Korea has locked down nearly a trillion dollars in chip partnerships during President Lee Jae-myung's visit to San Francisco, signing agreements worth a combined $950 billion with Nvidia, Broadcom, and other global tech firms. The deals, announced July 25 alongside Lee's "San Francisco Declaration on AI" speech, cement the country's role as the dominant manufacturing hub for advanced AI semiconductors and memory chips.
Samsung and Broadcom Ink $200 Billion Foundry Pact
Samsung Electronics signed a memorandum of understanding with Broadcom worth $200 billion through 2030, covering memory chips, foundry manufacturing, and advanced semiconductor packaging. The deal marks one of the largest single foundry commitments in the industry's history and positions Samsung's foundry business to take on TSMC for AI chip orders.
The partnership spans logic chip fabrication, high-bandwidth memory integration, and advanced packaging. Samsung has been investing heavily in its 3nm gate-all-around (GAA) process and is developing a 2nm node to win external foundry customers. Broadcom, which designs custom AI accelerators for hyperscalers like Google and Meta, had been evaluating foundry partners beyond TSMC for redundancy and capacity reasons. The current supply chain concentration around a single foundry in Taiwan has been a growing concern for US tech giants.
The two companies also agreed to cooperate on 2.5D and 3D stacking technologies that boost performance by placing memory and logic chips closer together. Samsung's advanced packaging group in Cheonan, South Korea, has been expanding capacity to handle rising demand for chiplet-based designs and heterogeneous integration.
SK Group Commits $750 Billion to Nvidia
The biggest single commitment came from SK Group, which signed long-term supply agreements worth $750 billion over five years. The package includes a partnership between SK hynix and Nvidia valued at more than $500 billion, covering high-bandwidth memory (HBM) chips for next-generation AI systems.
SK hynix currently dominates the HBM market with its 12-layer HBM3E memory, which is critical for Nvidia's Blackwell and Vera Rubin GPU platforms. The new deal locks in pricing and volume commitments through 2030 and includes joint development of HBM4, the next-generation memory standard that promises 50% higher bandwidth than HBM3E. SK Telecom separately plans to build a 2-gigawatt AI data center using Nvidia's Vera Rubin processors and SK hynix's HBM4 memory — a project valued at roughly $30 billion in infrastructure spending alone.
These long-term contracts give chipmakers the revenue visibility they need to justify the enormous cost of building leading-edge fabrication plants — currently running $20 billion to $30 billion per facility. Most of these agreements are designed to prevent a repeat of the pandemic-era chip shortage that squeezed automakers and electronics manufacturers for two years.
Korea's Strategy: AI Test Bed and Production Platform
President Lee Jae-myung presented Korea's vision as "a global AI test bed and production platform" during his San Francisco summit, moving beyond chip fabrication into system-level AI deployment. The government is backing this push with tax incentives, R&D subsidies, and streamlined permitting for semiconductor facilities.
"Korea aims to become an irreplaceable pillar of the global AI supply chain," Lee said. "Building on our world-class competitiveness in memory chips and related production capacity, Korea will serve as a trusted manufacturing base for AI semiconductors and as a reliable supply chain partner."
The push comes as the US-China technology war reshapes global chip supply chains. Korea is positioning itself as a neutral manufacturing hub that can serve customers on both sides of the divide, contrasting with Taiwan, which faces direct geopolitical risks from China's territorial claims. South Korea is the only major chip-producing nation with active free trade agreements with both Washington and Beijing.
Naver also announced plans to establish $10 billion worth of AI factories worldwide through a strategic investment agreement with Nvidia and an infrastructure partnership with Canadian asset manager Brookfield. The initiative will build gigawatt-scale AI computing facilities across North America and Asia. Hyundai Motor Group presented its strategy to integrate AI into manufacturing and mobility, expanding the use of robotics, autonomous production systems, and digital twins in its factories.
The Bigger Picture — A $1 Trillion Semiconductor Market
The global semiconductor market is on track to pass $1 trillion in 2026, according to the Semiconductor Industry Association. AI accelerators, high-bandwidth memory, and networking chips account for roughly 40% of that total, up from less than 10% just five years ago.
Korea's strategy is built on two structural advantages. First, it dominates memory chips — Samsung and SK hynix together control over 70% of the global DRAM and NAND markets. Second, its geographic and diplomatic position as a tech-savvy democracy with deep trade ties to both the US and China makes it a rare safe bet in an increasingly polarized tech world.
Korea's chip exports hit a record $16.7 billion in June 2026, and the government expects semiconductor-related capital investment to exceed $100 billion this year. The $950 billion in partnership agreements announced in San Francisco are mostly long-term supply contracts rather than equity investments, but they provide the demand certainty that fabs need to plan capacity expansions four to five years out.
Tower Semiconductor also announced a $3 billion expansion in Japan to boost AI chip production capacity, while Nvidia signed a $1.5 billion packaging deal with Amkor in Arizona. The broader industry is racing to build geographic redundancy into its supply chain.
Geopolitical Context and Cybersecurity
The semiconductor buildout has drawn attention from cybersecurity agencies. Korea's National Intelligence Service warned in a recent briefing that state-backed hackers have been aggressively targeting chip design files and manufacturing recipes in a campaign dubbed "Silicon Heist." The NIS advised foundries to air-gap sensitive process data and implement hardware-level security modules in fabrication equipment.
Geopolitically, Korea's chip diplomacy walks a narrow line. The US has been pressuring allies to restrict technology flows to China, while Beijing remains Korea's largest trading partner. President Lee's speech in San Francisco carefully avoided direct mention of China, instead framing Korea's role in universal terms as a "reliable partner that shares opportunities created by AI with the whole world."
The $950 billion in deals also sends a signal that Korea intends to stay ahead of China's rapidly advancing chip industry. Chinese memory maker CXMT is preparing for a public listing, and YMTC continues to develop advanced 3D NAND despite US export controls. Korea's aggressive long-term contracting is partly a preemptive move to lock in customers before Chinese alternatives gain traction.
The US government has also weighed in. The Department of Commerce recently announced a $225 million CHIPS Act award to Bosch for silicon carbide production, part of a broader push to domesticate chip manufacturing. Intel, meanwhile, posted blowout Q2 results driven by 25% revenue growth as AI server demand pushed its data center revenue up 59% to $6.3 billion.
For investors, the takeaway is clear: the AI semiconductor buildout is shifting from prototype volumes to mass production. The long-term supply agreements signed in San Francisco are the infrastructure contracts on which the next generation of AI data centers will be built. Companies that control the memory and logic supply chains — Samsung, SK hynix, Nvidia — are writing the playbook for the post-scarcity chip era.
The Korea Times reported that the deals "brought together leaders of Samsung Electronics, SK Group, Hyundai Motor Group and Naver together with Nvidia CEO Jensen Huang, OpenAI CEO Sam Altman, Broadcom CEO Hock Tan, and Anthropic representatives." Korea Times coverage