The global data center market is one of the most certain structural growth industries over the next decade. However, it is unlikely that all data centers will grow—instead, the market will likely be dominated by operators who have secured actual power access. 1. Until 2030: The Hyper-Growth Phase JLL forecasts that global data center capacity will nearly double from approximately 103 GW in 2025 to around 200 GW by 2030, with an annual growth rate of roughly 14%. This expansion will require up to $3 trillion in combined investment in buildings, power infrastructure, GPUs, and networks. Projected regional capacity by 2030: • Americas: 109 GW • Asia-Pacific: 57 GW • Europe, Middle East, and Africa: 34 GW AI’s share of total workloads could rise from about 25% in 2025 to 50% by 2030, and around 2027, inference demand may surpass training as the primary growth driver. 2. 2030–2036: Sustained Growth, But Speed Determined by Power The IEA projects global data center electricity consumption will rise from approximately 460 TWh in 2024 to 945–1,000 TWh by 2030 and to around 1,200–1,300 TWh by 2035. In the U.S., data centers could consume up to 11.8% of total electricity by 2030, with scenarios ranging from 9.5% to 15.3%. Starting from the estimated 200 GW global capacity in 2030, global capacity by 2036 could be: • Conservative: 280–310 GW — Driven by AI efficiency gains, adjusted investment, and grid delays • Base: 320–360 GW — Sustained expansion of AI agents, video, enterprise inference, and cloud services • Optimistic: Over 400 GW — Explosive growth in inference usage paired with successful expansion of generation and transmission infrastructure Note: These are not official projections but my own scenario estimates based on JLL’s 2030 capacity forecast and IEA’s 2035 power demand outlook. 3. Why Growth Will Continue for a Decade • Transition from AI training to constant inference serving billions of users • Migration of enterprise servers to cloud and colocation • National initiatives for data sovereignty and sovereign AI infrastructure • Computing demands from autonomous vehicles, robotics, and video generation • AI adoption across core industries: finance, healthcare, manufacturing • Rising demand for data storage, security, and backup Inference, unlike one-time large-scale training, repeats every time a user asks a question or runs a video or AI agent—so as AI permeates daily life, demand accumulates continuously. 4. The Real Bottleneck Data centers can be built in 1–3 years, but transmission infrastructure may take 5–15 years. Already, over 2,500 GW of global generation, storage, and large-demand projects are stuck in grid connection queues. Thus, the key determinant of location success will no longer be land cost—but the timing of power connectivity. Additional bottlenecks include: • Supply constraints for transformers, circuit breakers, and cables • Liquid cooling for high-density racks • Water usage and environmental regulations • Community opposition and debates over electricity cost pass-through • Large-scale borrowing and customer concentration risks 5. The Biggest Beneficiaries The greatest opportunities lie not in building data centers themselves—but in solving the bottlenecks: 1. Power generation, transmission, transformers, and switchgear 2. Liquid cooling, CDUs, and thermal management 3. Optical communications and high-speed networking 4. UPS systems, batteries, gas turbines, and microgrids 5. Colocation operators with secured power access 6. EPC contractors, commissioning services, and operational automation 6. Korea’s Opportunities and Limitations Korea has strengths in semiconductors, power equipment, construction, and telecom networks—but faces constraints from insufficient grid capacity in the Seoul metropolitan area, high land costs, and challenges in permitting and local acceptance. Therefore, Korea’s competitive edge will not come from simply building data centers—but from: • AI campuses linked to non-metropolitan power sources • Integrated power supply combining nuclear, renewables, and ESS • Export of liquid cooling and high-density power infrastructure • Domestic demand for sovereign AI and local data storage Conclusion Demand for data centers is highly likely to continue growing through 2036. However, the market’s criteria have already shifted. Corporate value will no longer be determined by announced GWs—but by actual connected MWs; not by contracted customers—but by reliably operational racks. Over the next decade, the winners will not be the companies that announce the largest data center capacities—but those that simultaneously secure power access, cooling solutions, capital funding, and community approval.