Naver Locks $10B From Nvidia and Brookfield for Gigawatt AI Data Center
South Korea’s Naver has secured commitments totalling roughly $10B to build what it calls an “AI factory” centred on the GAK Sejong data centre. Brookfield Asset Management is committing more than 10 trillion won ($9B+) to fund construction and power infrastructure. Nvidia’s $1B contribution arrives primarily as advanced GPU platform technology and supply commitment rather than a direct cash stake.
The deal makes Naver one of the better-capitalised sovereign AI infrastructure plays outside the US and China, anchored by a company that already operates South Korea’s dominant search, cloud, and e-commerce stack.
What Gets Built
GAK Sejong opens at 55MW — a meaningful initial deployment but a fraction of the stated roadmap. Naver’s target is gigawatt-scale capacity by 2029, an 18x expansion over roughly three years.
Brookfield’s capital covers both the physical data centre build and the power infrastructure required to sustain it. Nvidia’s role is strategic: Naver Cloud becomes a primary distribution node in an Nvidia-South Korea ecosystem that envisions deploying over 60,000 GPUs across government, enterprise, and research tenants.
The Sovereign AI Context
This follows a pattern now visible across multiple geographies. Japan committed $6.2B through a SoftBank-led consortium to a domestic AI foundation model. France placed a €10B bid for an EU AI gigafactory targeting 288,000 GPUs. The difference with Naver is that it is a commercial-first build: a dominant internet platform, not a government lab, driving the infrastructure.
That structure matters for utilisation. Sovereign AI projects built around state-funded research often struggle to fill capacity. A Naver-anchored facility starts with a paying anchor tenant already running large-scale inference workloads — search, recommendations, cloud services — before any government or third-party load is added.
Naver’s position also gives the facility a natural export layer: South Korean enterprises avoiding dependency on AWS, Azure, or Google Cloud can route through a domestically owned compute stack. The 1GW target by 2029 would make GAK Sejong one of the largest AI data centre campuses outside the US and China, comparable in stated ambition to some of the major US hyperscaler regional builds.
Nvidia’s Play
For Nvidia, the Korea commitment extends a pattern of co-investment in sovereign AI infrastructure rather than simply selling hardware at arm’s length. Similar arrangements include the Japan NIMS partnership (27,500 Rubin GPUs, 140MW), and strategic GPU allocation deals with European national labs.
Locking in Naver as a reference deployment for the next-generation GPU stack is worth more than the $1B in technology contribution: it secures a major regional distribution point and positions Nvidia GPU platforms as the default for South Korea’s AI buildout ahead of competing silicon from Huawei and domestic Korean chip development.