Naver's AI factory is being redefined as a sovereign AI infrastructure platform for South Korea

Overview

This report focuses on the investment implications of Naver's sudden pivot toward a neocloud and AI factory strategy. Bernstein believes Naver's early infrastructure buildout looks similar to neocloud companies such as CoreWeave, but its ultimate goal is more likely to be building a regional AI platform centered on South Korean sovereign AI, search, enterprise services, content, and a software stack. The report also discusses SPV financing, South Korea's national AI policy support, comparisons with CoreWeave and Nebius, and execution indicators that investors should monitor.

Core views

The core view is that Naver's AI factory should not be viewed merely as a capital-intensive GPU infrastructure project. In the short term, it does depend on GPUs, data centers, power, long-term customer demand, and project financing, so its economic metrics may resemble those of CoreWeave. However, if Naver can layer its search business, Korean-language AI, enterprise customers, content ecosystem, and sovereign AI positioning onto the infrastructure over the long term, its valuation framework should shift from a pure neocloud model toward a broader AI platform. The South Korean government's 8.4GW AI infrastructure plan further reinforces this framework, giving Naver's 1GW plan policy backing, potentially lower-cost capital, and power coordination support.

Analysis framework

The report combines cross-sectional comparison with strategic analysis: it first compares Naver with publicly traded neocloud companies such as CoreWeave, IREN, and NBIS across infrastructure, financing, customer demand, and leverage structures, then analyzes Naver's differentiation in search, Korean-language AI, sovereign AI, and full-stack integration. It subsequently uses a Meta-Blue Owl-style SPV case study to illustrate how Naver could place GPU, land, construction, power, and cooling assets into a project financing structure to reduce pressure on the parent company's balance sheet.

Methodology notes

• Business model comparisonNeocloud comparison framework

  Compare Naver with AI infrastructure operators such as CoreWeave, IREN, and NBIS

  This framework focuses on power capacity, GPU infrastructure, customer contracts, revenue backlog, financing structures, leverage, and software-stack depth to assess whether Naver should be valued according to a neocloud infrastructure logic in the short term.

• Financing structure analysisSPV and project financing framework

  Use special-purpose vehicles to isolate capital expenditures and debt

  The report assumes Naver could place GPUs, racks, power, cooling, and data center assets into SPVs, supported by infrastructure investors, policy banks, and project debt, while securing returns through long-term leases or capacity contracts.

• Strategic valuation frameworkSovereign AI platform framework

  Shift from compute leasing toward national AI infrastructure and a regional AI ecosystem

  This framework emphasizes policy support, Korean-language models, local content, enterprise customers, and export-oriented AI data center solutions, giving Naver a potential valuation logic that goes beyond pure compute monetization.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Naver / 035420.KS

  Core subject of the report; Bernstein rating: Outperform

  Strengths

  Has a search business, Korean-language AI assets, domestic distribution, enterprise relationships, a content ecosystem, and policy support; its AI factory could evolve into a regional AI platform.

  Weaknesses

  In the early stages, it remains highly dependent on capital expenditures, power, GPU supply, utilization, customer commitments, and financing execution.

  Comparison

  Its short-term infrastructure model resembles CoreWeave, while its long-term platform vision is closer to Nebius.

  Risks

  Execution below expectations, failure of SPV financing to reduce pressure on the parent company, anchor customer concentration, insufficient power costs or capacity, and platform monetization below expectations.

• CoreWeave / CRWV

  Neocloud comparison company; Bernstein rates CRWV Underperform

  Strengths

  Mature GPU cloud and orchestration capabilities, contract-backed expansion, and market sensitivity to contracted growth momentum.

  Weaknesses

  The report has concerns about its long-term fundamental business model and role in the infrastructure ecosystem.

  Comparison

  The economic characteristics of Naver's early-stage AI factory may resemble CoreWeave, but Naver has more local platform and sovereign AI assets.

  Risks

  Long-term business model, leverage, customer concentration, and positioning within the infrastructure ecosystem.

• Nebius / NBIS

  Long-term strategic comparison company; the report states that it is not covered

  Strengths

  Represents a path from infrastructure toward a broader AI platform.

  Weaknesses

  The report provides no Bernstein rating or detailed financial assessment.

  Comparison

  Naver's end state is closer to a Nebius-style regional AI platform than to simple CoreWeave-style GPU compute leasing.

  Risks

  The platformization path requires sustained execution across software, models, customers, and the ecosystem.

• Bitcoin miners / Emerging AI Infra

  Supply-side assets related to AI computing infrastructure

  Strengths

  Miners have approximately 30GW of power portfolios and have signed approximately 6.7GW with hyperscalers, neoclouds, and AI chipmakers, representing more than $130bn in contracts.

  Weaknesses

  Their transition to AI infrastructure remains constrained by power, financing, customers, and execution.

  Comparison

  Compared with Naver, miners are more focused on power and data center supply, while Naver is more focused on a policy-supported AI platform and regional ecosystem.

  Risks

  Contract execution, capital expenditures, energy costs, and fluctuations in AI demand.

Key data

• Naver ratingOutperformBernstein maintains its Outperform rating on Naver.
• Naver current price and target priceKRW 192,000 / KRW 330,000The pricing date in the table is July 10, 2026.
• Long-term AI factory development targetApproximately 1GWNaver plans to expand to approximately 1GW of AI data center capacity over five to six years.
• Initial deployment scale55MWThe report states that Naver is starting with a 55MW footprint and plans to continue expanding.
• End-2027 target100MW active powerThe initial phase will be achieved through leased facilities.
• Initial anchor customer capacity200MWThe anchor customer has not been disclosed, but has committed to an initial 200MW power capacity.
• Management revenue target20 trillion wonManagement targets an additional 20 trillion won in revenue from the AI factory over the next five to six years.
• South Korea's national AI infrastructure plan8.4GWUnder the South Korean government's plan, approximately 1GW is allocated to Naver, approximately 5GW to SK, and approximately 2.4GW to GS.
• Policy investment framework550 trillion wonThe national AI infrastructure project is included in a broader investment framework.
• Potential capital expenditure financing scaleApproximately $40-60bnThe report believes that policy support and SPV structures could improve the credibility of financing with limited pressure on the parent company's balance sheet.
• SPV investor target return7-9% IRRThe report assumes that infrastructure investors may participate in GPU SPVs at this return range.

Impact & implications

For investors, Naver's investment narrative has shifted from that of an overlooked South Korean internet company to an AI infrastructure and platform opportunity attracting attention. If SPV financing, policy funds, power supply, and anchor customer contracts are successfully implemented, the market may assign it a valuation higher than that of a traditional internet business or a pure capital expenditure project. However, if platformization fails to materialize, Naver may still be valued as a highly capital-intensive neocloud.

Risks

• Management may fail to execute the AI data center expansion and customer signings according to schedule.
• SPV, project debt, and policy financing may fail to effectively isolate parent-company leverage or reduce the cost of capital.
• Power, water resources, cooling, and high-voltage grid connections may fall short of expectations.
• The initial 200MW anchor customer may represent a high concentration, and demand visibility still needs to be validated.
• GPU supply, NVIDIA cooperation, long-term GPU prices, and utilization remain uncertain.
• If monetization of search, enterprise AI, content, and sovereign AI platform capabilities is insufficient, the market may continue to value Naver as a capital-intensive neocloud.
• Accounting treatment, lease structures, or residual-value guarantees could cause more debt or assets to return to the parent company's balance sheet.
• There are execution risks related to policy support, preferential electricity prices, policy bank funding, and export-oriented AI data center solutions.

What to watch

• Whether Naver discloses the identity of the initial 200MW anchor customer, contract duration, minimum usage, and pricing mechanism.
• Construction progress, power connections, and commissioning milestones along the 55MW, 100MW, and 1GW roadmap.
• Details of the SPV structure, including Naver's ownership percentage, external investors, debt recourse, IRR target, and accounting treatment.
• Policy funds, preferential electricity prices, grid capacity, and water resource arrangements under South Korea's 8.4GW national AI infrastructure plan.
• The phased delivery of the 20 trillion won AI factory revenue target.
• Whether Naver can combine its infrastructure with HyperClovaX, search, enterprise AI, content, and sovereign AI workloads.
• Partnership announcements involving NVIDIA, chipmakers, LLM companies, and infrastructure investors.
• Whether the market shifts from a pure neocloud valuation toward a platform-oriented AI infrastructure valuation.