The Investor
09 Nov 2026, 19:14
AMD EPYC Powers Google Cloud TPU 8tx as Confidential AI Infrastructure Expands
AMD (NASDAQ: AMD) announced that its 5th Gen EPYC processors are powering the host infrastructure for Google Cloud TPU 8tx, the first cloud TPU system to support hardware-enforced confidential computing.
The collaboration combines AMD EPYC CPUs with Google’s eighth-generation TPU architecture to support inference, fine-tuning and reinforcement learning workloads while protecting sensitive AI data and models during use.
AMD said EPYC processors handle workload orchestration, data preparation and high-throughput communication with TPU accelerators. The system uses AMD Secure Encrypted Virtualization technology to help keep prompts, model weights and intermediate data encrypted in CPU memory and during transfers between the host and accelerators.
The deployment expands AMD’s role as a host CPU provider for accelerated AI systems. The company said EPYC processors are now used as host CPUs across more than 10 leading hyperscalers and frontier AI labs, spanning systems built around AMD Instinct GPUs, NVIDIA GPUs and custom AI accelerator ASICs.
AMD and Google Cloud have already collaborated across multiple generations of EPYC-based virtual machines and confidential computing services. TPU 8tx extends that relationship into secure AI acceleration.
The announcement strengthens AMD’s position in AI infrastructure beyond GPUs, highlighting EPYC CPUs as a core component of increasingly specialized AI systems where security, memory bandwidth, I/O performance and accelerator coordination are critical.
AMD (NASDAQ: AMD) announced that its 5th Gen EPYC processors are powering the host infrastructure for Google Cloud TPU 8tx, the first cloud TPU system to support hardware-enforced confidential computing.
The collaboration combines AMD EPYC CPUs with Google’s eighth-generation TPU architecture to support inference, fine-tuning and reinforcement learning workloads while protecting sensitive AI data and models during use.
AMD said EPYC processors handle workload orchestration, data preparation and high-throughput communication with TPU accelerators. The system uses AMD Secure Encrypted Virtualization technology to help keep prompts, model weights and intermediate data encrypted in CPU memory and during transfers between the host and accelerators.
The deployment expands AMD’s role as a host CPU provider for accelerated AI systems. The company said EPYC processors are now used as host CPUs across more than 10 leading hyperscalers and frontier AI labs, spanning systems built around AMD Instinct GPUs, NVIDIA GPUs and custom AI accelerator ASICs.
AMD and Google Cloud have already collaborated across multiple generations of EPYC-based virtual machines and confidential computing services. TPU 8tx extends that relationship into secure AI acceleration.
The announcement strengthens AMD’s position in AI infrastructure beyond GPUs, highlighting EPYC CPUs as a core component of increasingly specialized AI systems where security, memory bandwidth, I/O performance and accelerator coordination are critical.
