Trending Topics
Trending
AMD EPYC processors power Google Cloud’s confidential AI infrastructure
AMD has announced its 5th-gen EPYC processors are powering the host infrastructure for Google Cloud Tensor Processing Unit (TPU) 8tx, a cloud-based AI accelerator system that uses hardware-enforced confidential computing to protect sensitive data and models.
According to AMD, the deployment combines AMD EPYC CPUs with Google’s 8th-gen TPUs, allowing customers to run AI inference, fine-tuning and reinforcement learning workloads while protecting data during processing.
The TPU 8tx system uses AMD’s Secure Encrypted Virtualisation (SEV) technology to extend confidential computing protections beyond the host CPU to connected accelerators. This includes encrypting customer prompts, model weights and intermediate data in CPU memory and while being transferred between the host and accelerators.
Executive leadership perspectives
“The most capable AI systems need a CPU foundation that can match the demands of the accelerators they host,” AMD computer and enterprise AI senior VP and general manager Dan McNamara said.
“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.
“Our work with Google Cloud brings that strength to TPU 8tx, helping customers scale AI with the performance and protection their most important workloads demand.”
AMD outlined how the EPYC processors handle workload orchestration, data preparation and communication with the TPU accelerators. They also provide the underlying processing and memory resources required to keep accelerated workloads running.
The system also verifies an accelerator’s identity and security configuration before granting it access to protected memory, with hardware isolation during accelerator execution.
“When teams move AI agents into production, they need confidence that their data and intellectual property stay private, even from the underlying infrastructure,” Google Cloud AI and computing infrastructure vice president and general manager Mark Lohmeyer 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.”
Expanding cloud partnership
The Google Cloud deployment builds on an existing relationship between the companies. AMD EPYC processors already support several Google Cloud Compute Engine virtual machine families, including C4D, N4D, H4D, Z4D and G4. The processors also underpin Google’s confidential computing services, including Confidential VMs, Confidential GKE and Confidential Space.
More from our Hardware section
About The Author
