as a Service and Bare Metal AMD Instinct. Together, these services are intended to cover infrastructure from bare metal AMD compute through to operated inference.

That gives Rackspace a way to pitch customers on managed AI infrastructure, rather than raw compute capacity alone. The company is presenting the stack as an option for enterprises that need AI systems operated with governance, sovereignty and accountability built into the service model.

AMD’s role spans both accelerated and general-purpose compute. Dan McNamara, senior vice president and general manager for Compute and Enterprise AI at AMD, described the arrangement as a way to match different enterprise AI workloads with the appropriate mix of AMD GPUs and CPUs.

Both companies expect to dedicate sales and marketing resources to identify enterprise customers for AMD compute-powered infrastructure. They will also commit personnel to jointly develop and pursue customer opportunities across regulated industries.

Deployment terms still need further approvals

The agreement establishes a commercial framework, but the press release includes several limits around execution. Individual deployment authorisations remain subject to separate execution, and some commercial terms, including pricing and financial parameters, remain subject to further agreement.

The companies also noted that any third-party financing required for planned deployments will depend on availability on terms acceptable to Rackspace.