AMD signs definitive deal with Rackspace for 30 MW of AI compute across global data centers, spanning 2026-2028

AMD and Rackspace Technology have entered into a definitive agreement for a large-scale deployment of AI compute capacity, a move that indicates continued momentum in the market for data-center infrastructure used to train and run machine-learning workloads.

Under the terms described in the announcement, Rackspace will deploy 30 MW of AMD-powered AI compute across its global data centers. The rollout is expected to start in late 2026 and continue through 2028, according to the report, and it follows an earlier memorandum of understanding that established the partners’ intent to work together on the project.

For AMD, the deal points to how its data-center-focused hardware can be paired with hosting and infrastructure providers that want to expand customer demand for AI services without building everything from scratch. For Rackspace, it is a concrete capacity commitment tied to a specific compute supplier and a multi-year build-out window.

The size of the capacity, 30 megawatts, matters because it offers a scale benchmark for AI infrastructure spending. While megawatts alone do not translate directly into revenue, large power allotments typically reflect plans for significant server and accelerator deployments, along with the electrical and cooling systems required to support sustained AI workloads.

The agreement is also notable because it is described as “definitive,” not merely exploratory. In deals like this, definitive terms usually indicate that key commercial and implementation details have been aligned, even if the public disclosures still do not provide fine-grained information such as the exact mix of products, the number of systems, or the expected ramp profile by site.

Sector-wise, the news fits into a broader pattern in which semiconductor vendors seek anchor deployments with data center operators and service providers. Those operators, in turn, pursue long-cycle capacity additions as cloud, enterprise, and government customers compete for access to AI compute for both inference (serving AI models to end users) and training (building models using large datasets).

Still, the report does not provide specifics that investors and customers typically want for impact assessment. It does not disclose financial terms, pricing, expected revenue contribution, the precise AMD technology stack to be used, or whether the compute capacity is allocated to specific Rackspace customers or sold as a capacity offering. Until additional company filings or detailed press releases emerge, the commercial magnitude and near-term accounting implications are uncertain.