Google Doubles Down on Intel CPUs for AI Systems, Signaling a Broader Industry Shift

Google is deepening its long-standing reliance on Intel, committing to deploy multiple future generations of Intel Xeon processors in its artificial intelligence data centers. This expanded partnership, announced Thursday, will see Intel's newest Xeon 6 CPUs handling both AI training and inference work, a notable win for Intel in a field where Nvidia has held a commanding lead.

Google's chief technologist for AI infrastructure, Amin Vahdat, pointed to Intel's processor roadmap as a foundation for meeting the escalating performance needs of Google's services. The agreement arrives as industry attention pivots back to the central processing unit. Nvidia's own AI infrastructure lead, Dion Harris, noted in March that CPUs are emerging as a limiting factor for advanced AI workloads, suggesting a need for more balanced systems beyond just powerful accelerators.

Intel CEO Lip-Bu Tan echoed that sentiment, stating that scaling AI effectively depends on this kind of system-wide balance. The deal provides a significant endorsement for Intel, which has recently secured substantial government investment and seen its stock price surge. The company manufactures its latest Xeon chips at its advanced Arizona fabrication plant.

Separately, the two tech firms confirmed ongoing collaboration on an Infrastructure Processing Unit (IPU), a specialized chip designed since 2022 to manage networking and storage tasks, thereby freeing up the main CPUs for primary workloads. This move complements Google's parallel development of its own custom AI accelerator (TPU) and, more recently, its Arm-based Axion CPU, indicating a multifaceted and strategic approach to building its AI infrastructure.

Source: CNBC