SK hynix to double wafer production within five years, chief says

SK hynix will "double its wafer production capacity within the next five years," according to SK Group Chairman Chey Tae-won on Tuesday. It marks the first time SK Group has publicly disclosed a concrete target for expanding production capacity.

“Building a new memory fab requires enormous investment and takes at least three years,” Chey told reporters at the SK hynix booth at Computex 2026 in Taipei, Taiwan.

“The memory bottleneck will continue through 2030," he said, referring to a memory shortage caused by surging AI-driven demand that continues to outpace supply.

SK hynix is already investing heavily in new facilities, including its M15X fab and advanced packaging P&T7 facility for memory chips in Cheongju, North Chungcheong, and a semiconductor cluster zone in Yongin, Gyeonggi. The Korean chipmaker has also made massive investments in an advanced packaging facility in the United States.

Chey’s remarks signaled the company’s will to stay ahead of the AI boom while narrowing the gap with its crosstown rival, Samsung Electronics, according to market observers.

Samsung captured 38.5 percent share of the global dynamic random-access memory (DRAM) market in the first quarter of this year, compared with 28.8 percent for SK hynix, according to market research firm TrendForce.

Chey expressed concern about the sharp rise in memory prices.

“Pricing is driven by supply and demand, so it’s not something I can control or decide,” he said. “But a sudden surge in [memory] prices could undermine the long-term sustainability of the entire AI industry.”

He also stressed that ramping up production would be no easy task.

“AI data centers and AI factories are facing multiple bottlenecks including financing, energy, graphics processing units and memory,” Chey said. “As manufacturers must also secure equipment, electricity and water, expanding capacity quickly enough to alleviate memory shortages is never a simple task.”

Earlier in the day, Chey toured the SK hynix exhibit with Nvidia founder and CEO Jensen Huang.

Huang left a message — “Please make more” — on a physical HBM4E wafer prepared by SK hynix. HBM4E is the Korean chipmaker’s seventh-generation high bandwidth memory (HBM) product.

The industry appears headed for an even stronger upswing in memory prices as persistent shortages collide with booming demand for HBM and dwindling supplies of conventional memory.

TrendForce forecast in its report on Tuesday that contract prices for HBM could rise severalfold by 2027. The firm expects HBM to account for 22 percent of total DRAM wafer input this year, up from 18 percent last year. The firm also projected that the figure would reach 30 percent by 2027.

As custom AI chips such as Nvidia’s next-generation Vera Rubin agentic AI platform and Google’s Tensor Processing Unit gain wider adoption, HBM is becoming an increasingly critical component of AI hardware. It consequently prompts suppliers to accelerate production expansion, according to market observers.

But the expansion of HBM production comes at a cost — reduced DRAM supply.

Because HBM chips are larger and more complex to manufacture, fewer chips can be produced from a single wafer than conventional DRAM. TrendForce expects that suppliers will see their pricing leverage grow as production capacity for conventional DRAM will be less available when HBM technology advances with a growing base die. Base die refers to the HBM chip's base layer.

This article was originally written in Korean and translated by a bilingual reporter with the help of generative AI tools. It was then edited by a native English-speaking editor. All AI-assisted translations are reviewed and refined by our newsroom.

BY LEE YOUNG-KEUN [lee.soojung1@joongang.co.kr]