Technology · AI
Nvidia, Samsung, and SK Group Chiefs Convene at AI Summit in San Francisco
Jensen Huang's joint appearance with Korean semiconductor leaders signals intensifying collaboration across the AI hardware ecosystem
KEY TAKEAWAYS
- ·Nvidia CEO Jensen Huang appeared with Samsung Electronics Executive Chairman Lee Jae-yong and SK Group Chairman Chey Tae-won at an AI industry event in San Francisco.
- ·The meeting underscores the strategic importance of Korean memory suppliers in the AI infrastructure race, particularly for high-bandwidth memory used in GPU clusters.
- ·Both Samsung and SK hynix are investing tens of billions in HBM production capacity and competing for engineering talent as AI demand reshapes the semiconductor sector.
High-Stakes Gathering
Nvidia CEO Jensen Huang shared the stage with two of South Korea's most influential technology executives at an AI industry gathering in San Francisco, a public display of alignment that reflects the interdependence shaping the semiconductor sector's AI push.
Samsung Electronics Executive Chairman Lee Jae-yong and SK Group Chairman Chey Tae-won joined Huang at the event, according to Samsung. The trio's appearance together marks a rare convergence of leadership from companies competing and collaborating across memory chips, AI accelerators, and data center infrastructure.
Asia's Memory Makers in the AI Race
The meeting comes as both Korean conglomerates work to secure their positions in the high-bandwidth memory market, a critical bottleneck for AI training systems. SK hynix has established an early lead in HBM3E supply to Nvidia, shipping advanced memory modules that pair with the chipmaker's H100 and H200 GPUs. Samsung has been working to close the performance and qualification gap, targeting volume shipments later this year.
Nvidia's reliance on these suppliers has grown in lockstep with demand for generative AI infrastructure. Hyperscalers in North America and Asia are ordering thousands of GPU clusters, each requiring terabytes of high-speed memory. The supply of HBM3E, which offers more than double the bandwidth of previous generations, remains constrained, giving leverage to the handful of manufacturers capable of producing it at scale.
Strategic Alignment Beyond Components
The San Francisco event signals more than transactional supplier relationships. Nvidia has deepened technical collaboration with both Korean firms, sharing roadmaps and co-engineering next-generation memory architectures. SK hynix has publicly discussed joint development efforts with Nvidia on HBM4, expected to enter production in late 2027. Samsung, meanwhile, has reorganized its memory division to prioritize AI-optimized products and has allocated additional cleanroom capacity to advanced packaging lines.
For Lee and Chey, the appearance alongside Huang offers a visible endorsement in a market where certification timelines and technical partnerships carry significant weight. Nvidia's validation of a memory supplier can determine whether that supplier captures a majority share of a product cycle or remains a secondary source.
Capital and Capacity Commitments
Both Samsung and SK hynix have committed tens of billions of dollars to memory fabrication and packaging expansion. SK hynix is building a new HBM packaging facility in South Korea, while Samsung has accelerated investment in its Pyeongtaek and Hwaseong campuses. The capital intensity reflects not only the technical complexity of stacking DRAM dies with sub-micron precision but also the expectation that AI infrastructure spending will sustain elevated memory demand through the end of the decade.
The companies are also navigating export controls and geopolitical constraints that shape where they can build capacity and whom they can serve. U.S. restrictions on advanced chip exports to China have reordered supply chains, pushing memory makers to prioritize customers in North America, Europe, and parts of Asia while managing legacy commitments in restricted markets.
Competitive Dynamics and Talent Flows
The race for AI memory leadership has intensified competition for engineering talent. SK hynix has offered performance bonuses tied to HBM shipment milestones, creating compensation disparities that have drawn attention from Samsung engineers. Both companies are expanding headcount in packaging, yield engineering, and failure analysis, roles that command premium salaries in a tight labor market.
Industry observers note that the public solidarity displayed in San Francisco may coexist with aggressive talent recruitment and IP protection efforts behind closed doors. The Korean semiconductor industry has a history of high-profile defections and legal battles over trade secrets, dynamics that have accelerated as memory technology becomes more central to AI competitiveness.
What Comes Next
The convergence of Nvidia, Samsung, and SK leadership at a single event reflects the consolidated nature of the AI hardware supply chain. With only a few companies capable of producing cutting-edge GPUs and matching them with the memory performance AI models require, the strategic relationships formed now will shape infrastructure deployment for years.
As Nvidia prepares its next-generation Blackwell architecture and Korean chipmakers ramp HBM3E and begin pilot production of HBM4, the partnerships on display in San Francisco will be tested by execution timelines, yield rates, and the ability to meet surging orders from cloud providers and enterprises building out AI capacity across Asia, North America, and Europe.
RELATED STORIES
Chip Rally and Meta's Cash Flow Slump Split Wall Street After Mixed Earnings Week
Priyanka V. Rao · 4 min
Seoul's KOSPI Jumps 13 Percent as Chip Stocks Rally on Microsoft Earnings
Cheng Wei-Lun · 3 min
MediaTek Raises AI Chip Target to 20% Market Share by 2027
Cheng Wei-Lun · 3 min
Mongolia Locks In Monthly Fuel Supply From South Korea
Cheng Wei-Lun · 3 min
Spot something wrong? Email editor@briefasia.com. We log every correction publicly.