NVIDIA CEO Jensen Huang will meet the leaders of Samsung Electronics and SK Group in New York to discuss AI chips, Seoul Economic Daily reports. The meetings bring the three companies that anchor the AI accelerator supply chain — the dominant designer of AI GPUs and the world's two largest memory makers — into direct negotiation over a product category whose supply, pricing and qualification status now shape quarterly revenue for all three.

The talks come as high-bandwidth memory (HBM) remains the critical gating component for AI accelerator production. Every current-generation NVIDIA data-center GPU ships stacked with HBM from either SK hynix, Samsung or Micron, and the balance of orders among those three suppliers has become one of the most closely watched variables in the semiconductor industry. SK hynix has led HBM supply to NVIDIA to date; Samsung has pursued qualification for the newest HBM generations and a larger share of the AI memory pool. Micron, the third supplier, is not reported to be part of the New York meetings.

For Samsung, the stakes are concrete. The company's memory division has flagged HBM as its principal growth vector, and winning or expanding NVIDIA orders would load its fabrication capacity with the industry's highest-value DRAM product. For SK — whose memory arm, SK hynix, has captured the largest share of HBM shipments to NVIDIA — the agenda is defending that position as demand for AI accelerators continues to outstrip memory supply.

For NVIDIA, the calculus runs the other way. The company needs enough qualified HBM to feed its accelerator roadmap, and dual- or triple-sourcing across the Korean suppliers plus Micron is the mechanism that keeps allocation and pricing in check. A face-to-face meeting between Huang and the top leadership of both Korean groups signals that those allocation and qualification decisions have risen to the chief-executive level rather than remaining a procurement matter.

The New York setting also carries weight beyond the deal specifics. Korea's two memory giants are competing for the same pool of AI-chip orders, and simultaneous meetings with their rival's most important customer give NVIDIA maximum leverage while giving each Korean side a reason to move quickly on volume, pricing or qualification commitments. Seoul Economic Daily, which first reported the meetings, did not disclose the agenda in detail, and none of the three companies has outlined expected outcomes.

Market attention will now turn to what emerges after the meetings: expanded HBM supply agreements, a shift in the share of orders between SK hynix and Samsung, or commitments tied to future NVIDIA product generations. Any confirmed reallocation of HBM capacity would move revenue between the two Korean suppliers and tighten or loosen the supply picture for AI accelerators through the coming quarters — the variable every cloud buyer and data-center operator is watching.