AMD shares jump 13% as Google Cloud drafts EPYC chips into TPU 8tx AI system

AMD shares climbed 13.26% on Monday, closing around $508.58, after the company said its fifth-generation EPYC processors will serve as the host infrastructure for Google Cloud's TPU 8tx AI system. The announcement landed on Nov. 9, 2026. The EPYC chips run on the Turin/Zen 5 architecture.

TPUs are Google's own AI accelerators, built to crunch the heavy math behind training and running models. Every accelerator cluster needs a host processor alongside it to move data and coordinate jobs. AMD just locked in that seat inside one of Google Cloud's flagship AI systems.

The TPU 8tx is the first cloud TPU deployed with hardware-enforced confidential computing, powered by AMD's Secure Encrypted Virtualization. SEV keeps a customer's workload encrypted and walled off at the silicon level on shared machines. The combination is aimed at secure inference, fine-tuning and reinforcement learning — a pitch to regulated enterprises and government customers handling sensitive data, where protections live in the hardware rather than in software controls and contractual promises.

This isn't AMD's first ride with Google Cloud. The two launched C4D and H4D virtual machines on fifth-generation EPYC chips in 2025. The new deal stretches that relationship from general-purpose cloud computing into Google's in-house AI hardware stack.

AMD is positioned as the preferred CPU host across more than ten hyperscalers and AI labs, supplying host CPUs alongside GPUs and custom ASICs. That setup means AMD gets paid even when a customer picks someone else's accelerator — when Google runs its own TPUs, AMD still holds the host slot.

One caveat: a 13.26% single-day jump reflects sentiment and expectations, not delivered revenue. What the TPU 8tx actually pays off depends on how widely Google rolls the system out and how much demand materializes from security-conscious customers.