Intel and Google partner on AI servers
Intel and Google announced a multi‑year collaboration to deploy Intel Xeon processors in Google’s next‑generation AI and cloud infrastructure and to co‑develop custom infrastructure processing units for networking, storage and security. The deal ties a major cloud operator to Intel’s server roadmap and signals continued vendor-level competition over AI datacenter stacks. For buyers and sellers in TMT, it highlights where infrastructure investments and vendor tie‑ups could influence valuation and capex assumptions. (x.com)
Google just made a very public bet that the chips running an artificial intelligence data center are not only the flashy graphics processors. On April 9, Intel said Google will keep using Intel Xeon central processing units across Google Cloud for artificial intelligence, inference, and general-purpose workloads under a multiyear deal. (intel.com) That sounds odd until you remember how an artificial intelligence server actually works. The graphics processor does the heavy math, but the central processing unit is the traffic cop that feeds data, schedules jobs, manages memory, and keeps the whole machine from idling. (intel.com) Google is also not just buying off-the-shelf chips. Intel said the two companies will expand co-development of custom application-specific integrated circuit infrastructure processing units, which are helper chips built for networking, storage, and security jobs inside the data center. (intel.com) Google has already been using that design pattern in public cloud products. Its C3 machine series combined fourth-generation Intel Xeon processors with Google’s custom Intel infrastructure processing unit, and its Z3 storage machines did the same while adding a newer local solid-state drive stack. (cloud.google.com, cloud.google.com) The practical point is that a cloud server wastes money when expensive graphics processors wait around for data. Intel said the new work is aimed at better performance, utilization, and energy efficiency at scale, which is corporate language for getting more output from the same rack of hardware. (intel.com) This also lands at an awkward moment for Intel. Google already sells Arm-based Axion central processing unit instances of its own, so this agreement says Google still wants Intel in parts of its future fleet even while it develops in-house alternatives. (msn.com) Google had already been one of the first big cloud operators to roll out Intel’s latest server generation. In August 2025, Google Cloud made C4 virtual machines based on Intel Xeon 6, code-named Granite Rapids, generally available and said those systems delivered up to 30% gains for general compute and up to 60% for machine-learning recommendation workloads. (cloud.google.com) So this is less a surprise wedding than a renewal of vows with new terms. Intel keeps a hyperscale customer tied to multiple future Xeon generations, and Google gets influence over custom plumbing chips that sit between the processor, the network, the storage drives, and the security layer. (intel.com, cloud.google.com) The bigger fight is over who owns the whole artificial intelligence stack, not just one chip socket. Nvidia sells graphics processors, Google builds tensor processing units, Amazon designs Graviton central processing units, and Intel is trying to stay indispensable by owning the control layer that keeps mixed fleets of chips working together. (intel.com, cloud.google.com) If that strategy works, Intel does not need to win every headline benchmark to stay inside the room. It just needs Google’s next generation of servers to keep needing Xeon processors and custom infrastructure processing units every time a customer spins up an artificial intelligence workload. (intel.com)
