Brookfield and Nvidia Partner with Naver on $10 Billion Sovereign AI Infrastructure Buildout in South Korea

Naver, Nvidia, and Brookfield Asset Management have unveiled a sweeping $10 billion proposal to construct a massive "sovereign AI" infrastructure project in South Korea, marking one of the largest single investments in localized computing power to date.

Announced during South Korean President Lee Jae-myung's visit to the AI Summit in San Francisco, the agreement represents a fundamental shift in how nations are approaching the artificial intelligence race. Rather than relying solely on software development and venture capital, countries are increasingly turning to heavy industrial infrastructure finance to secure the physical hardware required to train and deploy advanced models.[1][2]

The core financial mechanism of the partnership relies on a massive influx of traditional infrastructure capital, underscoring the sheer expense of modern AI facilities. Brookfield, a Canadian asset manager that currently oversees approximately $100 billion in assets across the global AI infrastructure value chain, has executed a nonbinding term sheet to provide up to $9 billion in project funding. This level of capital deployment is typically reserved for major public works, such as power plants or transit systems, illustrating the new scale of the AI industry.[5]

To complete the capital stack, Nvidia intends to inject $1 billion directly into Naver Corp as a strategic equity investment, while the South Korean technology giant will supply the remaining capital required to finalize construction and oversee daily facility operations. This tripartite financial structure highlights the immense financial burden of modern artificial intelligence, which single technology entities—even those as large as Naver—can no longer easily scale independently without specialized infrastructure partners.[2][3]

The technical roadmap for the project centers on a massive expansion of Naver's existing AI factory, which is located at the company's GAK Sejong hyperscale data center. The facility will scale from its initially planned 55-megawatt deployment to a staggering 200-megawatt capacity by 2028, representing a nearly fourfold expansion in just a few years. Naver has also stated a longer-term intention to eventually reach 1 gigawatt of dedicated AI infrastructure capacity at the site.[1][2][4]

According to Nvidia's technical disclosures regarding the project, the expanded South Korean infrastructure will utilize the Nvidia DSX platform, which is an end-to-end, co-designed architecture purpose-built to minimize token costs and maximize throughput. The buildout will incorporate the chipmaker's next-generation Vera Rubin and Blackwell computing architectures to deliver multi-tenant AI cloud capacity, providing enterprise-grade compute to domestic startups, established industries, and government organizations.[1][3][4]

The primary strategic claim driving this historic $10 billion deployment is the growing geopolitical necessity of "sovereign AI." Naver and South Korean policymakers argue that a nation must physically own and control its computing power rather than renting capacity from foreign hyperscalers like Amazon Web Services, Google Cloud, or Microsoft Azure. This localized approach is increasingly viewed as a matter of national security and economic independence in the intelligence era.

By establishing a domestically owned 200-megawatt facility, South Korea aims to ensure that highly sensitive enterprise, startup, and government data remains strictly within its own borders. This localized compute power is considered essential for training culturally and linguistically aligned foundation models, allowing the country to develop AI tools that understand the nuances of the Korean language and legal frameworks without exposing proprietary national data to US-based cloud providers.[1][3]

As direct evidence of this localized development approach, Naver is currently utilizing its existing infrastructure to develop a highly specialized "Seoul World Model." Built on top of Nvidia Cosmos world foundation models, this ambitious project utilizes proprietary urban street-view and spatial data to create highly accurate digital twins and autonomous agents specifically tailored to the South Korean market and urban environment.[3]

The sheer scale of the 200-megawatt target—and Naver's stated long-term goal of reaching 1 gigawatt of capacity—exposes the primary bottleneck currently constraining global artificial intelligence development: power generation. A 200-megawatt facility draws enough continuous electricity to power a small city, fundamentally shifting AI from a purely software-based endeavor into a heavy industrial challenge that requires grid-level coordination and massive cooling logistics.[2]

Brookfield's involvement as the exclusive capital partner for the project underscores this transition from software to heavy industry. By bringing massive capital deployment capabilities and global power infrastructure expertise to the table, the Canadian asset manager is addressing the physical realities of scaling intelligence that traditional chipmakers and software developers simply cannot solve on their own.[2][3][4]

For Nvidia, the multi-billion dollar partnership serves as a strategic mechanism to tighten its grip on where and how the world's largest AI workloads are built. By committing $1 billion into Naver and aligning with Brookfield's massive data center funding, Nvidia effectively ties its proprietary DSX platform directly into South Korea's national infrastructure, locking in a long-term hardware ecosystem that will be difficult for competitors to displace.[4]

Despite the historic scale and ambition of the announcement, the final execution of the sovereign AI factory remains contingent on several significant financial and logistical uncertainties. The headline $10 billion figure relies heavily on Brookfield's nonbinding term sheet, meaning the vast majority of the capital has not yet been fully secured or legally committed to the project.

Furthermore, Nvidia's $1 billion direct equity investment is subject to standard closing terms that place the burden of execution squarely on its South Korean partner. These conditions explicitly require Naver to successfully finalize at least $9 billion in binding project financing before the chipmaker's equity injection clears, meaning the entire deal hinges on Naver's ability to close the infrastructure funding in a tightening global capital market.[2]

Finally, scaling a single facility to a 200-megawatt—let alone 1-gigawatt—capacity will require unprecedented coordination with South Korea's national energy grid. The long-term viability of the sovereign AI project depends entirely on whether the local power infrastructure can reliably support the massive, continuous electricity demands of high-density accelerated computing without destabilizing regional supply or violating the country's environmental commitments.[4]