The growing demand for electricity is emerging as a critical factor shaping the future development of artificial intelligence (AI), underscoring a challenge that extends beyond advances in computing hardware and algorithms. Experts highlight that powering AI systems—from training large language models to operating daily services like ChatGPT and autonomous technologies—requires substantial and reliable energy supplies, prompting urgent attention to power infrastructure and sustainability.

This energy constraint is especially evident in China’s Taiwan region, a global hub for advanced semiconductor manufacturing responsible for producing 92 percent of the world’s cutting-edge chips essential to the AI industry. However, Taiwan’s electrical grid is under increasing strain as AI projects scale up. For instance, Nvidia’s proposed “giant AI supercomputer” is expected to initially require 20 megawatts of power, potentially rising to 100 megawatts in the future. Meanwhile, Taiwan is behind schedule in meeting its renewable energy target of sourcing 20 percent of electricity from renewables by 2025. Analysts warn that this energy bottleneck may limit Taiwan’s ability to fully leverage its semiconductor strengths in the expanding AI economy.

The challenge underlines a broader tension in the AI competition: success is not solely determined by semiconductor production but also by the capacity of regional power grids to support large-scale computing needs. Areas with flexible, modern energy infrastructure and available capacity have a competitive advantage over those facing slow grid expansion and aging systems.

Mainland China has grappled with these dynamics for years. The country’s digital economy has grown unevenly, concentrated in megacities along the coast, while significant renewable energy resources remain underutilized in western provinces such as Xinjiang and Sichuan. To address this mismatch, Beijing has implemented a strategy of relocating energy-intensive computing tasks to western regions where electricity is abundant, affordable, and increasingly green. Guizhou province exemplifies this approach, hosting large data center clusters from major companies including Tencent, Apple, Tesla, and Huawei. Guizhou offers a cool climate that reduces cooling costs and ample hydropower supply, turning the previously less-developed province into a vital AI infrastructure hub.

This approach offers multiple benefits: easing pressure on coastal power grids, making use of otherwise wasted renewable energy, and creating a spatial division of labor whereby innovation and algorithm development remain in coastal cities like Beijing and Shanghai, while computation shifts inland.

For Hong Kong and Taiwan, the situation calls for tailored strategies due to their limited land and energy resources. Taiwan faces the risk that ongoing energy constraints could undermine its semiconductor dominance if not addressed. Experts suggest that expanding renewable energy, reconsidering nuclear power, and establishing dedicated “AI energy zones” near existing substations are vital steps. Additionally, concentrating AI computing within designated industrial parks and outsourcing less critical processing might help manage grid loads.

Hong Kong confronts a different set of challenges, lacking significant domestic energy generation and space for large data centers. The city is expected to focus on high-value, low-energy AI sectors such as algorithm design, financial AI, and regulatory technology. Moreover, Hong Kong could serve primarily as a coordination hub, channeling investment and talent into the broader Guangdong-Hong Kong-Macao Greater Bay Area where nearby cities like Huizhou and Jiangmen possess more capacity for AI infrastructure.

Both regions must also prepare for intra-urban conflicts arising from competing demands on limited power supplies that affect essential services such as hospitals, schools, and housing. Questions of financing new energy infrastructure and prioritizing power allocation are not only technical but also political, requiring clear, equitable policies.

As AI continues to evolve, energy availability and grid readiness are emerging as crucial determinants of regional competitiveness. Mainland China’s energy-aware approach offers a reference point, but Hong Kong and Taiwan face urgent decisions on how rapidly they can adapt their energy systems to keep pace with AI’s growing demands.