(Bloomberg) — Chairman Chey Tae-won of SK Hynix (000660.KS) parent SK Group pledged to grow production of AI memory chips to meet a surge in demand from the global data center buildout.
The billionaire chief of South Korea’s second-largest conglomerate also called high-bandwidth memory a “monster chip” that’s generating enormous profits for SK Hynix when speaking at a conference in Washington on Feb. 20. The chipmaker’s share price has more than quadrupled over the past year on record earnings.
Most Read from Bloomberg
Photographer: SeongJoon Cho/Bloomberg
While Chey did not specify the scale of his chip firm’s expansion, SK Hynix said in January that its capital expenditure in 2026 will rise significantly from its spending last year to satisfy demand for HBM chips that are required to make accelerators designed by the likes of Nvidia Corp. to train and run artificial intelligence services.
US tech firms from Microsoft Corp. (MSFT) to Meta Platforms Inc. (META) are allocating about $650 billion this year for infrastructure that gives them an edge in the race to build AI technologies. That record spending is causing a global shortage of memory chips, a market that’s dominated by SK Hynix, its South Korean peer Samsung Electronics Co. (005930.KS), and US-based Micron Technology Inc. (MU).
SK Hynix has sold out its entire slate of memory chips in 2026, while Micron has done similarly with its HBM offerings.
But Chey also cautioned losses are still a possibility in the future due to potential changes in the competitive landscape caused by rapid technological shifts.
The average of analyst projections for SK Hynix’s annual operating profit for 2026 has risen to $70 billion in January from about $50 billion late last year, and some have revised that up again to more than $100 billion, according to Chey.
“That sounds like really good news,” Chey said, “but it could just as easily turn into a $100 billion loss.”
Chey also highlighted mounting infrastructure challenges. He said SK Group is now exploring building power plants alongside AI data centers, as failure to meet energy demand could be “disastrous.”
Most Read from Bloomberg Businessweek
©2026 Bloomberg L.P.