Google expands partnership with Intel for AI chips

Google has dedicated to utilizing a number of generations of Intel central processing items in its synthetic intelligence information facilities, an growth of an present partnership.

The web large has lengthy relied on Intel processors, relationship again to its earliest server rack ambitions practically three a long time in the past. Intel’s latest Xeon 6 CPUs will now run AI coaching and inference workloads, probably giving the chipmaker a stronger place in an AI market that is to date been dominated by Nvidia. 

“Their Xeon roadmap gives us confidence that we can continue to meet the growing performance and efficiency demands of our workloads,” Amin Vahdat, Google’s chief technologist for AI infrastructure, mentioned in an announcement Thursday.

No monetary phrases have been disclosed, nor did the businesses present a timeline for the settlement.

The deal lands as the CPU takes center stage within the subsequent part of the AI race. Dion Harris, Nvidia‘s head of AI infrastructure, advised CNBC in March that CPUs are “becoming the bottleneck” as agentic workloads transfer compute wants past the graphics processing items which have dominated AI to date.

“Scaling AI requires more than accelerators — it requires balanced systems,” Intel CEO Lip-Bu Tan mentioned in an announcement in regards to the Google deal on Thursday.

Intel, which has been struggling for years to maintain tempo with new traits in know-how, sold a ten% stake to the U.S. authorities in August, with the Trump administration touting the chipmaker’s skill to make superior chips on U.S. soil. The following month, Nvidia mentioned it will buy a $5 billion stake in Intel.

Shares of Intel have practically tripled prior to now 12 months, fueled by these investments.

Intel makes the newest Xeon processor on its most superior 18A know-how at its Arizona chip fabrication plant that opened last year. Despite pouring billions into the foundry aspect of its enterprise, Intel’s personal processors stay the biggest buyer on the new fab. 

But Tan posted on LinkedIn earlier this week that Elon Musk has tapped Intel to design, fabricate and bundle customized chips for SpaceX, xAI and Tesla at his formidable Terafab challenge in Texas, although no monetary particulars or timeline was introduced.

As a part of Thursday’s announcement, Google and Intel reiterated that they are collaborating on one other kind of chip, the infrastructure processing unit, or IPU, which the 2 corporations have labored on collectively since 2022. In a press launch, Intel mentioned this programmable accelerator is used to “offload networking, storage and security functions from host CPUs.” 

Google advised CNBC in an e-mail that the IPU was a first-of-its-kind chip when the businesses first collaborated on it 4 years in the past. Google mentioned it is designed to assist prospects higher make the most of the principle CPU in a standard information heart by taking up “overhead” duties, similar to routing community site visitors, managing storage, encrypting information and working virtualization software program.

For over a decade, Google has additionally developed its personal customized AI accelerator known as the tensor processor unit, or TPU. In 2024, Google additionally began making its own custom CPU, Axion, selecting an Arm-based design over Intel’s main x86 structure.

— CNBC’s Kristina Partsinevelos contributed to this report.

WATCH: How advanced packaging became the next bottleneck for making AI chips