Defying Geopolitical Risks! Power Sector Outperforms U.S. Stocks with Resilience and Elasticity from 'Inelastic Demand'

Amid geopolitical tensions and market headwinds, the power sector—backed by strong 'inelastic demand'—has become a safe haven for assets. Taishin Asset Management指出 that driven by AI and electrification, electricity is entering a 10- to 20-year golden era. Historical data shows its resilience far exceeds that of the S&P 500. Just as NVIDIA's Jensen Huang places energy (electricity) at the base of his 'AI Five-Layer Cake,' without power, everything else is meaningless. Goldman Sachs forecasts an explosive increase in data center power demand by 2030, making electricity the core lifeline of technology and investment.

Taishin Asset Management指出 that amid surging demand from industrial growth, AI, and electrification, energy investment has reached record highs, with power as the primary investment focus. The growth trend in electricity is expected to last at least 10 to 20 years. Due to its solid inelastic demand advantage, the sector demonstrates greater resilience than U.S. stocks during major panic events.

Statistics show that one year after the outbreak of the Russia-Ukraine war, the Smart Grid & Power Index rose 5.3%, significantly outperforming the S&P 500, which fell 7.7%. Two years later, the Smart Grid & Power Index gained 23.1%, surpassing the S&P 500's 20.4% increase. In the month following the escalation of U.S.-Iran tensions, the Smart Grid & Power Index surged 20.2%, far exceeding the S&P 500's 13.7% gain.

Lin Yu-Chuan, portfolio manager of the Taishin Global AI Power Fund,指出 that as agent-type AI becomes increasingly deployed, many decision-making steps and task executions will increase token consumption by hundreds or even thousands of times, driving a further surge in computing power demand. Goldman Sachs analysts estimate, based on current trends, that global token demand will exceed 24 times current levels by 2030. Goldman Sachs has raised its global data center power demand forecast, projecting it to reach 220% of 2023 levels by 2030, up from a previous estimate of 175%. Analysts also project that data center power demand in 2030 will grow by 905 terawatt-hours (TWh) compared to 2023, with approximately 60% coming from the United States—the world's largest power market.

Lin Yu-Chuan指出 that according to Jensen Huang's 'AI Five-Layer Cake' theory, the first layer is the Energy (Power) Layer: AI computation requires massive power, making electricity the mother of AI. The second layer is the Chips Layer, including integrated systems like GPUs and HBM, which convert power into computing power. The third layer is the Infrastructure & Cloud Layer, encompassing data centers, platforms, cloud services, networking, and cooling systems. The fourth layer is the Models Layer, where various AI training and inference models are developed. The fifth layer is the Applications Layer, hosting AI services such as autonomous vehicles, robots, and AI assistants. 'Without power, everything is just talk.'

*Disclaimer: References to individual stocks, funds, or futures products in this article are for informational purposes only and do not constitute investment advice. Investors should make independent judgments, carefully assess risks, and bear their own profits and losses.