The Manhattan AI Trinity Reshaping Global Chip Power

The table arrangement told the story

At a Manhattan ballroom on September 28, the seating chart said everything. Jensen Huang, the NVIDIA CEO, took the head table flanked by Lee Jae-yong of Samsung and Choi Tae-won of SK Group—the three men most responsible for the physical infrastructure of the AI era. This was not a casual celebrity encounter. It was a public ritual marking the formalization of what the industry has been quietly building for two years.

Huang received the Van Fleet Award from the Korea Society, an honorary gesture rooted in diplomatic goodwill. But the substance of the evening lay in what the three men said about each other—and what they deliberately did not say about China.

The public script: personal warmth masking structural alignment

Lee opened by invoking his father’s history with the same award— Chung Ju-yung received it two decades earlier—drawing a direct line from old-economy Korean entrepreneurship to the AI age. He then named three points of personal affinity with Huang: both are committed to AI, both have daughters named Madison, and both are fanatical about the Galaxy Z Fold 8.

The anecdote landed as humor. But the underlying point was sober. Samsung and NVIDIA are now positioned as the same kind of entity—AI-first, vertically integrated, willing to gamble capital on timelines that exceed normal corporate planning cycles.

Choi’s toast focused on the memory chip partnership. His framing was precise: NVIDIA designs the brain, SK produces the memory. He noted that Huang presses him for faster turnaround on every meeting, a rhythm Choi compared to Korean work culture. The metaphor was deliberate—it naturalized the interdependence between American design and Korean manufacturing as something culturally compatible rather than structurally coercive.

Huang’s acceptance speech completed the circle. He praised Korean construction speed, explicitly referred to Korea as a factory producing intelligence alongside NVIDIA, and declared he expected to work with Korean firms for another 33 years—the length of his existing partnership. On the question of rising costs from US investment mandates, he called supply chain resilience a positive outcome.

None of this is incidental choreography. It is the architecture of a commercial alliance being presented as friendship.

What English-language desks are missing

The meeting followed the July San Francisco AI Summit by roughly two months, suggesting deliberate sequencing rather than coincidence. Yet most English coverage treated it as a diplomatic photo opportunity—a friendly gathering between Silicon Valley and Korean conglomerates. That framing misses the structural shift on display.

What happened in Manhattan was the crystallization of a trilateral dependency: American AI design, Korean memory fabrication, and the US export policy framework that governs how those chips reach certain customers. Together, these three elements form a supply chain configuration that is simultaneously more efficient and more vulnerable than its predecessor.

The vulnerability is geopolitical. Korea’s memory chip exports face increasing pressure from US export controls targeting China. Samsung and SK are caught between complying with Washington’s restrictions and maintaining revenue from Chinese buyers, who remain critical customers. Huang’s comment about supply chain resilience being positive should be read as NVIDIA’s institutional position: US-based production investment is preferable to Chinese competition, even at higher cost.

Who wins and who loses

NVIDIA wins most directly. Securing long-term memory supply from Samsung and SK at favorable terms gives it cost predictability in a market where HBM (high-bandwidth memory) capacity is the binding constraint on AI chip production. Huang’s public praise reinforces the partnership narrative at a time when competitors like AMD and Intel are scrambling to secure alternative memory suppliers.

Samsung and SK win in the near term through guaranteed demand and technology access. Both companies are investing billions in US fabrication plants under CHIPS Act incentives, and their partnership with NVIDIA validates those expenditures as strategically sound. The cost of US production is absorbed through scale and government subsidies—for now.

The losers are less visible but real. Chinese AI developers who previously accessed NVIDIA chips through indirect channels now face tighter restrictions. Korean firms that relied on Chinese market revenue to subsidize their US expansion will feel margin compression. And the broader ecosystem loses diversity: when three companies control the design-memory-assembly chain, innovation paths narrow around their specifications.

The China question, unasked

Nothing at the event addressed China directly. That silence is the most consequential element. Korea’s semiconductor industry generates substantial revenue from Chinese customers, and export controls have already forced Samsung and SK to restructure parts of their business. The trinity being formalized here operates within the US security framework—which means China is increasingly an external constraint rather than a market partner.

Huang’s comment about Korean factories building “intelligent production” alongside NVIDIA should be read as a commitment to keep advanced chipmaking outside Chinese jurisdiction. Whether that aligns with Korean economic interests over the next decade remains unresolved.

What happens next

Expect deeper joint announcements between NVIDIA and Korean memory producers through 2027. The timeline aligns with when Samsung and SK’s US fabs come online and when next-generation HBM requirements will intensify. Huang’s 33-year commitment is rhetorical, but the commercial contracts backing it are real and likely already underway.

The Van Fleet Award ceremony functioned as a public commitment device—making the alliance visible to policymakers, competitors, and markets. That visibility matters. It signals to Washington that Korean firms are aligned with US tech strategy, which strengthens their case for continued subsidies and favorable trade treatment. It also signals to Beijing that the window for alternative partnerships is narrowing.

This is not a momentary alliance. It is the institutionalization of a supply chain order—one where AI dominance flows through Manhattan boardrooms, Seoul fabrication plants, and US policy levers in a single configured loop. The question for readers outside Korea is whether that loop sustains competitive innovation or simply locks in whoever controls the nodes.

The jacket auction at 550,000 dollars suggests the evening was also a fundraising exercise. But the real transaction was far more expensive and far more consequential.