Artificial intelligence now comes up far more often in everyday conversation. Hope and anxiety about AI exist side by side, but few would dispute that it will shape our lives and our future in a major way.
As AI transforms economic activity across the board, the energy underpinning it is being overlooked. AI is usually discussed in terms of graphics processing units, models, agents and the cloud. Behind training, inference clusters and the expansion of data centers, however, lies a simple truth: a stable supply of electricity that can be brought online quickly.
AI's computing demand, whether for training or inference, translates directly into power consumption. The structural feature of the AI industry is that advances in software convert straight into demand for hardware and energy. Every internet search is now tied to AI as well, so electricity use is bound to rise sharply. The International Energy Agency projects that global data center power demand will reach 1,000 terawatt-hours in 2026, 2.2 times the 460 TWh recorded in 2022, and 1,580 TWh by 2034, or 3.4 times that level. Those figures are 1.68 times and 2.65 times South Korea's 2025 electricity consumption of 596 TWh, the eighth-largest in the world.
The biggest obstacle to the spread and development of AI is pressure on the power supply. What keeps executives at the front line of the AI industry awake is not the precision of algorithms or the performance of semiconductors. It is the more fundamental question of how to secure electricity. Until now the debate has been about which power sources to choose. The question now is which power sources AI requires.
Countries are using every available means to secure electricity. The United States is pursuing an "all-of-the-above" strategy, investing simultaneously in nuclear, renewables and gas. Germany is adding gas-fired generation alongside renewables in a bid to stabilize its grid and lower electricity bills at the same time. Japan has designated green transformation strategy zones and is strengthening its power mix with renewables, nuclear, gas, hydro and hydrogen.
Big Tech companies competing in AI, including Google, Microsoft, Amazon Web Services and Meta, are choosing gas-fired generation first because they need power urgently. Gas plants can be built relatively quickly next to data centers that must run around the clock, and they can supply electricity directly on-site without going through the transmission grid. Taiwan, home to TSMC, is also expanding renewables while planning to add 25.98 gigawatts of liquefied natural gas generation by 2035 to guard against intermittency.
Global Energy Monitor reported that gas-fired capacity under development in the United States tripled from a year earlier to 252 GW as of 2025, while capacity under development worldwide rose 31% to 1,047 GW. It said new gas-fired projects would increase further in 2026 and set another record. Rising demand for gas-fired generation has created a shortage of gas turbines, and buyers are racing to lock in orders. Turbine makers are expanding capacity yet are quoting delivery dates four or more years out. None of these numbers are policy targets. They are figures the market is generating on its own.
Gas-fired generation is one of the most important strategic dependencies shaping the future of global AI. Even during the worst of the crisis over the closure of the Strait of Hormuz in the war between the United States and Iran, growth in gas demand from the power sector, driven by AI, continued. The IEA has also projected that natural gas demand will keep rising into the 2030s. The starting point that now moves the energy industry is not energy but AI. In the end, the story of AI becomes the story of gas.
The era of competition over AI, a new form of national infrastructure, has already begun. Data centers, the physical foundation of AI, drive growth in infrastructure industries such as semiconductors, electrical equipment, batteries and cooling systems, and they extend further into power, land, heat and water. If South Korea is to turn the growth of AI into an opportunity to raise its national competitiveness, it must read the megatrend beneath the surface accurately and act on it. Only by supplying electricity quickly and reliably, within the constraints of time and technological reality, can the country lead the global AI competition. Given the reality that smart energy makes powerful AI, the capacity to supply electricity should be understood as a factor that will determine South Korea's competitiveness and its future.
