NAVER Builds a Sovereign AI Supercomputer Starting at 55 Megawatts With Nvidia, Scaling to Gigawatt

[Distant Kernel]

NAVER, South Korea's dominant internet company and the operator of one of the most widely used search engines in Asia, announced a partnership with Nvidia to expand its sovereign AI infrastructure using the Nvidia DSX platform, starting at 55 megawatts of AI compute capacity at its GAK Sejong data centre in South Korea with plans to scale toward gigawatt-level capacity. The partnership will support NAVER's next-generation HyperCLOVA X large language models, a Seoul World Model using urban spatial data for Seoul-specific applications, and an AI Agent Platform set to launch in Korea in the second half of 2026.

Sovereign AI, the concept of nations and major technology companies building their own AI infrastructure and model development capability rather than depending entirely on US providers, is one of the most significant structural themes in the AI industry in 2026. For South Korea specifically, NAVER's infrastructure build represents a country with a strong technology industrial base making a deliberate investment in domestic AI capability that does not route all data through foreign infrastructure or depend on foreign-developed models for critical applications. The Seoul World Model in particular, using urban spatial data specific to Seoul's geography, demographics and infrastructure, is the kind of specialised application that a domestically developed and operated model is better positioned to serve than a general-purpose model trained primarily on global English-language data.

Fifty-five megawatts starting capacity is a substantial initial commitment, roughly equivalent to the power consumption of a mid-size industrial facility, and the gigawatt scaling ambition puts NAVER's long-term infrastructure plan in the same conceptual territory as hyperscale cloud providers like AWS and Azure rather than typical corporate IT build-outs. The Nvidia DSX platform providing the underlying accelerated computing foundation means this is Blackwell-generation hardware from the outset.

10810naver-corp.com

[IdlePhoenix]

55 megawatts as the starting point rather than the destination is the detail that signals how seriously NAVER is treating this investment. 55 megawatts is already a hyperscale-class commitment. Calling it a starting point implies a trajectory that would make NAVER one of the largest AI compute operators in Asia

[Sienna74]

Sovereign AI being the frame for this investment is not just corporate messaging. South Korea has genuine strategic reasons to want domestic AI infrastructure that does not route sensitive applications through foreign data centres or depend on foreign companies for model updates and access

[RomanReigns96]

The Seoul World Model concept is the most interesting specific application here. A model trained on Seoul-specific spatial, demographic and infrastructure data can answer questions about Seoul that a general-purpose global model simply cannot, not because of capability limitations but because the relevant data is not in its training set

[BigDogShane10]

HyperCLOVA X competing with GPT and Claude in the Korean language and Korean cultural context is a different product from English-language AI assistants even at comparable capability levels. Linguistic and cultural specificity matters enormously for enterprise and consumer applications in non-English markets

[Baz_26]

The AI Agent Platform launching in Korea in the second half of 2026 positions NAVER to capture the agentic AI wave domestically in the same way it established search dominance by building Korean-specific infrastructure rather than relying on Google

[Dave_37]

Nvidia winning this partnership alongside the HPE AI Factory expansion and the OpenAI Jalapeño collaboration shows the company benefiting from AI infrastructure investment at every level simultaneously, from Korean internet companies to American defence contractors to Chinese competitors building alternative supply chains

[Glenn84]

The gigawatt scaling ambition puts NAVER's plan in conversation with the energy infrastructure reality covered elsewhere this week. Where does one gigawatt of clean, reliable power come from for a Korean AI data centre, and what does its build-out timeline look like?