Rackspace Technology and AMD say they have signed a definitive agreement to roll out an initial AI compute footprint of 30 MW across Rackspace’s global data centers. The announcement was made on June 16, 2026, with the build planned in phases.

According to the companies, the deployment window runs from late 2026 through 2028. While 30 MW may sound technical, it matters because the power draw is a practical measure of how much real computing capacity a data center can support for AI workloads. In plain terms, it points to a concrete ramp-up rather than a vague plan.

The deal centers on an AMD-based hardware stack. Rackspace and AMD said the compute will combine AMD Instinct GPUs and AMD EPYC CPUs, integrated into Rackspace’s enterprise AI cloud architecture.

Why the rollout matters

AI services tend to hit infrastructure limits quickly, especially when customers want higher performance without delays. A phased deployment helps operators bring capacity online over time, instead of trying to install everything at once. That can also make it easier to align supply and support as demand shifts.

For Rackspace customers, the goal is capacity that is specifically designed for enterprise AI needs using AMD’s GPU and CPU pairing. For the industry, it’s another example of how data center operators are turning energy and hardware planning into an “AI infrastructure” roadmap.

Rackspace and AMD did not provide further details in the announcement about operational specifics beyond the phased 30 MW footprint and the hardware components included in the architecture.

Source: Rackspace Technology