By Hamza Faisal
Nvidia, Korean chip startup Rebellions and SK Telecom announced on Monday a collaboration to build South Korea's sovereign AI infrastructure, linking Korean-designed AI processors into Nvidia's data-centre platform through its NVLink Fusion technology. The deal was unveiled in Seoul at Nvidia AI Day Seoul 2026, and it is backed by South Korea's Ministry of Science and ICT.
Under the agreement, Rebellions will adopt NVLink Fusion to connect its next-generation R200 neural processing units to Nvidia's open infrastructure platform for rack-scale inference, using Nvidia's MGX rack architecture for system development. SK Telecom plans to deploy the resulting systems in its AI data centre to deliver full-stack sovereign AI, spanning infrastructure, proprietary models and public-facing services, including those under Korea's national "AI for All" project.
Korean chips plug into Nvidia's platform
NVLink Fusion is Nvidia's way of letting third-party accelerators work inside its data-centre ecosystem rather than outside it. Instead of building a separate stack around its own chips, a chipmaker can connect its processors to Nvidia's high-bandwidth, low-latency interconnect and tap into the servers, networks and software that already run on it — a shortcut that Nvidia says shortens time to market for custom-chip developers.
Rebellions is developing the R200 for low-latency, energy-efficient inference on frontier models, including multitrillion-parameter mixture-of-experts models. The company's R200 systems are expected to be ready for deployment in 2028. Rebellions' current ATOM-Max chips are already running inside several SK Telecom services, handling around 14 million requests and more than 4 billion tokens a day as of September 2026.
SK Telecom to run the systems in its AI data centres
For SK Telecom, the deal adds a Korean-made inference chip to an AI factory plan that is already heavily built on Nvidia hardware. The carrier plans a gigawatt-scale AI factory in Korea on Nvidia's DSX AI Factory architecture, with an operational target of 2027, starting on the Blackwell platform and moving later to the upcoming Vera Rubin generation. The factory is meant to serve SK Group companies, including SK hynix and SK Telecom, as well as outside customers.
The collaboration builds on earlier steps. In April 2026, SK Telecom, Rebellions and Arm signed an agreement on a joint AI inference server, and in July 2026 Nvidia and SK Group widened their relationship, with talk of a potential investment commitment of more than $500 billion — a figure the companies described as a possibility, not a signed deal. No financial terms were disclosed for Monday's announcement, and the release gives no purchase commitments or deployment dates beyond the 2028 target for the R200 systems.
A wider Korean push at AI Day Seoul
The announcement was one of several Korean partnerships unveiled at the Seoul event. Nvidia also announced physical-AI work with Hyundai Motor Group, Doosan Robotics, LG Electronics and SK hynix, and said SK Telecom, LG AI Research and Upstage are using datasets from its open Nemotron models in South Korea's government-backed race to build a homegrown foundation model, with two winners due to be chosen in early 2027.
For South Korea, the aim is to cut reliance on foreign AI technology while keeping its own chipmakers inside the dominant data-centre ecosystem. For Nvidia, the deal extends NVLink Fusion — and with it, its platform — to another country's sovereign AI ambitions, one custom chip at a time.
