Google Expands Partnership with Intel for AI Data Center Chips

Google has committed to using multiple generations of Intel central processing units in its artificial intelligence data centers. This agreement expands an existing partnership between the companies. Google has relied on Intel processors since its early server operations nearly three decades ago.

Intel's Xeon 6 CPUs will support AI training and inference workloads in Google's data centers. The companies announced the expansion on Thursday. No financial terms or timeline for the agreement were provided.

Vahdat, Google's chief technologist for AI infrastructure, stated that Intel's Xeon roadmap supports the performance and efficiency needs of Google's workloads.

“Their Xeon roadmap gives us confidence that we can continue to meet the growing performance and efficiency demands of our workloads.”

Intel CEO Lip-Bu Tan commented on the need for balanced systems in scaling AI, beyond accelerators alone. The partnership also involves ongoing collaboration on infrastructure processing units, or IPUs, since 2022. IPUs offload networking, storage, and security functions from host CPUs.

Google described the IPU as a first-of-its-kind chip when the collaboration began four years ago. It handles tasks such as routing network traffic, managing storage, encrypting data, and running virtualization software. This allows better utilization of main CPUs in data centers.

has developed its own custom AI accelerator, the tensor processing unit, or TPU, for over a decade. In 2024, Google introduced its custom CPU, Axion, based on Arm architecture rather than Intel's x86 design. Intel produces its latest Xeon processors using 18A technology at a fabrication plant in Arizona that opened last year.