Investment Acceleration: Trillions of Won Targeting "Computing Power Bottleneck"
South Korean tech giant Naver announced that it will invest over one trillion won in AI infrastructure by the fiscal year 2026 and simultaneously embed AI agent capabilities into its flagship business scenarios such as search and shopping. The company management emphasized that this round of investment is not a "pilot-level increment," but foundational construction aimed at scalable AI applications, prioritizing the resolution of training/inference computing power, ultra-low latency interconnectivity, and high availability issues.
Strategic Motivation: Adding Software Capability to Manufacturing Advantages
Naver proposed leveraging South Korea's engineering advantages in manufacturing industries such as semiconductors, automobiles, and batteries, combined with its capabilities in large language models, search enhancement, and multimodal algorithms and platform capabilities, to create an "industry x AI" amplification effect. The core pathway is first to validate within its own ecosystem (search, content, commercialization) and then to open up platform capabilities to empower a wider range of industry chain partners.
Infrastructure Route: Dual-Driven by GPU Cluster and Data Center
The company intends to focus on increasing high-end GPU clusters, optimizing high-bandwidth interconnect and storage architecture, and establishing a "physical AI testbed" from Seongnam headquarters to the Sejong data center, forming cross-campus computation collaboration and remote redundancy. Besides training clusters, Naver will also expand low-latency, cost-sensitive inference clusters to support high-concurrency requests in consumer-targeted search and shopping scenarios.
Ecosystem Construction: AI Agent Initially Deployed in “Search-Shopping-Service”
Naver plans to position the AI agent as a "center of action" for users, covering processes such as information retrieval, price comparison, ordering fulfillment, and after-sales consultation. For merchants, the agent will support intelligent listing, ad placement, and customer service automation; for the platform, the focus will be on establishing "explainable, traceable" content and transaction risk control mechanisms to ensure both experience and compliance.
Industry Collaboration: "Long-term Contract + Localization" with Computing Power Suppliers
Against the backdrop of global GPU supply shortages, Naver has secured its initial-scale GPUs and will enhance supply certainty and delivery speed through long-term procurement agreements and localized deployment. The company is also exploring cooperation with partners in power, cooling, and renewable energy to reduce PUE and per-unit computational carbon emissions, enhancing the sustainable operations of data centers.
Security and Compliance: From "Model Trustworthiness" to "Platform Trustworthiness"
For the deployment of AI agents, Naver will strengthen three lines of defense: "data sovereignty - privacy computing - content safety": using tiered permissions and minimized invocation constraints for training and inference; introducing red team testing, output auditing, and watermark labeling; and establishing transparent disclosure and appeal channels for commercial content and advertisements to simultaneously improve efficiency while reducing the risk of misinformation and misuse.
Market Impact: "Anchor" of South Korean AI Infrastructure
The substantial infrastructure investment could elevate South Korea's AI computing power and application capabilities to a new level:
- For the industrial chain: Stimulating orders for servers, networks, cooling, energy, and IDC engineering, forming a supporting cluster effect;
- For developers: Opening model and API ecosystems, lowering entry barriers for small and medium-sized enterprises and university teams;
- For users: By upgrading perceivable search and shopping experiences, converting to higher retention and commercial efficiency.
Risks and Variables: Three Challenges to Overcome
Firstly, supply and cost: The delivery pace and cost fluctuations of top-tier GPUs remain critical variables.
Secondly, energy consumption and location choice: Power access, cooling solutions, and the proportion of green energy will determine the TCO curve.
Thirdly, compliance and governance: Once AI agents are scaled, establishing an auditable loop for privacy, copyright, and content safety is needed.
From "AI Capability" to "AI Productivity"
If computing power and data centers are commissioned as planned, Naver's AI agents are expected to achieve a "three-fold loop of efficiency enhancement-revenue increase-cost reduction" in core business, and spill over to more industries in a platformized manner. For South Korea, this trillion-level investment round is not only an intensification of enterprise competition but also the starting point of a quantum leap in the nation's AI infrastructure capacity.
