Nvidia, SK Telecom and Rebellions Sign a Deal to Pair Korean AI Chips With Nvidia GPUs in One Rack
A memorandum signed in Seoul on Monday would link Rebellions' inference chips to Nvidia's NVLink, with SK Telecom deploying the racks, and the government hopes it becomes a template for others.
Outspoken Digest Technology Desk
Monday, November 9, 2026/2 min read
South Korea's Ministry of Science and ICT announced on Monday that Nvidia, SK Telecom and the chip startup Rebellions have signed a memorandum of understanding to combine Korean-designed neural processing units with Nvidia GPUs on Nvidia's NVLink Fusion platform, according to The Korea Herald. The signing took place at Nvidia's AI Day Seoul event at Coex, and the hardware is meant for an AI data centre SK Telecom is building.
It is a small paper step with a large ambition: giving a home-grown chip maker a seat inside Nvidia's own architecture.
What exactly did the three companies agree to?
Per the joint release carried by StockTitan, Rebellions plans to connect its next-generation Rebel R200 neural processing units to Nvidia's platform for inference workloads, linking the R200 chips within a rack using NVLink. Nvidia's MGX rack architecture would guide system development. The companies describe a division of labour in which Nvidia GPUs handle prefill, the processing of the input, while Rebellions' NPUs handle decode, the generation of the output, aiming to improve throughput and power efficiency. That is a planned approach, not a demonstrated one.
The R200 systems are expected to be ready for deployment in 2028. SK Telecom plans to deploy them in its AI data centre and connect them to its own models and public services, including under Korea's "AI for All" project, the release says. The release gives no price, chip volume or power figure.
Why does the Korean government care?
The ministry sees a template. The Korea Herald reports that it wants to extend this partnership model to other domestic NPU makers and is pursuing similar links with AMD. On the same day it held an AI start-up partnership forum with Nvidia, attended by 22 Korean companies, and pledged support for follow-up work including joint research, commercialisation and overseas sales.
The logic, as the ministry frames it, is that access to Nvidia's platform lets local chip designers go beyond selling standalone products, and that Rebellions gets to test its chips in Nvidia's data centre design and look for overseas customers. Inference chips are in demand because they aim to cut running costs, and the ministry argues that mixing NPUs and GPUs can help data centres balance performance and power use. Korea's Deputy Prime Minister and science minister, Bae Kyung-hoon, said in the release that the government will actively support the collaboration.
What should we watch next?
Two things. First, whether other Korean chip firms announce similar arrangements, as the ministry hopes. Second, whether 2028 holds: an MOU commits no money, and the gap between a signed memorandum and a rack that runs production traffic is long. For Nvidia, the arrangement keeps Korean silicon inside its ecosystem rather than beside it. For Rebellions, it is a chance to be tested at scale.
Published in The Outspoken Digest
Editorial desk
Outspoken Digest Technology DeskSoftware, hardware, artificial intelligence and what they change for everyone else.
Newsletter
The Digest, in your inbox
One edition, sent when it is ready. No noise, and your address is never passed on.
Read Next
More Technology →
Starship Flight 15 Is Set for Friday With the First Attempt to Catch the Ship, and Vulcan Returns the Same Day
Nov 9, 2026/2 min read
Researchers Find 1,465 Exposed Artex AI Hacking Consoles Online, and Microsoft Ties the Tool to the ConnectWise Breach
Nov 8, 2026/2 min read
Astronauts Install the Space Station's Eighth and Final Roll-Out Solar Array in a 7-Hour Spacewalk
Nov 6, 2026/2 min read
Zipline Raises $1.5 Billion at a $21.5 Billion Valuation as Walmart Commits to 300 More Drone Delivery Sites
Nov 6, 2026/2 min read