Intel–Google AI chip partnership
Intel and Google announced a multiyear collaboration to advance AI and cloud infrastructure that pairs Intel Xeon processors with custom IPUs for balanced systems. The deal signals hyperscalers and chipmakers are trying to diversify beyond single‑vendor hardware stacks for datacenters. (x.com).
Google and Intel said on April 9 that they are extending their chip partnership for several future server generations, and Google Cloud will keep using Intel’s Xeon processors while the two companies co-develop custom infrastructure chips for its data centers. (intel.com) (cnbc.com) A modern data center does not run on one kind of chip any more. One chip does the main computing, another chip trains artificial intelligence models, and another chip handles the plumbing that moves data, storage, and security checks around the building. (intel.com) (datacenterdynamics.com) The main computing chip here is the central processing unit, which is the general-purpose brain of a server. Intel says its Xeon line will keep powering Google Cloud systems used for artificial intelligence inference, general cloud workloads, and the orchestration work that ties bigger systems together. (intel.com) (newsroom.intel.com) The new piece is the infrastructure processing unit, which is a helper chip that takes networking, storage, and security chores off the main processor the way a warehouse conveyor belt takes lifting work off the people packing boxes. Google and Intel have been working on that idea since at least 2021, when they introduced the first co-designed chip in this category under the code name Mount Evans. (datacenterdynamics.com) (datacenterknowledge.com) That helper chip matters because every networking packet or storage request handled somewhere else frees the main processor to do work customers actually rent. Data Center Dynamics says the infrastructure processing unit offloads those infrastructure tasks so server central processing unit cycles can be used for revenue-generating jobs instead. (datacenterdynamics.com) Google is not starting from zero here. In July 2025, Google Cloud said it was the first leading hyperscale cloud provider to bring Intel Xeon 6 into general availability, and those C4 virtual machines claimed up to 30% better general compute performance and up to 60% better machine learning recommendation performance than the prior generation. (cloud.google.com) Intel needs deals like this because the artificial intelligence boom has made graphics processing units from Nvidia the center of the market, while central processing units risk looking like background parts. Intel’s pitch in this announcement is that artificial intelligence systems still need “balanced systems,” with central processors coordinating data movement and custom infrastructure chips keeping the whole machine fed. (cnbc.com) (intel.com) Google has its own reason to do this. The biggest cloud companies have spent the last few years trying to avoid depending on a single vendor for every layer of their artificial intelligence stack, so a multiyear commitment to Intel central processing units plus co-designed infrastructure chips gives Google another route besides buying ever more accelerator-heavy systems from one supplier. (techcrunch.com) (datacenterdynamics.com) The detail to watch is not a single chip launch but the phrase “multiple generations.” That means Google is betting several years of cloud build-out on Intel’s server roadmap, and Intel is getting something more valuable than a one-time order: a seat inside one of the world’s biggest data center planning cycles. (intel.com) (cnbc.com)
