SK Group and Nvidia unveiled a partnership worth more than $500 billion on the sidelines of Friday's Korea-US technology summit, with the headline commitment a 2-gigawatt data center to be built by SK Telecom.
What gets built
The facility is due online in 2027, powered by Nvidia's next-generation Vera Rubin compute platform and fed by SK hynix HBM4 memory. At 2 gigawatts it ranks among the largest AI data centers any single operator has committed to build, and it places Korean memory directly inside Nvidia's forward roadmap rather than in the spot market.
Beyond the data center
Nvidia used the same day to commit 50,000 Blackwell GPUs to Hyundai Motor Group for its DRIVE Hyperion autonomous-driving platform, and to open a joint agentic-AI laboratory with KAIST — described as the first joint AI lab between a Korean university and a global technology company.
What was not disclosed
No memory-specific dollar figure, no GPU volume, no HBM4 base-die specification and no exclusivity term appears in any of the announcements. The two sides also signed letters of intent to co-develop next-generation AI memory — an intent to collaborate, not a priced supply contract, and the distinction will matter when the first invoices are written.
How it fits the bigger number
This $500 billion is not additional to the $950 billion package announced at the summit. It sits inside SK Group's $750 billion memory commitment. Reporting that sums the two produces a figure that does not exist.
A dating trap
Nvidia's corporate blog post covering the Korean partnerships carries a July 23 publication date with updates running through late on July 24 — it is a rolling page, not evidence the deals were announced a day early. Reuters, Nikkei and Korean outlets all place the announcements on July 24.