SAN JOSE, California, March 20 (Reuters) - The same booming AI demand that has caused the chairman of South Korea's SK Group to predict shortages of high-bandwidth memory chips could also cause tight supplies for storage drives, an executive from the South Korean firm's U.S.-based AI subsidiary Solidigm told Reuters this week.
Memory chips that sit right next to computing chips are critical to the servers sold by Nvidia and others, and earlier this week at Nvidia's annual developer conference, SK Group Chairman Chey Tae-won told reporters that shortages of high-bandwidth memory could last until 2030 because of demand for AI systems.
But new AI-powered software tools can now make sense of huge datasets that humans could rarely wring business value from, and Nvidia introduced several new technologies this week aimed at helping move data more quickly from storage drives to its chips.
“The storage system is going to get pounded," Nvidia CEO Jensen Huang said on Monday during a keynote addressat the company's developer conference.
That is translating into increased demand for solid state drives, said Greg Matson, senior vice president at Solidigm, the unit of SK Hynix that sells storage drives to businesses. Speaking on the sidelines of Nvidia's conference this week, Matson said AI systems coming later this year could require 35% more storage than previous systems.
"It's going to be tight," Matson said of supplies of storage memory between now and 2030.
"We'll be coming out with higher-density drives from a silicon perspective later this year, and even expanding our manufacturing output as well," Matson said. "But can we keep up? No, we can't. I could sell twice as much as I am today."
(Reporting by Stephen Nellis in San Jose, California; Editing by Mark Porter)
