The Pulse
AMD says its EPYC CPUs power Google Cloud’s TPU 8tx hosts
AMD’s fifth-generation EPYC processors provide the host infrastructure for Google Cloud’s TPU 8tx. The system combines TPU acceleration with hardware-enforced confidential computing, the companies say.
AI.info Team · Written with AI · fact-checked ·
EPYC processors provide the host layer
AMD’s fifth-generation EPYC server processors provide the host infrastructure for Google Cloud’s TPU 8tx, which AMD calls the first cloud TPU system with hardware-enforced confidential computing. AMD’s Nov. 9 announcement describes an expanded collaboration that pairs AMD’s host CPUs with Google Cloud’s TPU acceleration.
The combination is intended to support inference, fine-tuning and reinforcement-learning workloads, including sensitive AI work in the cloud, AMD says. AMD describes its processors as supplying CPU performance, Infinity Guard security features and compatibility with x86 software.
In the TPU 8tx system, EPYC processors handle workload orchestration, data preparation and high-throughput communication with the TPU accelerators. AMD says the processors use AMD Secure Encrypted Virtualization, or SEV, to extend confidential-computing protection beyond the CPU to connected accelerators.
How AMD describes the protections
An accelerator’s identity and security configuration are checked before it receives direct access to protected memory. Customer prompts, model weights and intermediate data remain encrypted in CPU memory and while moving between the host and accelerators, with hardware isolation during accelerator execution.
Those protections are intended to help enterprises use sensitive information in reasoning and agentic AI applications, AMD says. AMD’s announcement also says EPYC-based SEV technologies support workloads in Google Cloud services including Confidential VMs, Confidential GKE, Confidential Space, Confidential Managed Service for Apache Spark and Confidential Dataflow.
Google Cloud vice president and general manager Mark Lohmeyer said pairing EPYC processors with the company’s eighth-generation TPU architecture brings hardware-enforced confidential computing to cloud TPUs for the first time. He said the system provides silicon-level data protection alongside the low-latency performance needed for modern reasoning models.
EPYC’s wider AI role
AMD says EPYC CPUs serve as host processors across systems using GPUs and custom AI accelerator ASICs, and are used by more than 10 leading hyperscalers and frontier AI labs. The company lists systems using its Instinct GPUs, Nvidia GPUs and custom silicon among the architectures supported by EPYC processors.
In those systems, host CPUs prepare data, schedule work and connect AI workloads with surrounding applications and services. AMD presents EPYC as a CPU platform that can be used across different accelerator choices, supported by the x86 software ecosystem.
A continuing collaboration
AMD says its work with Google Cloud has spanned multiple EPYC generations and workloads including general-purpose, compute-intensive, storage-intensive and accelerated computing. AMD’s current Google Cloud Compute Engine offerings named by AMD include C4D, N4D, H4D, Z4D and G4 virtual machines.
The companies have worked together on security technology, including reviews of AMD security processor firmware by Google security experts, AMD also says. AMD says TPU 8tx extends that collaboration to confidential AI acceleration.