Nvidia has widened its South Korea push with fresh agreements involving SK Hynix, SK Telecom, Naver, and Doosan Group, a move that ties the chipmaker’s AI ambitions more tightly to the country that already sits near the center of advanced memory manufacturing. The deal set is aimed at boosting memory production and speeding up artificial intelligence deployments, with Jensen Huang framing SK Hynix as Nvidia’s biggest memory supply partner and hinting that demand will keep rising fast.
That is not a small bet. AI server builds are still being limited by memory availability as much as by raw accelerator supply, and South Korea’s biggest names are clearly trying to lock in a longer-term position before rivals do. Samsung remains the other heavyweight in the country, but Nvidia’s latest round of agreements shows how much leverage memory suppliers now have in the AI stack.
SK Hynix stays at the center of Nvidia’s memory supply
Huang said Nvidia is already buying billions of dollars’ worth of SK Hynix products every year, and that figure will grow substantially. The two companies have worked together for more than two years, so this is less a new romance than a very public renewal of vows.
The logic is simple: AI systems chew through memory, and the companies that can reliably ship it are suddenly strategic partners rather than just vendors. In a market where every major player wants more inference capacity, more training clusters, and more data-center scale, memory is becoming the bottleneck everyone politely complains about.
SK Telecom, Naver and Doosan build around Nvidia’s stack
SK Telecom said its plan is a gigawatt-scale cloud AI platform using Nvidia technology in South Korea, with the first AI data center based on that setup expected to open next year. Naver and Doosan are also moving in the same direction, building their own AI data centers around Nvidia’s technology.
- SK Telecom: gigawatt-scale AI cloud platform using Nvidia technology
- Naver: AI data center buildout tied to Nvidia tech
- Doosan Group: AI data center plans plus an energy solution it says can help Nvidia’s data centers
Doosan’s energy pitch is the most interesting wrinkle here. Data centers are not just silicon problems anymore; they are power, cooling, and infrastructure problems dressed up as software stories, and companies that can help on the utility side get a seat at the same table as the chip vendors.
Why South Korea matters in Nvidia’s AI expansion
South Korea is already one of the world’s most important electronics and memory manufacturing hubs, with SK Hynix and Samsung dominating the conversation around high-end memory. Nvidia’s latest agreements lean into that reality rather than pretending the AI boom can be built in isolation, which is the sort of practical move that tends to win when supply chains get tight.
The bigger question is how quickly these partnerships turn into capacity. If the promised data centers come online on schedule, Nvidia gets more demand locked into its ecosystem while its Korean partners gain a clearer role in the AI buildout. If not, the announcements still serve a purpose: reminding everyone else that the memory side of AI is now just as strategically contested as the chips themselves.