TodayToday's edition
Seoul's vice science minister told parliament on Nov. 9 that 2026 GPU orders stay on schedule despite the NVIDIA CEO probe, though the probe's impact is uncertain and purchases go through server suppliers.
Chips & Power DeskEdited by Ege Erten
2 min readAI-assisted
NVIDIA data-center hardware image from the company's Nov. 8 post on its sovereign AI collaboration with Korea's science ministry; it is illustrative, not the GPUs delivered to Korea.
Image: NVIDIA
Source · Press image from the NVIDIA blog, used for news reporting.
South Korea's Vice Minister of Science and ICT, Ryu Je-myung, said on Nov. 9 that 10,000 NVIDIA GPUs bought under a 2025 supplementary budget have arrived and are in use, and that deliveries of contracted volumes should continue without disruption.
Ryu made the statement to a parliamentary budget committee, according to Reuters. GPUs planned for 2026 are under contract and expected to enter service this year, he said, and the ministry is monitoring developments while discussing prices and schedules with NVIDIA Korea. "We expect production and the delivery schedule for contracted volumes to proceed without disruption," Ryu said, per Reuters.
The Korea Herald added that Ryu called the impact of the investigation into NVIDIA uncertain, and that deliveries under this year's orders are already underway. The government buys the processors through server suppliers such as Hewlett Packard rather than directly from NVIDIA, the paper said.
Of the 260,000 GPUs announced on Oct. 31, 2025, only the government's 10,000 are confirmed in use; installed counts for the rest are not confirmed.
Illustration (AI-generated)
The 10,000 GPUs are the first slice of a much larger program. On Oct. 31, 2025, NVIDIA said it would supply 260,000 GPUs to the Korean government and four conglomerates, the Korea Herald reported at the time: 50,000 Blackwell GPUs for the government, 50,000 each for Samsung, SK and Hyundai, and 60,000 for Naver Cloud. The 10,000 now in use equal 20% of the government's 50,000 and about 3.8% of the 260,000 total. Voltgrid has seen no official count of how many of the other 250,000 have been installed.
For data-center planners, the point is timing. GPUs that arrive on schedule keep Korean AI capacity builds on their power and cooling timelines, while a delay would leave committed racks and grid connections waiting for hardware. Ryu's statement covers only government-contracted volumes, not the four conglomerates' orders.
NVIDIA and Korea's Ministry of Science and ICT went ahead with their joint events. On Nov. 8 NVIDIA said it was expanding sovereign AI collaboration with the ministry, including a path for Korean NPU and XPU makers to join its NVLink Fusion ecosystem, as Voltgrid reported for Rebellions and SK Telecom. On Nov. 9 NVIDIA said at the MSIT-NVIDIA AI Startup Partnership Forum that it had invested in Korean interconnect startup Point2 Technology and planned investments in Motif Technologies, RLWRLD and Upstage. Ryu attended the forum.
The remarks follow NVIDIA's board opening an independent investigation on Nov. 8 into reports about CEO Jensen Huang, as Voltgrid reported, and Huang's Nov. 9 email to staff, which we covered in our latest story. None of the reported allegations has been established, and the probe has not reported any findings.
Reuters said NVIDIA did not immediately respond to a request for comment on the Korean statement. Open questions remain: whether US authorities tighten export oversight, which the Korea Herald says could raise compliance costs for Samsung and SK hynix, and whether three US House committee chairs get the closed-door briefing they requested for Nov. 18.
