Nvidia’s Seoul AI Day: 260,000 GPUs, Huang Investigation

By Amara Osei

More than 260,000 graphics processing units (GPUs) form the core of South Korea’s deployment plan, announced by Nvidia in October last year, as allegations concerning Chief Executive Jensen Huang drew attention away from cooperation announcements at the company’s second annual AI Day in Seoul.

On the 8th, Nvidia’s board launched an independent investigation into reports that a foreign intelligence service had targeted Huang. The board said there was “no basis to conclude that national security or Nvidia’s confidential information had been compromised.”

What did Nvidia agree in Seoul?

On the 9th, Nvidia signed a memorandum of understanding on NVLink Fusion cooperation with SK Telecom (017670), Rebellions and the Ministry of Science and ICT. A separate agreement to link domestically developed AI chips with Nvidia systems was signed as scheduled this year.

The October announcement covered more than 260,000 GPUs for deployment by the Korean government and Korean companies. Attention at the event centered less on the new agreements than on the delivery timetable for that supply plan and the high-bandwidth memory chain supporting it.

Takeaway: The Seoul agreements put cooperation on the record, while the deployment timetable remained the central market question.

Why did questions turn to Jensen Huang?

Allegations about Huang’s private life surfaced around the event, including allegations of an affair with a figure linked to Chinese intelligence. Madison Huang, the CEO’s eldest daughter and Nvidia’s senior director of product marketing for Omniverse and robotics, canceled her planned trip to Korea.

At the roundtable on the 9th, held immediately after the NVLink Fusion signing, Nvidia offered little comment on the allegations. Jesse Clayton, senior manager for AI infrastructure, and other company officials answered only pre-submitted questions. Reporters objected to that format, and Nvidia agreed to take questions from the floor.

One reporter asked why Madison Huang had not traveled to Korea and whether uncertainty over the CEO’s position could disrupt GPU deployments in South Korea or Nvidia’s cooperation on HBM. An Nvidia official said the question “was not related to the day’s event” and that company limits on its response reflected Madison Huang’s absence.

Nvidia officials at the venue declined to specify how the allegations might affect the company’s Korean business. The exchange left the leadership issue unresolved while linking it directly to deployment and memory-supply questions.

Takeaway: Nvidia allowed the allegations into the session without resolving them, leaving leadership risk attached to Korean GPU and HBM plans.

How do Samsung and SK hynix factor into HBM4?

Samsung Electronics (005930) and SK hynix (000660) have expanded memory cooperation with Nvidia, centered on HBM supplied to the chipmaker. Both companies are pursuing sixth-generation HBM, or HBM4, for Nvidia’s next-generation AI platform, Vera Rubin.

Industry watchers said Huang’s status could become a variable in next-generation HBM supply. If a successor management team reviews procurement strategy after Huang steps down, volume allocations among suppliers and joint development plans could change, according to that view.

Takeaway: Both suppliers are pursuing HBM4 for Vera Rubin, but industry watchers see procurement and allocation risk if the CEO exits.

What is Nvidia’s board investigating?

The board said it would review all available information and that it had confidence in Jensen Huang and the management team. The inquiry therefore ran alongside the deployment and HBM4 agreements presented in Seoul.

Takeaway: The board has reported no basis for a compromise finding, while its review leaves the leadership question open.