Rackspace and AMD unveil enterprise AI cloud tie-up

Rackspace and AMD have signed a memorandum of understanding for a multiyear partnership on enterprise AI infrastructure, centred on adding AMD GPU capacity to Rackspace's managed AI offering.

The agreement sets out a framework for an Enterprise AI Cloud aimed at regulated organisations and workloads that require tighter controls over security, governance and accountability. The memorandum of understanding is not a binding commercial agreement.

Rackspace is presenting the AMD tie-up as part of a broader shift in its AI strategy. It wants to provide a full stack of services for customers running AI projects, covering infrastructure, inference and operational management, rather than simply offering access to raw computing resources.

The move reflects a broader debate in the enterprise market about how AI systems transition from trial projects into everyday operations. Many businesses still rent graphics processing units by the hour from cloud providers and manage integration, security and oversight themselves, a model that can leave them carrying costs and operational risk internally.

Under the proposed arrangement, Rackspace would integrate AMD Instinct GPUs and EPYC CPUs into its managed environment. It would assemble and run the technology stack for customers that want private or hybrid AI systems, including organisations with sovereignty or compliance requirements.

The companies outlined four areas they expect to cover under the partnership: a managed Enterprise AI Cloud, an Enterprise Inference Engine designed to retain business context and session history across queries, an inference service built on dedicated AMD GPUs, and bare-metal AMD Instinct systems for customers that want direct access to hardware.

The aim is to offer a single operator responsible for each layer of the stack. In practice, Rackspace would take responsibility for system availability, scaling and performance in the managed services it plans to offer.

The announcement also highlights the growing importance of private cloud models for AI workloads. Companies in regulated sectors such as finance, healthcare and government have been looking for ways to use AI systems while retaining tighter control over where data is stored, how models are run and who is responsible when systems fail or produce inaccurate results.

For AMD, which has been expanding its position in AI computing, the Rackspace relationship provides a route into managed enterprise environments. The chipmaker has been seeking to broaden adoption of its Instinct accelerators and EPYC processors as businesses look for alternatives in the AI hardware market.

Rackspace said it has spent the past two years building a group of AI partners that includes Palantir, Uniphore, Rubrik and VMware. It is positioning that network as a way for customers to combine infrastructure, software and operational support as they move AI systems into production.

Gajen Kandiah, Chief Executive of Rackspace, described the move as a response to customer concerns about accountability in operational AI systems. "As enterprises move AI out of the lab and into production environments, they're asking who they can trust to run it there," he said.

"Governing AI infrastructure in regulated environments with defined accountability is not something you bolt on after the fact. It must be built in from the start. Rackspace and AMD are building exactly that and, in doing so, establishing a new category of enterprise AI infrastructure that the market has been asking for."

AMD also framed the deal around the need for a stronger computing base as AI workloads expand beyond experiments. "Enterprise AI is quickly moving from experimentation to production, and that requires a compute foundation engineered for performance and efficiency at scale," said Dan McNamara, Senior Vice President and General Manager, Compute & Enterprise AI at AMD.

"Our collaboration with Rackspace delivers AMD AI compute into managed, private and governed environments so enterprises can deploy AI with the performance and flexibility their workloads demand."