AMD EPYC processors to power Google Cloud TPU 8tx

AMD says its fifth-generation EPYC processors power the host infrastructure for Google Cloud TPU 8tx.[1] Google Cloud says it paired those processors with its eighth-generation TPU architecture inside its AI Hypercomputer.[1] AMD describes the collaboration as combining EPYC performance, Infinity Guard security and x86 compatibility with Google Cloud TPU acceleration.[1] The statement from AMD carries a timestamp of November 9, 2026, at 8 a.m. EST.[1] AMD says organizations can run sensitive AI workloads at scale in the cloud without compromising data protection.[1]

AMD says EPYC CPUs handle host processing in systems using GPUs and custom accelerators, including preparing data and coordinating workloads.[1] On TPU 8tx, AMD says those host-processing capabilities are combined with hardware-rooted security to protect data while it is in use.[1] Google Cloud says the pairing delivers silicon-level data protection and low-latency performance for modern reasoning models.[1]

Google Cloud says the pairing brings hardware-enforced confidential computing to cloud TPUs for the first time.[1] AMD characterizes the approach as enabling sensitive AI workloads to run at scale in the cloud without compromising data protection.[1] Google says its eighth-generation TPU portfolio includes TPU 8tx and TPU 8i Max, bringing its total to four new TPU systems in one year.[2] The two named systems have different stated aims: Google describes 8tx as designed for secure inference and 8i Max as designed for massive context.[2]

The supplied coverage reports a rise in AMD shares after the announcement, but the figures differ: GuruFocus reported about 4% in Monday morning trading, while Livemint reported 12.5%.[3][4] The evidence does not establish that the reports refer to the same measurement time. AMD says EPYC processors are the host CPU of choice at more than 10 leading hyperscalers and frontier AI labs.[1] It also says its EPYC generations have developed performance, memory-bandwidth and I/O capabilities for accelerated systems.[1]