Chip large Qualcomm introduced Monday that it’s going to launch two new synthetic intelligence chips – setting the stage for a head-to-head race with billionaire Jensen Huang’s Nvidia.
It’s a serious shift for Qualcomm, which has lengthy centered on manufacturing semiconductors for wi-fi exercise and cell phones – not for the power-hungry knowledge facilities which might be powering the AI growth.
Shares in Qualcomm jumped 11% Monday to shut at $187.68.
Qualcomm — headed by Brazilian-born CEO Cristiano Amon — stated its new, super-powered chips can be obtainable in liquid-cooled server racks, which assist prospects like cloud service suppliers save on operational prices.
Nvidia and AMD additionally promote their graphics processing models, or GPUs, in full-rack techniques. Giant language fashions like OpenAI’s ChatGPT depend on these highly effective techniques.
Nonetheless, Qualcomm is going through a steep uphill battle. Nvidia already dominates about 90% of the information middle market and boasts a large $4.6 trillion market cap.
Nvidia’s chips had been used to coach OpenAI’s ChatGPT, nonetheless extensively seen as the preferred AI chatbot.
However earlier this month, OpenAI broadened its provide chain. It introduced plans to begin shopping for chips from rival AMD, and even probably take a stake within the firm.
The AI race has proven no indicators of slowing down. Practically $6.7 trillion in capital expenditures can be spent on AI infrastructure via 2030, in response to a McKinsey estimate.
Qualcomm’s new AI200 and AI250 chips will go on sale in 2026 and 2027, respectively, the corporate stated in a press launch.
A rack would require 160 kilowatts of energy, much like rack options from Nvidia.
Qualcomm’s chip-based accelerator playing cards will assist 768 gigabytes of reminiscence – greater than comparable choices from Nvidia and AMD.
The corporate declined to touch upon the worth of its AI chips, playing cards or racks.
“Qualcomm AI200 and AI250 are designed for frictionless adoption and fast innovation,” Durga Malladi, the agency’s senior vp and common supervisor for expertise planning, stated in a press release Monday.
“Our wealthy software program stack and open ecosystem assist make it simpler than ever for builders and enterprises to combine, handle, and scale already skilled AI fashions on our optimized AI inference options.”
Qualcomm’s new AI chips can be primarily based on its Hexagon neural processing models, or NPUs, that are used for its smartphone chips.
The corporate goals to promote its new chips and different components individually, so prospects can design their very own racks in the event that they’d like. Nvidia or AMD might even turn out to be potential shoppers for some knowledge middle components, Malladi advised reporters final week.
“What now we have tried to do is make it possible for our prospects are ready to both take all of it or say, ‘I’m going to combine and match,’” he stated.
