Rackspace and AMD have officially finalized their partnership to build a significant AI infrastructure with a capacity of 30 megawatts (MW) across Rackspace's global data centers.
This agreement turns their earlier memorandum of understanding (MOU) into a concrete action plan, with deployments scheduled from late 2026 through 2028. You might wonder how large 30 MW is. While it's smaller than the massive facilities used by tech giants like Meta, it represents a very substantial capacity for enterprise use. To put it in perspective, this could power roughly 20,000 to 37,000 high-end AMD Instinct AI GPUs. This isn't just about scale; it's about providing a specialized service.
The core idea behind this partnership is to create a 'governed AI' cloud. First, think about industries like healthcare or finance. They operate under strict regulations like HIPAA and need to ensure their data and AI models are secure, private, and compliant. They can't simply use a generic public cloud. Rackspace aims to be the specialized provider for these companies, offering a managed environment where accountability is clear 'from silicon to outcomes'.
Second, this is where AMD comes in. Instead of relying on closed, proprietary systems, this new cloud will be built on AMD's open-standard hardware. This includes its powerful Instinct GPUs and EPYC CPUs, assembled in a rack-scale design called 'Helios'. This open approach, supported by standards like the Ultra Ethernet Consortium, gives customers more flexibility and avoids vendor lock-in, which is a common concern in the tech world.
Finally, the timing and credibility are key. AMD has already proven its ability to deliver AI chips at a massive scale through a major deal with Meta. This demonstrated supply chain strength gives Rackspace the confidence to commit to a multi-year plan built on AMD's technology roadmap. This deal positions Rackspace as a key enabler for enterprises that need powerful AI capabilities but cannot compromise on security and governance, creating a distinct alternative in the competitive cloud market.
- Governed AI: An AI system operated within a strict framework of rules, security protocols, and compliance standards, often required in regulated industries like healthcare and finance to ensure accountability and data privacy.
- MOU (Memorandum of Understanding): A non-binding agreement between two or more parties that outlines the terms and details of a mutual understanding or agreement, often serving as a preliminary step before a formal, definitive contract.
- Inference: The process of using a trained AI model to make predictions or decisions based on new, live data. For example, an AI model analyzing a medical scan to detect anomalies is performing inference.