Powering the AI Race: Taiwan, China, and the Hidden Cost of Electricity

Artificial Intelligence (AI) has emerged as a dominant technological force that shapes our daily lives. Since OpenAI released ChatGPT in 2022, every industry has begun to transform, investing heavily in AI-related infrastructure to enhance manufacturing efficiency. To secure technological leadership, most corporations and governments focus on semiconductors and computing power. However, competitions in these two sectors ultimately reflect more visible constraints: energy and geopolitical issues. In addition to electricity-intensive infrastructure, AI development relies heavily on the global supply chain, particularly in East Asia. This makes the entire AI industry exposed to both energy and geopolitical risks.

Data centers form the backbone of AI infrastructure, supporting machine learning and deep learning by providing computing power and large data storage. As the demands of AI models increase, companies are committing unprecedented capital to expand data centres. There are currently 11,038 data centres operating across 174 countries. Global investment is projected to reach nearly $7 trillion, and demand for data centre capacity is expected to triple by 2030.

Behind the AI boom lies a harsh physical reality: data centres require a vast amount of energy to operate effectively. In 2025, these facilities consumed 448 TWh to power the computing system. Specifically, the United States hosts approximately 48% of global data centres and accounts for 4.4% of total US electricity consumption, roughly equal to France's annual electricity consumption. Cooling requirements also impose significant water stress: 5 million gallons of water are consumed per day to maintain hardware reliability.

The dependency on natural resources makes the whole AI industry risky. Despite large-scale technology companies adopting green energy sources such as nuclear and solar power due to ESG policies, grid electricity remains reliant on fossil fuels. However, fossil fuel production is concentrated in a few regions, including the United States, Russia, and the Middle East, increasing instability in these regions. For instance, in 2026, the oil crisis broke out in the Strait of Hormuz, where 1/5 of the global oil supply passes, resulting in shortages and volatility in crude oil prices. As electricity costs surge, the AI centre's operational costs increase significantly, reducing profit margins and slowing AI investment.

Taiwan, an island in East Asia, has been at the forefront of the AI revolution. It accounts for the majority of advanced semiconductor production, which is necessary to support AI workloads such as large language model training and data-intensive cloud services. Specifically, Taiwan Semiconductor Manufacturing Company (TSMC), the world's largest dedicated semiconductor producer, accounts for over 64% of dedicated contract chipmaking and produces more than 90% of the world's most advanced logic chips, which it sells to AI technology giants such as Nvidia, Microsoft, and AMD. Other than TSMC, Foxconn, Wistron, and MediaTek also play significant roles in the AI supply chain.

The dependency has been reflected in the consumer market: due to the Iran War, helium exports have been disrupted, driving up its price. This has raised semiconductor costs, forcing Sony to raise the price of its game console, including the PS5 by $100 in 2026. This demonstrates how supply chain and geopolitical pressures can easily impact manufacturing processes and be transmitted to consumers.

However, this supply chain is becoming increasingly vulnerable as the conflict between China and Taiwan has escalated over the past ten years. Since 1945, when the communist party defeated the nationalist party and forced them to retreat to Taiwan, the Taiwan Strait has remained one of the most politically sensitive regions in the world. In 2022, China launched 11 missiles into the waters surrounding Taiwan and conducted large-scale military exercises, causing what is widely referred to as the Fourth Taiwan Strait Crisis.

Even though China chose not to invade Taiwan initially, maritime and aerial quarantine and blockades can generate destructive damage to semiconductor production. As mentioned, manufacturing processes require significant amounts of electricity. Regarding TSMC only, it consumed over 2.5% of Taiwan’s total electricity generated. However, Taiwan’s energy sector is heavily reliant on imports: over 95% of the oil, coal, and natural gas used in energy production is imported, and these imports accounted for over 75% of Taiwan's total energy generation. Although Taiwan’s government is proactively developing sustainable energy, including solar and wind, it still takes time to mitigate the risk.

Besides, TSMC requires stable front-end manufacturing supplies because raw materials, such as chemicals and gases, cannot be easily stored in large quantities. According to the records, the 2011 Tōhoku earthquake and tsunami in Japan reduced semiconductor production by 1/4, as Japan was one of the largest suppliers of its raw materials. Additionally, the inferior quality of photoresists purchased from Dow Chemicals in 2019 generated massive losses of 5.5 billion US dollars.

If foundries in Taiwan were to halt production due to blockades from China, the S&P 500, which reached a record high recently because of AI, is likely to collapse due to the drop in Magnificent Seven, including Nvidia, Apple, Microsoft, Google, Meta, Amazon, and Tesla, that significantly drives S&P 500 gains. The global economy is expected to lose up to 10 trillion US dollars in the first year, surpassing the damage from the 2008 financial crisis and the COVID pandemic.

These two concerns emphasise that the AI ecosystem is built on a fragile foundation, where geopolitical and energy fluctuations can easily destabilise the entire supply chain. This reflects the paradox in which AI is initially promised unprecedented efficiency, yet is constrained by physical and political systems. The future of AI will not be determined solely by technology and computing power, but by each nation's ability to ensure energy and supply chain stability.

References

Campagnola, D. (2025, August 22). Chips made Taiwan indispensable. AI can make it unstoppable. Lowy Institute. https://www.lowyinstitute.org

Goldman Sachs. (2024, May 14). AI is poised to drive 160% increase in data center Power demand. Goldman Sachs. Retrieved March 26, 2026, from https://www.goldmansachs.com

Kumar, D. (2026, January 9). A $10-trillion reckoning: how a Chinese invasion of Taiwan would upend markets. https://invezz.com

United Nations. (2025, April 7). Artificial intelligence: How much energy does AI use? UNRIC.org. https://unric.org

Vision of Humanity. (2025, June 16). The world’ s dependency on Taiwan’ s semiconductor industry is increasing. Vision of Humanity. https://www.visionofhumanity.org