Global competition in artificial intelligence is entering a phase in which access to advanced memory is becoming as important as the availability of powerful AI accelerators. South Korea’s Samsung Electronics and SK Group are pursuing large scale partnerships with major U.S. technology companies, including Nvidia and Broadcom, with a combined potential value of approximately $950 billion. At VeyronNewsBrief, I view these agreements as evidence of a structural transformation in the semiconductor market: AI development increasingly depends on manufacturers’ ability to supply data centers with sufficient volumes of high speed memory and computing infrastructure.
Approximately $750 billion is linked to long term SK Group projects with U.S. technology companies, including Nvidia. SK Hynix, one of the world’s leading producers of high bandwidth memory, is positioned at the center of these plans. The companies are also expected to develop next generation infrastructure involving future Nvidia Vera Rubin platforms and HBM4 memory. I believe this integration strengthens SK Hynix’s position in one of the fastest growing segments of the memory market, where supply reliability and technological leadership are becoming critical requirements for AI system manufacturers.
Samsung Electronics, meanwhile, is expanding its cooperation with Broadcom in projects potentially valued at approximately $200 billion through 2030. The partnership covers memory supplies, semiconductor foundry manufacturing and advanced packaging technologies. At VeyronNewsBrief, I emphasize that the agreement has strategic value for Samsung on two fronts. It provides an opportunity to strengthen the company’s presence in AI memory while increasing utilization of its foundry operations, which continue competing for major orders to manufacture advanced processors.
The rapid growth of HBM demand is particularly significant. Modern AI models require enormous quantities of data to move between processors and memory at extremely high speeds, making bandwidth one of the principal constraints on computing infrastructure. I analyze this trend as a fundamental change in the traditional memory cycle. If construction of AI data centers continues at the current pace, HBM manufacturers could maintain high factory utilization and stronger pricing power for considerably longer than during previous DRAM cycles.
However, the headline figure of $950 billion should not be interpreted as guaranteed revenue already secured by Samsung and SK Hynix. It represents the combined potential value of long term projects, some of which will unfold over several years and remain dependent on continued expansion of AI infrastructure. At VeyronNewsBrief, I note that investors should distinguish between the strategic value of these agreements and confirmed commercial orders. Actual HBM shipment volumes, customer capital expenditure, manufacturing utilization and contract margins will provide a more accurate measure of their financial impact.
For Nvidia, closer cooperation provides an opportunity to secure access to one of the most critical components required for future generations of AI accelerators. For Broadcom, reliable Samsung capacity could support the expansion of customized AI chips developed for major cloud computing companies. I see this as part of a broader shift toward a model in which technology corporations reserve manufacturing capacity years in advance. Building new semiconductor fabrication and advanced packaging facilities takes considerable time, making supply chain security a competitive advantage in its own right.
For the United Kingdom and London, the implications are primarily connected to investment flows and the expansion of domestic AI infrastructure. British funds have an additional reason to look beyond processor manufacturers and assess opportunities across memory, data centers, energy infrastructure and networking equipment. For London as a global financial center, projects of this scale could generate additional demand for debt and infrastructure financing. At the same time, Britain’s semiconductor ecosystem may benefit indirectly from rising global demand for architectures and technologies supporting increasingly complex AI systems.
Looking ahead, I expect demand for HBM to remain elevated as new generations of AI accelerators reach the market. However, the risk of excessive investment cannot be ignored. If data center construction slows, aggressive capacity expansion could eventually restore pricing pressure across the memory industry. At Veyron News Brief, I view the proposed $950 billion in partnerships primarily as an indicator of where global technology capital is moving. For investors, confirmed orders and sustainable cash flows should remain the central benchmarks, as these will ultimately determine whether the extraordinary AI ambitions of Samsung, SK Hynix and their U.S. partners translate into durable profitability.
