AMD EPYC CPUs host Google Cloud TPU 8tx for confidential AI

Google Cloud and AMD have broadened their collaboration around Google Cloud TPU 8tx, a cloud Tensor Processing Unit system whose host infrastructure is powered by 5th Gen AMD EPYC processors. The system is intended to help enterprises run their most sensitive AI workloads in the cloud without weakening data protection.

Accelerated AI systems depend on host processors to prepare data, coordinate workloads and manage communication with accelerators. AMD EPYC CPUs are the host CPU of choice for more than 10 leading hyperscalers and frontier AI labs, spanning GPUs and custom AI accelerator ASICs. Protecting data while it is in use is the requirement this collaboration addresses.

In TPU 8tx systems, 5th Gen EPYC processors orchestrate workloads, prepare data and manage high-throughput communication with TPU accelerators. The design uses AMD Secure Encrypted Virtualization (SEV) capabilities to extend confidential computing protection to connected accelerators. Before an accelerator receives direct access to protected memory, its identity and security configuration are verified. Customer prompts, model weights and intermediate data remain encrypted in memory and in transit.

AMD and Google Cloud have worked together across several generations of EPYC processors, supporting general-purpose, compute-intensive, storage-intensive and accelerated workloads. The companies' work on EPYC has continued through the current Google Cloud Compute Engine C4D, N4D, H4D, Z4D and G4 virtual machines. The companies have also collaborated for years on security technology, including deep reviews of AMD security processor firmware by Google security experts.

Dan McNamara, senior vice president and general manager, Compute and Enterprise AI, AMD, said, “The most capable AI systems need a CPU foundation that can match the demands of the accelerators they host. Across generations of EPYC, we have built leadership capabilities in the performance, memory bandwidth and I/O that keep accelerated systems productive, backed by years of engineering collaboration with the world's leading cloud providers.”

Mark Lohmeyer, vice president and general manager for AI and Computing Infrastructure, Google Cloud, said, “By pairing 5th Gen AMD EPYC processors with our 8th-generation TPU architecture inside the AI Hypercomputer, we are bringing hardware-enforced confidential computing to cloud TPUs for the first time.” The expanded collaboration is described as helping enterprises run modern reasoning and agentic AI workloads in production without compromising on data protection.