Google reportedly taps AMD to design next-generation TPU — hybrid AI ASIC could integrate on-package CPU cores for reinforcement learning
Google may be building a TPU with on-package CPU cores specifically for agentic and reinforced learning workloads, according to a rumor.
Google is partnering with AMD to develop the next generation of its TPUs, according to a report. The collaboration may involve integrating CPU cores on the TPU for reinforcement learning and other CPU-intensive tasks. This would be AMD's first major involvement in a custom AI ASIC project. The partnership could also indicate Google's interest in a new type of TPU that combines its proprietary accelerator technology with on-board general-purpose cores.
Google has been increasing CPU resources for its TPU inference systems, with a ratio of one CPU to every two TPUs for the 8i models. Bringing CPU cores directly into the TPU package could potentially improve performance and reduce power consumption. AMD has experience in developing data center-grade designs with both CPU and accelerator chiplets, making it a potential partner for Google's new TPU.
Written by urgent.news from Tom's Hardware's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
