Intel–Google infrastructure partnership

Intel and Google said on April 9 they signed a multiyear deal to build more of Google Cloud’s artificial intelligence infrastructure around Intel Xeon processors and jointly designed infrastructure chips. (intel.com) Intel said Google Cloud will keep using multiple generations of Intel Xeon central processing units for artificial intelligence inference, cloud services, and general computing workloads. The companies also said they will expand co-development of custom infrastructure processing units, or network-and-data-movement chips, built as application-specific integrated circuits. (intel.com) Google already sells cloud servers based on Intel chips, including instances with Intel Xeon processors for enterprise workloads. Its artificial intelligence infrastructure page says Google uses a global data-center network and offers customers a mix of cloud tensor processing units, graphics processing units, and central processing units. (cloud.google.com 1) (cloud.google.com 2) A central processing unit is the general-purpose manager in a server, while an infrastructure processing unit handles networking and data traffic so the main chips spend more time running models. Intel said that split is becoming more important as artificial intelligence systems mix different kinds of processors inside the same data center. (intel.com) The deal lands as cloud providers are trying to lower the cost of serving models after training, a stage called inference that handles live user requests. Intel said Xeon will continue to power Google Cloud infrastructure for inference, and Google has separately been expanding graphics processing unit capacity and managed model services on its cloud platform. (intel.com) (cloud.google.com 1) (cloud.google.com 2) The announcement also shows Google keeping a mixed hardware strategy instead of relying on a single chip vendor. In March 2026, Google Cloud announced new artificial intelligence infrastructure tied to Nvidia products, including support for Vera Rubin systems and broader Nvidia software integration. (cloud.google.com) Intel framed the partnership as evidence that central processing units still matter in artificial intelligence clusters even as accelerators get most of the attention. Chief Executive Lip-Bu Tan said scaling artificial intelligence "requires balanced systems," with general-purpose compute and purpose-built infrastructure working together. (intc.com) For Intel, the agreement adds a large cloud customer to its effort to defend Xeon’s role inside artificial intelligence data centers. For Google, it preserves flexibility in the lower layers of cloud infrastructure while it keeps building services on top for training, inference, and orchestration. (cnbc.com) (intel.com) The next test is whether this kind of shared design work changes costs and performance for cloud customers, not just the supplier list inside Google’s server racks. For now, both companies are betting that the busiest artificial intelligence systems will keep running on a mix of general-purpose processors, accelerators, and custom networking silicon. (intel.com) (cloud.google.com)