South Korea's largest internet company made its name on search boxes and shopping carts. Now Naver is wagering that its future runs through electrical substations and racks of Nvidia silicon. In a deal that ranks among the most ambitious sovereign-AI plays yet announced, the Seongnam-based company has lined up roughly $10 billion in outside capital to more than triple a hyperscale AI factory outside Seoul, with a stated long-term target of one gigawatt of compute dedicated to Korean and U.S. developers.
The three-way arrangement, unveiled during Korean President Lee Jae-myung's visit to an AI summit in San Francisco, pairs Naver with two heavyweight partners: Nvidia, which plans to invest $1 billion directly into Naver Corp., and Brookfield, the more than $1 trillion asset manager that has signed a nonbinding term sheet to fund up to $9 billion of the buildout. Naver will cover the remaining amount. The plan scales the initial NVIDIA DSX AI factory at Naver's GAK Sejong data center from 55 megawatts to 200 megawatts by 2028 — with a runway to gigawatt scale beyond that.
The Deal in Numbers
The economics are structured so that Naver's national champions status and Brookfield's balance sheet do the heavy lifting, while Nvidia supplies both chips and a strategic stake. Crucially, Nvidia's $1 billion is conditional: the company said its investment is "subject to customary closing conditions and NAVER finalizing at least $9 billion of committed financing for the project, separate from NVIDIA's planned investment." In other words, the Brookfield financing has to firm up before Nvidia's equity check clears.
The 200-megawatt facility is expected to run on Nvidia's forthcoming Vera Rubin and current Blackwell platforms, delivered through the DSX stack that Nvidia markets as a codesigned bundle of chips, systems, software and facilities tuned to minimize the cost per token of output. Naver, already an Nvidia Cloud Partner, will operate the site as multi-tenant infrastructure — effectively a domestic hyperscaler renting production-scale compute to startups, enterprises and government bodies in Korea, plus AI innovators in the United States.
A National Champion Goes Vertical
For Naver, the move is a deliberate expansion beyond the ad-and-commerce engine that still pays the bills. "NVIDIA's planned strategic investment and our infrastructure supply agreement with Brookfield have propelled NAVER's vision for the AI factory business into a robust execution phase," said Haejin Lee, the company's founder and chairman. "Leveraging the solid partnerships with our global partners, we will drive technological innovation, foster a sovereign AI ecosystem and spearhead efforts to strengthen South Korea's AI competitiveness."
The infrastructure will also feed Naver's own model ambitions. The company is retooling its HyperCLOVA X models to build on Nvidia's Nemotron open models, and it became the first Korean firm to join the Nemotron Coalition. Naver plans to launch an AI agent platform in Korea in the second half of the year and is developing a "Seoul World Model" trained on its proprietary street-view and spatial data.
The Sovereign-AI Trend
The Naver deal is one node in a fast-thickening web of "sovereign AI" arrangements in which governments and their national tech champions build domestically controlled compute rather than renting it from U.S. cloud giants. In the same week, Nvidia announced expanded AI-factory partnerships with SK Group and a joint research lab with KAIST — a coordinated Korean push timed to the presidential summit. Jensen Huang, Nvidia's founder and CEO, framed the logic bluntly: "AI factories are the infrastructure nations need to compete and innovate in the intelligence era. Together, NAVER, NVIDIA and Brookfield are building sovereign AI infrastructure at the scale needed to fuel Korea's startups and industries."
For Nvidia, sovereign deals are a way to diversify demand beyond a handful of American hyperscalers while embedding its full stack as the default national platform. For Brookfield, which manages roughly $100 billion across the AI-infrastructure value chain and about $12 billion of Korean assets, the appeal is patient-capital exposure to power-hungry data centers. Sikander Rashid, Brookfield's global head of AI infrastructure, cast the theme as a durable one: "As AI adoption accelerates across the global economy, access to trusted, sovereign and scalable AI infrastructure is becoming an increasingly important strategic priority for companies and countries."
The dynamic it sets up is a national champion — Naver — positioning itself as Korea's answer to AWS or Microsoft Azure, backed by the same chipmaker and the same class of infrastructure financiers that underwrite the American giants. The bet is that Korean enterprises, regulators and startups will prefer compute that lives onshore and answers to a Korean operator.
The Cost Shows Up in the Books
The ambition is already visible in Naver's financials. In second-quarter results reported Aug. 7, the company posted record revenue of 3.39 trillion won (about $2.4 billion), up 16.2% year over year, powered by AI-infused advertising and commerce. Its new "AI Briefing Ads" delivered click-conversion rates more than 30% higher than traditional search ads. Yet operating profit slipped 0.2% to 520.3 billion won as operating expenses jumped 19.8% on preemptive AI-infrastructure, logistics and payments spending. Management has called 2026 a peak investment year, and the AI factory is not expected to generate revenue until 2027.
That is the core tension of the sovereign-AI wager: the core business remains resilient enough to fund the buildout, but the payoff on a gigawatt of compute is years out and hinges on demand materializing at the scale Naver is provisioning for.
What to Watch
- Financing close: Nvidia's $1 billion hinges on Naver locking in at least $9 billion of committed financing. Watch whether Brookfield's nonbinding term sheet converts to firm commitments — the whole structure depends on it. - Power and timeline: Reaching 200 megawatts by 2028, let alone a gigawatt, is as much a grid-and-power problem as a chip problem. Delays in energy procurement would ripple through the plan. - Demand proof: With AI-factory revenue not due until 2027, the first real test is whether Korean and U.S. developers actually rent the capacity — and at what margin. - The sovereign race: Naver is one of several Korean champions courting Nvidia. How Naver, SK Group and others carve up domestic and export AI demand will shape who becomes Korea's default AI cloud.
"AI factories are the infrastructure nations need to compete and innovate in the intelligence era."— Jensen Huang, Founder and CEO, NVIDIA