A 2 trillion won proof point for Naver’s AI factory
Naver’s latest update turns its Nvidia-linked AI factory plan into a live commercial story: the Korean internet group has secured an initial compute-service supply contract worth about 2 trillion won for a U.S.-based AI inference company, with revenue expected to start after commercial service begins in the first half of 2027 . The agreement is significant because it is not merely a chip-purchase headline; it is Naver showing that it can convert access to scarce Nvidia infrastructure into contracted AI compute revenue.
The structure also matters. According to Naver’s November 12 announcement, the AI factory revenue will be reflected from next year as the business being developed with Brookfield and Nvidia begins to contribute to earnings . Naver CEO Choi Soo-yeon said the company had secured an initial customer through the roughly 2 trillion won supply contract and would seek stability and profitability by adding large customers and high-growth global AI firms . That framing puts the deal in the same category as the new generation of “AI factory” projects: data-center capacity, power, cooling, GPUs, financing and cloud operations packaged into a compute service.
Nvidia is the capacity story behind the revenue
The customer contract is with a U.S. AI inference company, but Nvidia sits at the center of the infrastructure stack. Chosunbiz reported that Naver is negotiating with Nvidia on a server purchase order that would be described as the largest of its kind in Asia, as the company tries to secure Nvidia’s next-generation Vera Rubin platform ahead of rivals . That distinction is important: Nvidia is not the buyer of the 2 trillion won compute contract; rather, Naver’s ability to source and operate Nvidia systems is what makes the contract commercially credible.
Naver’s management also gave technical detail on the first deal. CFO Kim Hee-cheul said the initial supply contract represents roughly 20 megawatts of power capacity, with the equipment mix including Nvidia’s B300 platform and the next-generation Vera Rubin platform . The contract period varies by chip type but runs about four to five years, and the roughly 2 trillion won figure refers to cumulative revenue across the contract period rather than annual revenue . That clarification is crucial for investors and competitors: the deal is large, but it is a multi-year capacity commitment, not a single-year revenue surge.
Why Vera Rubin and liquid cooling are central
The market signal is not just “Naver bought GPUs.” The sharper point is that Naver is trying to position itself as a supplier of high-density AI compute at a moment when customers want the latest Nvidia systems but suitable data-center space remains scarce. Chosunbiz quoted Naver as saying the market is marked less by oversupply than by a shortage of the latest GPUs, especially Vera Rubin, and by a shortage of liquid-cooled data-center space able to host them . Seoul Economic Daily similarly reported that Naver sees a competitive advantage in having power and liquid-cooling infrastructure available for next-generation Nvidia GPUs .
That is why the server order under discussion with Nvidia carries strategic weight beyond its purchase value. If Naver can secure large volumes of Vera Rubin-class infrastructure, it can offer regional and global customers a faster route to inference capacity than building their own facilities. Seoul Economic Daily reported that Naver is seeking to secure next-generation GPU platforms in advance through large-scale orders and is pursuing what it described as Asia’s largest procurement order to offer Vera Rubin capacity ahead of competitors .
A 200 MW target already looks too small
Naver’s capacity roadmap has also shifted from ambition to execution. The company had planned to launch the AI factory business at 55 MW in the first half of 2027, expand to 100 MW by the end of that year and reach 200 MW in 2028, with a long-term goal of 1 GW of AI compute infrastructure . Edaily reported that CEO Choi told investors on the third-quarter call that contracted data-center capacity already exceeds the 200 MW target announced for 2028 .
That capacity update gives the 20 MW first contract more context. One deal fills only a slice of the first phase, but it validates the commercialization path. IT Chosun reported that Naver is discussing supply contracts with more than 10 companies for the remaining 180 MW, with counterparties centered on North America and other non-Asian AI companies . In other words, the first 2 trillion won agreement is not being presented as the finish line; it is the first booking against a much larger infrastructure buildout.
Revenue starts in 2027, margins improve later
Naver is guiding investors to a staged monetization curve rather than instant profit. The company said AI factory revenue should begin to be recognized after commercial service starts in the first half of 2027 . IT Chosun reported that Naver expects the AI factory business to contribute operating profit and net income attributable to controlling shareholders from 2027, while margins may start lower because there is a gap between equipment arrival and service launch .
The profitability story becomes more ambitious from 2028. According to the same report, Naver expects margins to improve substantially once operations are fully running, with management pointing to margins well above double digits from 2028 . That forecast depends on utilization, timely equipment delivery and continued demand for premium inference capacity, but it explains why Naver is moving aggressively on procurement and customer signing at the same time.
The financial backdrop: AI spending is already visible
The AI factory update arrived alongside Naver’s third-quarter 2026 results. Naver reported revenue of 3.4724 trillion won, up 10.7% year on year, while operating profit fell 20.9% to 451.3 billion won as the company spent more on AI competitiveness, including infrastructure, content and offline data . Yonhap also reported that Naver’s net profit fell 45.2% to 402.5 billion won, underscoring the near-term earnings cost of the AI investment cycle .
That makes the 2 trillion won contract especially important. It gives Naver a concrete answer to a common investor concern: when does AI infrastructure spending become revenue? The answer, based on the company’s latest disclosure, is that the first booked AI factory revenue begins next year and then accumulates over a four- to five-year contract period .
What this means for Asia’s compute market
For Asia’s AI infrastructure market, the deal is a signal that hyperscale compute purchasing is no longer only an American cloud story. Naver is trying to assemble a model in which a Korean platform company, Brookfield-backed infrastructure, and Nvidia’s GPU roadmap combine to serve global AI customers . If the Nvidia server negotiations close at the scale being reported, Naver would become a more visible regional supplier of premium inference capacity rather than simply a domestic internet platform.
The risk is execution. Naver still has to turn contracted capacity into working service, receive and deploy high-end Nvidia systems, keep utilization high, and sign enough additional customers to fill the remaining megawatts. But the company has now put numbers around the thesis: 20 MW, roughly 2 trillion won over about five years, more than 10 additional customer discussions, and a capacity base that management says already exceeds its 2028 target .
The short version: Naver has booked the first major commercial proof point for its Nvidia-centered AI factory strategy. The bigger test is whether this initial 2 trillion won contract becomes the template for a broader Asian compute platform.
Comments
Be the first to comment.