The Korea Herald

ssg
피터빈트

Naver, Intel team up to counter Nvidia’s dominance in AI chips

Joint lab aims to nurture software ecosystem based on Intel’s chips

By Jo He-rim

Published : April 10, 2024 - 21:55

    • Link copied

Ha Jung-woo (right), head of AI Innovation at Naver Cloud, speaks with Intel CEO Pat Gelsinger during the Intel Vision 2024 keynote speech session, in Phoenix, Tuesday. (Intel Corporation) Ha Jung-woo (right), head of AI Innovation at Naver Cloud, speaks with Intel CEO Pat Gelsinger during the Intel Vision 2024 keynote speech session, in Phoenix, Tuesday. (Intel Corporation)

US chip giant Intel is teaming up with Naver, Korea's leader in artificial intelligence, in an apparent move to counter Nvidia's dominance in the burgeoning AI chip market.

Unveiling its more powerful next-generation AI accelerator, Gaudi 3, on Tuesday, Intel CEO Pat Gelsinger said the cutting-edge central processing unit has been enhanced to perform better than Nvidia's graphic processing unit "at a fraction of the cost." He also said the company is working together with global tech firms such as Naver, Bosch and IBM to expand its AI chip supply chain.

“Innovation is advancing at an unprecedented pace, all enabled by silicon -- and every company is quickly becoming an AI company,” Gelsinger said, delivering his keynote speech at the Intel Vision 2024 held in Phoenix.

“Intel is bringing AI everywhere across the enterprise, from the PC to the data center to the edge. Our latest Gaudi, Xeon and Core Ultra platforms are delivering a cohesive set of flexible solutions tailored to meet the changing needs of our customers and partners and capitalize on the immense opportunities ahead.”

Intel, the world's biggest chipmaker by revenue, and other global tech firms such as OpenAI and Microsoft, are each working to reduce their reliance on Nvidia, which is currently controlling 95 percent of the global market with its GPUs.

Intel said the Gaudi 3 AI accelerator has been enhanced to perform 50 percent on average better inference and 40 percent on average better power efficiency when compared to Nvidia's H100.

While the price has not been revealed, it is anticipated to be lower than Nvidia's H100, which sells at around $25,000 to $40,000 per unit.

In its race against Nvidia, Intel announced that it is joining hands with a number of global tech firms, including Naver. At the keynote session on Tuesday, Naver's top executives made a surprise appearance to officially announce their partnership.

"Naver was the third company globally to announce our own hyper-scale Gen AI model, HyperCLOVA X, and for this model to work with our services and ecosystem, we must focus on R&D for powerful and cost-effective computing power. That's why we are very much excited about working together with Intel," Naver CEO Choi Soo-yeon said in a video message prepared in advance.

Naver CEO Choi Soo-yeon speaks in a pre-recorded video message during the Intel Vision 2024 keynote speech session, in Phoenix, Tuesday. (Intel Corporation) Naver CEO Choi Soo-yeon speaks in a pre-recorded video message during the Intel Vision 2024 keynote speech session, in Phoenix, Tuesday. (Intel Corporation)

"I can't wait to see what Naver and Intel can create for a new AI chip environment, and the innovations this will bring."

Ha Jung-woo, head of AI Innovation at Naver Cloud, appeared on the stage to explain how Naver and Intel established a joint lab to expand the Gaudi software ecosystem.

With Intel, Naver plans to establish an LLM training infrastructure on Gaudi 2 and offer commercial cloud instances of this LLM training infrastructure to global enterprises and public sectors for their competitiveness and innovations in the new generative AI sector, Ha explained.

"This collaboration aims for Naver to deliver our most powerful LLM on high performance and cost-effective, and for the vision of AI everywhere, beyond our effort for AI sovereignty," Ha said.

Naver has been using Nvidia's GPU chips as the main tool for its AI services, but it recently turned to Intel as the main supplier largely due to Nvidia's price hikes and supply delay.

Intel's Gaudi 3 accelerator will be available to original equipment manufacturers, including Dell Technologies and Hewlett Packard, in the second quarter of 2024, according to the tech giant. General availability is anticipated in the third quarter.