Chey Tae-won vows to double SK Hynix wafer capacity within 5 years, sees memory bottleneck through 2030

SK Group chairman makes first appearance at Computex 2026, Asia's largest IT show

Memory bottleneck to persist at least 5 years as AI expands

HBM4E sample unveiled for the first time

Nvidia CEO Jensen Huang signs HBM4E base die with 'Please Make More'

Nvidia CEO Jensen Huang visits the SK Hynix booth at Computex 2026 in Taipei, Taiwan, on Monday and poses for a photo with SK Group Chairman Chey Tae-won. [SK Hynix]

SK Group Chairman Chey Tae-won, making his first appearance at Computex 2026 — Asia's largest IT trade show, held in Taiwan — said Monday that memory bottlenecks will persist through 2030 and pledged to double SK Hynix's total wafer production capacity within five years.

Nvidia CEO Jensen Huang also visited the SK Hynix booth that day, signing an HBM4E — seventh-generation high bandwidth memory — wafer with the message "Please Make More."

Chey laid out the specific targets during a media interview at the company's booth on the Computex exhibition floor.

He said the memory bottleneck would last at least five more years and that SK Hynix was "pushing capacity expansion at full speed." Building new memory fabs requires enormous investment and takes a minimum of three years, he added, but the company plans to double wafer production capacity over the next five years despite those challenges.

SK Hynix is channeling large-scale investment into its M15X and P&T7 facilities in Cheongju, a semiconductor cluster in Yongin and an advanced packaging factory in Arizona.

Chey's remarks reaffirmed the company's commitment to medium- and long-term capacity expansion, underpinned by that investment strategy.

Nvidia CEO Jensen Huang signs an HBM4E (seventh-generation high bandwidth memory) base die with the message "Please Make More" during a visit to the SK Hynix booth. [SK Hynix]

Chey attributed the memory bottleneck expected to last through 2030 to AI's ongoing expansion. "AI is also in an expansion phase while some degree of supply shortage is occurring," he said. "More capacity and KV cache are needed — meaning more memory is needed. Everyone is scaling up their AI centers, which requires enormous amounts of memory."

He added that the new AI PC Nvidia CEO Jensen Huang announced Monday also demands large amounts of memory simultaneously.

On capital expenditure, Chey said the company does not calculate total spending in advance. "Our attitude right now is to supply whatever is needed," he said. Equipment prices, construction costs, land prices, water costs and electricity rates will all rise, making an exact figure impossible to pin down, he added. "But we have to produce," he said.

The previous day, after the Nvidia GTC 2026 keynote, Chey met with Huang and executives from both companies to discuss the future of AI memory cooperation.

SK Hynix unveils HBM4E sample for the first time

SK Hynix put an HBM4E sample on public display for the first time at its Computex booth Monday.

Huang visited the booth and signed both an HBM4E wafer and a 192-gigabyte SOCAAM2, writing "Please make more ;)" on each.

On the development status of HBM4E, the seventh-generation product, Chey said the timeline depends entirely on the customer. "There is currently only one customer for HBM4E, so it is entirely up to their schedule," he said. "We need to be ready whenever the customer is ready, and we will be."

Chey also outlined plans to build AI factories in collaboration with SK Group affiliates including SK Telecom. "We produce memory chips for AI, but we are also taking on the challenge of how we can build 'AI factories' with an eye to the future," he said. "Many AI factories capable of producing more AI will be needed, and because I believe this will benefit all of humanity, I want to build more AI factories."

'Alliance with TSMC better than ever'; more Taiwan partnerships needed

Chey expressed confidence in the three-way alliance among SK Hynix, Nvidia and TSMC, calling it "better than ever."

"We are collaborating with TSMC on the HBM4 base die," he said. "We rely on TSMC and have the best partnership we have ever had."

On competition, he said he did not know who would win or lose, but that customers want HBM4 and SK Hynix simply delivers it. "We don't think of it as competition," he added.

Chey cited the importance of cooperation with Taiwanese companies in the AI infrastructure ecosystem as the reason for his Taiwan visit.

"The more we expand our AI business, the more we need better and more Taiwan partnerships," he said. He added that he wanted to visit partners in person — not only TSMC but also Foxconn and Acer — to see how those relationships would develop going forward.

"I look forward to deepening Taiwan's AI-related development further," Chey said, "and I think we will be seeing each other more often." After the interview, he visited Foxconn in Taiwan and Essencore, an overseas SK Group affiliate.

go@heraldcorp.com