Korean chip startup Rebellions will connect its Rebel R200 neural processing units to Nvidia’s platform through NVLink Fusion, and SK Telecom plans to deploy the systems in its AI data centre, Nvidiaannounced on 9 November. The companies plan to use Nvidia GPUs for prefill and Rebellions NPUs for decode, two stages of generating an AI response.
Nvidia says Rebellions’ R200 systems are “expected to be ready for deployment in 2028”. The release says Korea’s Ministry of Science and ICT is supporting collaboration between Nvidia and Korean NPU companies, “beginning with Rebellions and SK Telecom”.
Business Pill · Sovereign AI Explained
An evergreen Business Engineer explainer on sovereign AI. It teaches the general idea only and does not discuss this announcement.
The key insight: Nvidia says Korean chip startup Rebellions will link its R200 NPUs to Nvidia’s platform through NVLink Fusion, with Nvidia GPUs planned for prefill, Rebellions NPUs for decode and SK Telecom as the planned operator; the R200 systems are expected to be ready for deployment in 2028.
Under the agreement, Rebellions will use NVLink Fusion, Nvidia’s interconnect for linking other companies’ chips to its systems, to connect R200 NPUs within a rack, and will build on Nvidia’s MGX rack architecture, the press release says. An NPU is a processor built for AI calculations; Rebellions says the R200 is optimised for low-latency, energy-efficient inference, including multitrillion-parameter mixture-of-experts models.
SK Telecom plans to deploy the systems in its AI data centre and connect them with its own foundation models, including A.X K2, and with public services under the government’s “AI for All” project. The release says the MGX design lets GPU and NPU systems share rack footprint, power delivery and cooling.
The planned split of one AI response between Nvidia GPUs (prefill) and Rebellions R200 NPUs (decode), linked by NVLink Fusion and deployed by SK Telecom, as described in the companies’ press release of 9 November 2026. Diagram by FourWeekMBA; no chip counts, prices or power figures were given.
How the Work Is Split
The companies plan “disaggregated inference”, which the release describes as assigning different stages of AI response generation to the processor best suited to each task. Prefill reads the prompt; decode then produces the answer token by token. Here Nvidia GPUs would do the first and Rebellions NPUs the second, “to improve inference throughput and power efficiency”, the release says.
Rebellions lists eight backers in the release: Aramco, Arm, Kindred Ventures, KT, Mirae Asset Group, Samsung, SK Hynix and SK Telecom. That makes SK Telecom both an investor in Rebellions and the planned operator of the systems, by our reading of the release.
What Korea’s Minister Said
Korea’s Deputy Prime Minister and Minister of Science and ICT, Bae Kyung-hoon, said the collaboration “enables Korean-developed NPUs to enter the computing infrastructure market alongside NVIDIA”. The minister added that the government will “further advance cooperation with NVIDIA across a range of areas, including physical AI and securing GPU supply”.
Nvidia’s founder and chief executive said in the release that “With NVLink Fusion, Korea’s chipmakers can connect their custom ASICs to NVIDIA’s AI factory platform”. Rebellions co-founder and CEO Sunghyun Park called the announcement “a turning point”, and SK Telecom CEO Jung Jai-hun said it “validates the potential of next-generation sovereign AI infrastructure”.
Connecting the Coverage
Earlier on 9 November, Korea Says Nvidia GPU Deliveries on Track, 10,000 in Use reported, citing Reuters, that Rebellions said a planned visit by an Nvidia senior director to its headquarters had been cancelled because of internal circumstances at Nvidia. That piece also set out the more than 260,000 Nvidia GPUs announced for Korea in October 2025. Nvidia’s company-level context this week is in Nvidia Board Opens Independent Probe 9 Days Before Results.
As we read the release, the split gives each chip one stage of the same answer: Nvidia GPUs read the prompt and the Korean NPUs generate the reply, so neither replaces the other.
As we read the backer list, SK Telecom sits on both sides of the plan, as one of eight named investors in Rebellions and as the company that plans to deploy the systems.
As we read the minister’s remarks, the government ties the NPU work to the wider Nvidia relationship, naming physical AI and securing GPU supply as further areas of cooperation.
Bae Kyung-hoon, Korea Deputy Prime Minister and Minister of Science and ICT, in the press release
“enables Korean-developed NPUs to enter the computing infrastructure market alongside NVIDIA”
Three Implications
FOR REBELLIONS The release says the MGX ecosystem gives Rebellions access to an established supply chain of original design manufacturers and component suppliers.
FOR SK TELECOM SK Telecom says it will link the systems with its A.X K2 models and services developed under the ‘AI for All’ project.
FOR DATA-CENTRE DESIGN The release says the MGX rack architecture lets GPU and NPU systems share rack footprint, power delivery and cooling.
The framework says: “Inference is a revenue engine: continuous, distributed across millions of applications and agents, incurring marginal cost with every query.”
As we read it, a marginal cost on every query is why the companies frame the split around inference throughput and power efficiency. The framework predates this announcement and does not discuss Nvidia, Rebellions or SK Telecom.
What Is Not Established
The release gives no price, chip volume, rack count or power figure for the systems, and no deployment site beyond SK Telecom’s AI data centre. It does not say how many R200 chips will be paired with how many Nvidia GPUs, or which Nvidia GPU generation. The 2028 date is Nvidia’s expectation for Rebellions’ systems. We did not contact Nvidia, Rebellions or SK Telecom.
This Week In The AI Supercycle
The Great Convergence of Capital, Power, and Geopolitics
This week’s issue: for two years the AI build-out was priced as a supply story, and last week as a credit story. This week it was priced as a permission story, and the answer arrived as a cost, not a ban.
Nvidia, Rebellions and SK Telecom plan rack-scale systems in which Nvidia GPUs handle prefill and Korean-designed R200 NPUs handle decode, linked by NVLink Fusion and expected to be ready for deployment in 2028. Korea’s science minister described it as Korean NPUs entering the market alongside Nvidia.
A note on sourcing. The plans, dates and quotes are from the NVIDIA, Rebellions and SK Telecom press release and an NVIDIA blog post, both published on 9 November 2026. The count of backers and the reading of SK Telecom’s two roles are ours. We did not contact the companies.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
