Google, Intel deepen AI tie-up

Google and Intel said on April 9 they expanded a multiyear deal to keep Intel chips inside Google Cloud and to build new data-center processors together. (intel.com) The agreement keeps Intel Xeon central processing units in Google Cloud for artificial intelligence, inference, and general-purpose computing, according to Intel’s announcement. The companies also said they will expand work on custom application-specific integrated circuit infrastructure processing units, or data-center chips that move network and storage traffic so main processors can focus on customer workloads. (intel.com) Google has used that design before: its C3 virtual machines, introduced in 2022, paired 4th Gen Intel Xeon Scalable processors with Google’s custom Intel infrastructure processing unit. Google said those machines were built for compute-heavy applications such as high performance computing, web serving, and video encoding. (cloud.google.com) The tie-up lands as cloud companies are spending heavily on the hardware behind generative artificial intelligence. In November 2025, Google said it would invest $40 billion in Texas through 2027, including new cloud and artificial intelligence infrastructure and new data-center campuses in Armstrong and Haskell counties. (blog.google) Google is also building more of its own artificial intelligence silicon. In April 2025, the company introduced Ironwood, its seventh-generation Tensor Processing Unit, and said it was designed specifically for inference, the stage when a trained model generates answers for users. (blog.google) That leaves room for several kinds of chips inside one data center. Intel said the new agreement reflects “heterogeneous” systems, meaning different processors handle different jobs: graphics processing units and Tensor Processing Units for model math, central processing units for orchestration and memory-heavy work, and infrastructure processing units for moving data around the network. (intel.com) Intel has been pushing that message as it tries to win more cloud spending back from rivals led by Nvidia and Advanced Micro Devices. Reuters reported on April 9 that shifting artificial intelligence workloads have revived demand for traditional computing chips, giving Intel an opening to sell more central processing units into data centers even as graphics chips dominate model training. (reuters.com) For Google, the arrangement adds another supplier to a stack that already includes its own Tensor Processing Units and outside accelerators. CNBC reported the company committed to use multiple generations of Intel chips in its artificial intelligence data centers as part of the expanded partnership. (cnbc.com) The immediate result is not a new consumer product but a deeper bet on who controls the machinery behind cloud artificial intelligence. Google gets more custom plumbing for its data centers, and Intel gets one of the world’s biggest cloud operators to keep buying in. (intel.com)