Sign up to save your podcasts
Or
Share SK chief vows to ease memory chip bottleneck with new mega fabs
Share to email
Share to Facebook
Share to X
Share to email
Share to Facebook
Share to X
Or
SK chief vows to ease memory chip bottleneck with new mega fabs
This article is by Lee Jae-lim and read by an artificial voice.
SK Group Chairman Chey Tae-won said Monday that the conglomerate is ramping up investment in next-generation semiconductor plants to tackle growing memory supply bottlenecks, as demand for high-performance chips surges amid the global AI boom.
The company is building massive new facilities in Cheongju, North Chungcheong and Yongin, Gyeonggi, with a combined investment capacity of more than 140 trillion won ($98 billion).
"We are receiving more requests for memory supply than ever before - so much so that it has become a serious operational challenge," Chey said at the annual SK AI Summit held at Coex in southern Seoul. "If we fail to meet this demand, some of our customers could face disruptions in their own operations. Meeting these commitments is critical to maintaining trusted, long-term relationships."
SK hynix, the group's chipmaking unit, is pushing ahead to meet the surge in demand. Its new M15X fab in Cheongju, built with an investment of over 20 trillion won, has completed construction and begun equipment installation two months earlier than scheduled. The facility will produce next-generation dynamic random-access memory chips once operations begin.
The company's Yongin semiconductor cluster, a separate project worth more than 120 trillion won, is set to begin operations in 2027. The vast campus will host four large fabs, each with production capacity equivalent to six M15X plants.
"Our strategy is to secure the manufacturing space first and then expand equipment installation in line with market needs," Chey said. "It requires significant upfront capital, but it's essential to preventing future supply shortages."
With Big Tech firms rolling out increasingly powerful processors for AI, memory bandwidth has emerged as a key bottleneck holding back performance. Even the most advanced graphics processing units (GPUs) cannot fully operate without enough bandwidth between processors and memory. The most practical solution, for now, is to increase the number of high bandwidth memory units paired with each GPU - driving an unprecedented appetite for advanced memory chips.
Chey said SK hynix's leadership in memory technology is now "beyond question."
"Our technological capabilities are already well recognized across the industry," he said. "Even Nvidia - famous for its relentless innovation - no longer questions our development speed. That reflects their confidence in our readiness."
SK is also deepening partnerships with multiple Big Tech players to build out more AI infrastructure in Korea. Last week, the company and Nvidia announced a deal for the supply of 50,000 Blackwell processors, which SK will use to automate its semiconductor production process and create a "virtual factory" tailored to SK hynix's operations.
In parallel, SK hynix will supply memory chips to OpenAI's $500 billion Stargate project, while SK Telecom plans to build an OpenAI-customized data center - dubbed the Korean version of Stargate - in the country's southwest. SK and Amazon Web Services are also co-investing $5 billion to establish a 100-megawatt AI data center in Ulsan, slated to begin operations in 2027.
View all episodes
By Newsroom of the Korea JoongAng DailyThis article is by Lee Jae-lim and read by an artificial voice.
SK Group Chairman Chey Tae-won said Monday that the conglomerate is ramping up investment in next-generation semiconductor plants to tackle growing memory supply bottlenecks, as demand for high-performance chips surges amid the global AI boom.
The company is building massive new facilities in Cheongju, North Chungcheong and Yongin, Gyeonggi, with a combined investment capacity of more than 140 trillion won ($98 billion).
"We are receiving more requests for memory supply than ever before - so much so that it has become a serious operational challenge," Chey said at the annual SK AI Summit held at Coex in southern Seoul. "If we fail to meet this demand, some of our customers could face disruptions in their own operations. Meeting these commitments is critical to maintaining trusted, long-term relationships."
SK hynix, the group's chipmaking unit, is pushing ahead to meet the surge in demand. Its new M15X fab in Cheongju, built with an investment of over 20 trillion won, has completed construction and begun equipment installation two months earlier than scheduled. The facility will produce next-generation dynamic random-access memory chips once operations begin.
The company's Yongin semiconductor cluster, a separate project worth more than 120 trillion won, is set to begin operations in 2027. The vast campus will host four large fabs, each with production capacity equivalent to six M15X plants.
"Our strategy is to secure the manufacturing space first and then expand equipment installation in line with market needs," Chey said. "It requires significant upfront capital, but it's essential to preventing future supply shortages."
With Big Tech firms rolling out increasingly powerful processors for AI, memory bandwidth has emerged as a key bottleneck holding back performance. Even the most advanced graphics processing units (GPUs) cannot fully operate without enough bandwidth between processors and memory. The most practical solution, for now, is to increase the number of high bandwidth memory units paired with each GPU - driving an unprecedented appetite for advanced memory chips.
Chey said SK hynix's leadership in memory technology is now "beyond question."
"Our technological capabilities are already well recognized across the industry," he said. "Even Nvidia - famous for its relentless innovation - no longer questions our development speed. That reflects their confidence in our readiness."
SK is also deepening partnerships with multiple Big Tech players to build out more AI infrastructure in Korea. Last week, the company and Nvidia announced a deal for the supply of 50,000 Blackwell processors, which SK will use to automate its semiconductor production process and create a "virtual factory" tailored to SK hynix's operations.
In parallel, SK hynix will supply memory chips to OpenAI's $500 billion Stargate project, while SK Telecom plans to build an OpenAI-customized data center - dubbed the Korean version of Stargate - in the country's southwest. SK and Amazon Web Services are also co-investing $5 billion to establish a 100-megawatt AI data center in Ulsan, slated to begin operations in 2027.