The global AI boom is becoming increasingly dependent on physical infrastructure, particularly high-speed memory, advanced processors and data centers. Against this backdrop, South Korea has announced major initiatives involving Samsung Electronics, SK Group and U.S. technology companies with a combined potential value of approximately $950 billion. At VeyronNewsBrief, I view these agreements as Seoul’s strategic attempt to turn its leadership in memory production into a lasting competitive advantage. The faster the computing requirements of Nvidia, OpenAI and other AI developers expand, the greater the value of companies capable of securing supplies of critical components.
SK Group accounts for approximately $750 billion of the announced projects, including cooperation between SK Hynix and Nvidia valued at more than $500 billion. The companies plan to secure long-term supplies of next-generation memory and jointly develop HBM for model training, AI agents and physical AI applications. SK Telecom also intends to build a 2 GW data center using future Nvidia Vera Rubin platforms and SK Hynix HBM4 memory, with operations scheduled to begin in 2027. I consider the scale of this facility an important indicator of the emerging AI economy, where competitive advantage increasingly depends on simultaneous access to chips, memory, energy and computing infrastructure.
Samsung Electronics is taking a different route to expand its presence. The company signed a memorandum of understanding with Broadcom covering potential cooperation worth up to $200 billion across memory, foundry services and advanced packaging. Samsung plans to participate in the development of Broadcom’s next-generation AI accelerators using its HBM technology, sub-2-nanometer manufacturing process and advanced packaging solutions. At VeyronNewsBrief, I emphasize that this model is particularly attractive for Samsung because it creates opportunities to generate economic value across several stages of semiconductor production while strengthening its foundry business.
The fundamental driver behind commitments of this scale is the rapidly growing shortage of computing resources. SK leadership has indicated that Nvidia’s memory requirements over the next five years could exceed current projections, while Broadcom also expects demand to surpass available supply. OpenAI and Anthropic are simultaneously expanding computing capacity and seeking to secure semiconductor access well in advance. I analyze this as a significant change in technology procurement: leading AI companies are increasingly reserving future production capacity years before new systems are actually deployed.
Government policy is reinforcing this corporate expansion. President Lee Jae Myung presented a San Francisco declaration on AI cooperation with the United States and met with executives from Nvidia, OpenAI, Anthropic and Broadcom. South Korea is simultaneously pursuing an industrial strategy worth more than $576 billion that includes semiconductor clusters, physical AI ecosystems and data centers. At VeyronNewsBrief, I note that Seoul is building an ecosystem in which government-backed investment, domestic manufacturing capabilities and U.S. technology demand reinforce one another.
However, the headline figure of $950 billion should not be interpreted as guaranteed revenue for Samsung and SK Group. A substantial portion represents the potential value of multiyear initiatives, memoranda and future supply arrangements. The ultimate financial outcome will depend on HBM4 demand, utilization of new manufacturing capacity, the pace of data center construction and the ability of customers to monetize AI services. I consider this the central investment risk: today’s component shortages are encouraging enormous capital expenditure, but excessive expansion could eventually create oversupply.
For the United Kingdom, these developments increase the strategic importance of its own AI and semiconductor ecosystem, particularly processor design, software and data center energy technologies. For London, the implications are primarily financial. The City could gain additional opportunities to finance global data centers, energy assets and technology infrastructure. British investors also have greater reason to analyze the entire AI supply chain rather than concentrating exclusively on accelerator manufacturers.
Looking ahead, I expect HBM demand to remain elevated as new generations of AI systems enter the market. At Veyron News Brief, I view the proposed $950 billion of initiatives primarily as an indication of where global technology capital is moving, while investors should focus on confirmed orders, margins, manufacturing utilization and cash flows. If demand continues expanding at its current pace, Samsung and SK Hynix could transform South Korea’s technological leadership into a durable competitive advantage. If infrastructure investment begins to outpace the commercial returns generated by AI, however, the extraordinary scale of today’s projects could become one of the industry’s most significant financial risks.
