Key Takeaways
U.S. firms take aim at Korean chipmakers Samsung Electronics and SK Hynix
- Samsung and SK Hynix will announce long-term memory chip supply deals with US tech companies during President Lee's San Francisco visit
- Global tech firms will also reveal strategic AI data center investments tied to the agreements
- President Lee will meet NVIDIA CEO Jensen Huang and OpenAI CEO Sam Altman as part of the summit
Samsung Electronics and SK Hynix will announce large, long-term memory chip supply contracts with US technology companies this week, according to a senior South Korean presidential official. The deals will be unveiled during President Lee Jae Myung's visit to San Francisco on July 25-26, cementing what Seoul describes as a deepening tech and economic partnership between the two nations.
Kim Yong-beom, the president's policy adviser, told reporters that the figures involved would be substantial. "We expect a lot of very large and meaningful figures to be announced," Kim said. He added that global technology companies would also reveal "strategic investments" in artificial intelligence data centers, though he did not specify amounts or partners.
Who's at the table in San Francisco?
President Lee's itinerary includes meetings with NVIDIA CEO Jensen Huang and OpenAI CEO Sam Altman. Both companies depend heavily on high-bandwidth memory chips for their AI hardware. NVIDIA's H100 and successor GPUs require HBM modules that Samsung and SK Hynix dominate, holding over 80% of the global HBM market between them.
The visit also features an AI summit with South Korean business leaders: Samsung Electronics Chairman Jay Y. Lee, SK Group Chairman Chey Tae-won, and Hyundai Motor Executive Chair Euisun Chung. The presence of Hyundai hints that discussions may extend beyond chips into broader AI infrastructure and autonomous vehicle technology.
Why these deals matter for AI supply chains
Memory chips have become a bottleneck for AI development. Training large language models and running inference at scale requires massive amounts of high-bandwidth memory. SK Hynix supplies HBM to NVIDIA, while Samsung has been racing to catch up after quality issues delayed its HBM3 qualification with major customers.
Long-term supply contracts give US tech giants predictable access to these critical components. For Samsung and SK Hynix, the deals lock in revenue and justify the billions they're investing in US manufacturing capacity under the CHIPS Act.
Related analysis on the infrastructure demands of AI expansion
The CHIPS Act backdrop
Both Korean chipmakers have already committed substantial investments in the United States. Samsung received a preliminary CHIPS Act award of $6.6 billion for its Texas chip production facilities in April 2024. SK Hynix announced a $450 million HBM advanced packaging facility in Indiana the same year.
Combined with previous commitments, the two companies have pledged over $75 billion in US semiconductor investments. The new supply deals announced this week will likely be tied to these manufacturing expansions, creating a tighter integration between Korean chipmakers and American tech customers.
| Company | CHIPS Act Award | US Facility | Key Product |
|---|---|---|---|
| Samsung | $6.6 billion | Texas | Logic chips, HBM |
| SK Hynix | $450 million | Indiana | HBM advanced packaging |
Strategic implications beyond chips
Kim's mention of AI data center investments suggests the announcements will cover more than component supply. US hyperscalers, including Google, Microsoft, Amazon, and Meta, have been expanding data center capacity at unprecedented rates to support AI workloads. Korean firms could be positioning themselves as infrastructure partners, not just suppliers.
President Lee's visit to San Francisco is a stopover en route to South America, where he will attend summits with regional leaders. The timing, sandwiching AI deals between diplomatic meetings, underscores how semiconductor supply has become a matter of national strategy rather than pure commerce.
Another example of Asian tech giants making strategic US moves
Logicity's Take
These deals are less about securing supply, which was never truly at risk, and more about signaling. For Samsung, a high-profile contract with NVIDIA or OpenAI would help rehabilitate its HBM reputation after the quality setbacks that let SK Hynix pull ahead. For US tech companies, locking in Korean suppliers reduces exposure to potential trade restrictions on Chinese memory. The real question is pricing. Long-term contracts typically come with volume discounts, so watch whether these deals compress margins for the chipmakers even as they boost revenue visibility.
What to watch for this week
The official announcements are expected Friday and Saturday. Key details to look for include contract durations, whether specific chip types like HBM3E or HBM4 are named, and which US companies are confirmed as customers. Any joint venture or co-investment structures would signal deeper partnerships beyond simple supply agreements.
Samsung and SK Hynix shares will likely react to the headline numbers. But for tech decision-makers tracking AI infrastructure costs, the more important signal is whether these deals stabilize or further tighten the HBM supply that constrains GPU availability.
Frequently Asked Questions
When will Samsung and SK Hynix announce the US chip deals?
The deals are expected to be announced during President Lee Jae Myung's San Francisco visit on July 25-26, 2026.
What type of chips are covered in the deals?
The contracts focus on memory chips, likely including high-bandwidth memory (HBM) used in AI accelerators like NVIDIA GPUs.
Which US companies are involved in the chip supply deals?
Specific customers have not been confirmed, but meetings with NVIDIA CEO Jensen Huang and OpenAI CEO Sam Altman suggest they may be involved.
How much are Samsung and SK Hynix investing in US chip production?
Combined, the two companies have committed over $75 billion in US investments, including a $6.6 billion CHIPS Act award for Samsung and $450 million for SK Hynix.
Why are memory chips important for AI development?
AI training and inference require massive data throughput. High-bandwidth memory chips enable GPUs to process data fast enough for large language models and other AI workloads.
Need Help Implementing This?
If your organization is evaluating AI infrastructure investments or tracking semiconductor supply for hardware planning, Logicity can connect you with analysts covering chip supply chains. Contact our team for briefings on memory market dynamics.
Source: Tech-Economic Times / ET
Huma Shazia
Senior AI & Tech Writer
Produced with AI assistance and reviewed by the Logicity editorial team. Learn more in our Editorial Policy.
