Nvidia, Rebellions and SK Telecom Join Forces on Korea’s Sovereign AI Push
Nvidia, South Korean chipmaker Rebellions and SK Telecom have announced a collaboration to build sovereign artificial intelligence infrastructure in South Korea, pairing Korean-designed processors with Nvidia’s computing platform in a deal backed by the country’s government.
The agreement was unveiled on Monday at Nvidia AI Day in Seoul, held at the COEX convention centre, and centres on Rebellions’ next-generation Rebel R200 neural processing units, or NPUs. The chips will be connected to Nvidia’s open infrastructure platform through NVLink Fusion, the company’s interconnect technology, and built into rack-scale systems using Nvidia’s MGX server design.
SK Telecom plans to deploy the combined systems in its AI data centre, drawing on its own foundation models, including A.X K2, and public services developed under South Korea’s “AI for All” programme. The project is supported by the Ministry of Science and ICT, which aims to connect domestic semiconductor development with the infrastructure needed to deliver AI at scale.
What “sovereign AI” means for Seoul
Sovereign AI has become shorthand for computing infrastructure, models and services that a country can shape around its own language, industries and priorities, rather than relying entirely on foreign systems. For South Korea, that means AI that works fluently in Korean, serves local companies and public services, and runs on hardware in which domestic firms have a stake.
Rebellions, founded in Seoul, designs AI chips for inference — the stage at which a trained model answers questions and generates output. Its R200 is its next-generation processor, and the Nvidia collaboration is its most prominent international tie-up to date. By linking the R200 to Nvidia’s platform, the partners intend to split inference work between Nvidia’s graphics processors and Rebellions’ NPUs, a division they say should improve output capacity and energy efficiency.
The collaboration also forms part of a wider agreement between Nvidia and the Ministry of Science and ICT covering chips, data centres, models, skills and agent security. Under the “AI for All” programme, Nvidia is working with three consortia and four Korean foundation-model builders, including SK Telecom’s A.X K2, LG AI Research’s K-EXAONE 2.0, Upstage’s Solar Open 2 and Motif Technologies’ Motif-3, with deployment recipes to help developers run those models at scale.
A 2028 target, with the details still to come
The announcement sets a timetable rather than a finished product. Rebellions aims to have its R200 systems ready for deployment in 2028, and the partners have not said how many systems SK Telecom will install, or given a firm date for live service. Engineering, integration and deployment all remain ahead.
No financial terms were disclosed, and the companies published no measured performance figures or order values. The memorandum of understanding signed by the three companies at the Seoul event sets out the plan but leaves the commercial scale of the project open.
Rebellions chief executive Park Sung-hyun presented the plan at the Seoul event, where the company said its chip would anchor rack-scale inference systems built on Nvidia’s MGX design and run in SK Telecom’s data centres.
Nvidia’s week in the spotlight
The Seoul announcement comes during a turbulent stretch for Nvidia. The company’s board opened an independent investigation into Jensen Huang after reports that the chief executive had been targeted by a foreign intelligence effort, a story that has drawn calls in Congress for further scrutiny. Huang has told staff he will cooperate with the inquiry.
Markets have been volatile too. The Nasdaq plunged 5.5 percent into correction territory on November 5 as investors questioned AI spending, then the Nasdaq bounced 1.1 percent on Friday as chip stocks recovered. The Seoul deal lands in that unsettled mood, offering a concrete example of the demand Nvidia says still underpins its business: governments and telecoms operators building AI infrastructure of their own.
Nvidia has also faced continued scrutiny over its China business, including questions about who ultimately buys its most powerful chips. That backdrop makes the Korean project a useful counterpoint for the company — a partnership built openly with a government ministry and a major telecoms operator, around a product designed in the country where it will run.
What happens next
The test for the collaboration will be delivery. Rebellions must finish the R200 and prove it works efficiently alongside Nvidia’s GPUs in a live data centre, while SK Telecom must turn the combined systems into services that Korean businesses and public bodies actually use.
If it succeeds, South Korea will have a template other countries are watching closely: a sovereign AI stack built from a mix of domestic and foreign technology, rather than a single supplier’s platform. If the 2028 timetable slips, it will join a long list of AI infrastructure announcements whose ambitions outran their engineering. For now, the three partners have put their names to a plan — and the clock is running.
