Naver Lands $1.5 Billion AI Computing Deal With US Inference Firm

Naver has signed a 2 trillion-won ($1.5 billion) computing services contract with a U.S. AI inference company, securing the first customer for its AI infrastructure business months before the planned commercial launch in the first half of next year. The deal, announced Thursday alongside the company's third-quarter earnings, marks the first concrete revenue commitment for the Korean internet giant's push into AI computing, a market where hyperscale data center capacity has become one of the scarcest commodities in the global technology supply chain.

The contract covers computing services built on next-generation graphics processing units (GPUs), with the initial supply of services scheduled for 2027. The first deal covers 20 megawatts of capacity.

"This is a meaningful contract that confirms actual market demand and the initial commercial viability of the AI factory business," CEO Choi Soo-yeon said during the company's third-quarter earnings call Thursday. "Revenue is also expected to be generated sequentially as commercial services begin in the first half of next year."

AI factory is Nvidia's term for next-generation data centers that generate tokens — the basic units of output from large language models — rather than simply storing data or hosting websites.

Why the deal matters for Naver's AI factory bet

Naver is developing its AI factory business in partnership with the Canadian asset manager Brookfield and U.S. chipmaker Nvidia. The company plans to expand capacity from an initial 55 megawatts next year to 200 megawatts by 2028.

The company said it has already secured contracted data center capacity exceeding that initial 200-megawatt target for 2028. It is also in talks with more than 10 companies about additional AI infrastructure supply contracts, following the first 20-megawatt deal worth 2 trillion won.

The pipeline matters because AI infrastructure is capital-intensive and carries substantial execution risk. Choi said Naver is pursuing long-term contracts with major customers while expanding its client base to include global AI companies with strong growth potential. The strategy aims to reduce Naver's reliance on individual clients while seeking stable, higher-margin business.

For Naver, the stakes extend beyond a single contract. The company, which dominates Korean search and operates commerce, fintech and content platforms, is betting that selling AI computing capacity can open a new revenue stream less dependent on its mature domestic advertising market.

What happened to third-quarter earnings?

The announcement came as Naver reported record quarterly revenue but weaker operating profit, with AI-related investment weighing on earnings.

The numbers show the tension:

• Sales reached 3.47 trillion won, up 10.7 percent year-on-year — the company's sixth consecutive quarterly revenue record.
• Operating profit fell 20.9 percent to 451.3 billion won.
• Operating margin stood at 13 percent.

By business segment, the breakdown was uneven. Platform revenue rose 7.3 percent year-on-year to 1.95 trillion won, supported by advertising growth and increased purchases through Naver's commerce service and membership programs. Financial platform revenue gained 14.1 percent to 487.7 billion won. Global business revenue increased 15.9 percent to 1.04 trillion won, driven by growth in its consumer-to-consumer marketplaces, including Wallapop in Europe and Poshmark in North America.

The 20.9 percent profit decline reflects the cost of the AI buildout. Naver is funding data center expansion, GPU procurement and model development simultaneously, and the earnings call made clear that management views the 2 trillion-won contract as evidence the investment can convert into revenue.

What comes next: logistics, agents and expansion

Beyond AI infrastructure, Naver outlined two additional strategic moves.

First, the company plans to establish a dedicated logistics subsidiary and gradually expand its delivery infrastructure to strengthen its commerce business. Choi estimated the investment to bolster delivery capabilities would approach 1 trillion won through 2030. She added that the company would seek appropriate financing and revenue structures to limit the impact on cash flow and earnings — a signal that Naver intends to avoid funding the buildout entirely from its operating cash.

Second, Naver plans to launch a personal AI agent app by the end of this year. The app will allow users to carry out tasks such as shopping and making reservations through Naver's services, rather than simply receiving search results or recommendations. The move positions Naver in the emerging category of agentic AI (systems that take actions on a user's behalf rather than only producing text or links), where global competitors are racing to convert chatbot interfaces into transactional tools.

The sequencing is notable. Naver is entering the agent market through its own services — search, commerce, reservations — where it already controls the transaction endpoints, rather than competing head-on with general-purpose assistants.

Can the AI factory scale profitably?

The 2 trillion-won contract answers one question for Naver: whether customers will commit to its AI factory before it opens. It leaves others open.

Initial GPU supply arrives in 2027, meaning revenue recognition is sequential and partly deferred. The talks with more than 10 additional companies suggest demand, but no further contracts were announced. And the profit decline shows that the transition period carries real cost: Naver absorbed a 20.9 percent drop in operating profit in the same quarter it signed its first billion-dollar-scale infrastructure deal.

The company's stated approach — long-term contracts, a diversified customer base of global AI companies, stable higher-margin business — is designed to address concentration risk. Whether the pipeline of more than 10 potential customers converts at terms comparable to the first 20-megawatt deal will determine whether the AI factory becomes a genuine second engine for Naver or a capital-intensive sideline.

With commercial launch set for the first half of next year, capacity expansion from 55 to 200 megawatts by 2028, contracted capacity already beyond that target, and a personal AI agent app due by year-end, Naver has committed to a dense delivery schedule across infrastructure and consumer products. The first half of next year, when commercial services begin and sequential revenue starts flowing, will be the first measurable test.

By Marcus Bennett