South Korea's chip giants (Samsung, SK Hynix) announce $700 billion in AI-related semiconductor and infrastructure deals
South Korea's two largest chip companies each emerged from a San Francisco AI summit with sprawling partnerships valued at nearly $700 billion combined, positioning the nation as a critical supplier to America's AI buildout.
On July 24, SK Group announced a partnership with NVIDIA valued at more than $500 billion over five years. Under the deal, SK Telecom will construct a two-gigawatt AI factory in Korea using NVIDIA's Vera Rubin systems, with the first phase expected online in 2027. The facility will run on SK hynix's HBM4 memory, and SK hynix separately committed to a long-term arrangement to supply and help develop next-generation memory for NVIDIA. "By leveraging SK hynix's AI memory and SK Telecom's AI infrastructure capabilities, SK will collaborate with NVIDIA to build a world-class AI factory," SK Group Chairman Chey Tae-won said in a statement, adding that the goal was to help Korea "become a global hub that drives AI innovation."
The second deal paired Samsung Electronics and Broadcom in a memorandum of understanding expected to be worth more than $200 billion through 2030, covering memory, chip manufacturing and packaging. Samsung will supply high-bandwidth memory for Broadcom's AI accelerators and manufacture Broadcom chips, including networking silicon, on its two-nanometre process. The companies also plan to collaborate on advanced packaging that stacks memory and logic more closely together.
For Samsung, this deal carries significance beyond its headline value. The company's contract manufacturing arm holds an estimated 7% of the global foundry market, far behind Taiwan's TSMC, and has struggled to win flagship customers. Winning a designer of Broadcom's stature represents a public endorsement of Samsung's 2nm roadmap, which has faced persistent questions over production yields. Analysts point to the appeal of Samsung's vertically integrated model—keeping logic manufacturing, memory and packaging under one roof—at a time when TSMC's leading-edge capacity is fully booked and prices are climbing.
Read together, the two deals reflect a rivalry taking shape in AI computing. SK hynix, the current leader in high-bandwidth memory, is deepening its alignment with NVIDIA and its graphics processors. Samsung, still closing the memory gap, has attached itself to Broadcom, whose custom accelerators power data centres at hyperscalers including Google and Meta. Korea's two memory makers have effectively positioned themselves in opposite camps of the AI computing race, with memory as the common thread: demand for cutting-edge HBM outpaces industry supply, and Samsung's output for this year is already committed.
The announcements formed part of a broader Korea-US semiconductor initiative tied to President Lee Jae Myung's visit to the United States, alongside parallel commitments from other Korean companies including NAVER and Hyundai. Seoul's signal was unmistakable: it intends to supply the American AI expansion rather than observe it from the sidelines.
Yet the figures demand careful reading. The SK-NVIDIA partnership rests on letters of intent, while the Samsung-Broadcom pact is a memorandum of understanding. Neither constitutes a binding contract, and the companies disclosed no committed investment, project timeline or commercial terms. The dollar totals represent expected value over five years, not money spent or firm orders. NVIDIA's shares were little changed following the news.
Execution poses the greater question, particularly for Samsung. The Broadcom collaboration will only convert to revenue once Samsung passes customer qualification, lifts its 2nm yields to stable production levels and demonstrates capability in mass-producing the advanced packaging the deal requires. While yields have reportedly improved, they continue to hover near the industry's floor for dependable production.
What emerged from the week were two of the year's largest chip announcements and a clear statement of intent. Whether the headline valuations materialize will depend on binding contracts, manufacturing capacity and yields still being resolved.