Intel and Google join on AI chips

Most artificial intelligence jobs in a data center do not run on one giant chip. They run like a kitchen line, with one processor moving data, another serving storage and networking, and a separate accelerator doing the heavy math. (intel.com) Intel and Google said on April 9, 2026 that they are extending that division of labor together for multiple years inside Google Cloud. Intel said its Xeon central processing units will keep powering Google Cloud systems for artificial intelligence, inference, and general-purpose workloads. (intel.com) A central processing unit is the traffic cop of a server: it schedules work, moves data, and keeps the rest of the machine fed. Intel said that role gets more important as artificial intelligence systems become “heterogeneous,” meaning they mix different kinds of chips in one setup. (intel.com) Google Cloud is already using Intel’s latest Xeon 6 chips in its C4 and N4 virtual machines. Intel said those instances handle everything from coordinating large training jobs to running low-latency inference, which is the step where a model answers a live user request. (intel.com) Google separately said its C4 virtual machines based on Intel Xeon 6 were generally available in 2025, with bare metal options, local solid-state drives, and extra-large shapes. Google said those machines delivered up to 30% better general compute performance and up to 60% better machine-learning recommendation performance than the prior generation. (cloud.google.com) The newer piece is the chip they are building together. Intel said the companies are expanding co-development of custom infrastructure processing units, which are application-specific chips built to offload jobs like networking, security, and data movement so the main processors can spend more time on customer workloads. (intel.com) Google has been down this road before with Intel. In 2022, Google Cloud introduced its C3 machine family with a custom Intel infrastructure processing unit, saying the chip helped move platform tasks off the main server and improved efficiency for storage and networking. (cloud.google.com) That makes this week’s announcement less like a brand-new alliance and more like a decision to keep standardizing a bigger slice of Google Cloud around Intel parts. Data Center Dynamics reported that Google committed to multiple future generations of Intel central processing units, while the companies did not disclose financial terms or a timeline for the custom chip work. (datacenterdynamics.com) The timing matters for Intel because cloud companies have been designing more of their own silicon and buying more graphics processing units for artificial intelligence. This deal gives Intel a clearer place in that stack by tying its Xeon chips to Google Cloud’s own infrastructure design instead of selling a standalone processor and hoping it wins on price. (techcrunch.com) It also shows what enterprise artificial intelligence hardware is starting to look like in practice: not one winner, but bundles. Google supplies the cloud, Intel supplies the host processors, both sides shape the infrastructure chip, and customers rent the finished system as a service instead of piecing it together rack by rack. (intel.com)