In a podcast on January 6, Musk claimed that China would outpace the United States in artificial intelligence computing power, with its primary edge stemming from its power supply capabilities.

Based on his projections, by 2026, China's power generation capacity could soar to roughly three times that of the United States, empowering it to sustain energy-guzzling artificial intelligence data centers.

Musk is of the opinion that power generation capacity will emerge as a critical constraint in scaling up artificial intelligence systems.

The focal point of the AI race is undergoing a transformation. For an extended period, chip performance, algorithmic innovation, and data scale have been regarded as the three cornerstones of AI development. Yet, Musk's latest viewpoint unveils a more fundamental factor: electricity.

He remarked, "People tend to underestimate the challenges of power supply." As global AI data centers experience rapid growth, power demand is skyrocketing at an alarming pace.

Industry analysts have observed that the power consumption of numerous AI data centers rivals that of a small city. A November 2025 report from Goldman Sachs highlighted that power shortage bottlenecks could impede the United States' progress in the global AI competition.

In stark contrast to the United States' predicament, China has been consistently expanding its energy capacity. Goldman Sachs predicts that by 2030, China may possess approximately 400 gigawatts of reserve power capacity, exceeding three times the current total power demand of global data centers.

This implies that China will have ample power to fuel the growth in data center demand while also catering to the power requirements of other industries.

China's power advantage transcends mere generation scale; it signifies a systemic lead. China's installed power generation capacity accounts for one-third of the global total, with one in every three units of electricity consumed being green power.

This dual advantage of scale and structure lays a robust foundation. China's power system is capable of full-chain optimization, spanning raw materials, power generation, transmission, distribution, and electricity consumption, showcasing remarkable system synergy.

Notably, China stands as the sole nation in the world to have achieved independent development of a complete set of ultra-high-voltage equipment and established a comprehensive technical standard system. This technological autonomy provides a solid assurance for the stable operation of the power system.

China also guarantees the stability and reliability of its power system through the collaborative efforts of national strategies and market mechanisms. Electricity prices have long been among the lowest globally, offering a cost advantage for the operation of energy-intensive data centers.

According to The Wall Street Journal, some Chinese data centers now incur less than half the electricity costs of their U.S. counterparts.

Musk's prediction doesn't solely hinge on power. In the same podcast episode, he also presented a thought-provoking perspective: Over time, restrictive measures such as U.S. semiconductor export controls to China may lose their significance, as China is poised to resolve its chip issues.

There's a phenomenon of diminishing marginal returns in chip performance at the cutting edge, which could make it easier for China to catch up with the United States even without access to the most advanced chips.

As of June 2025, the total number of standard racks in operational computing power centers in China has reached 10.85 million, with intelligent computing power hitting 78.8 exaflops per second and storage capacity surpassing 1,680 exabytes.

More significantly, China's innovation in the artificial intelligence realm hasn't been hampered by external restrictions. According to Huatai Securities, the iteration paths of AI models in China and the United States are diverging: Overseas efforts are shifting from pre-training to post-training and reinforcement learning, relying on computing power to gain performance edges; whereas China, constrained by computing power, is focusing more on architectural and algorithmic refinements, employing attention optimization techniques to enhance training and inference efficiency and cost-effectiveness.

At the policy level, there are also proactive efforts to foster independent innovation in chips. In January 2026, eight departments, including the Ministry of Industry and Information Technology, issued the Implementation Opinions on the Special Action for 'Artificial Intelligence + Manufacturing,' explicitly proposing to strengthen the supply of artificial intelligence computing power, promote the coordinated development of intelligent chip software and hardware, and support breakthroughs in key core technologies such as high-end training chips and edge inference chips.

Duan Weiwen, a researcher at the Chinese Academy of Social Sciences, pointed out that artificial intelligence is currently in a phase of intense hype. Even if a balance between energy efficiency, green power, and chip technology can be achieved to some extent driven by national competition, if total energy consumption rises and the beneficiaries are highly concentrated, it will lead to higher and potentially irreversible environmental liabilities.

Meanwhile, some experts have also proposed more fundamental solutions. Wang Zhong, Director of the Science Fiction Industry Development Center at the Capital Institute of Science and Technology Development Strategy, believes that the current AI development path centered around large language models is unsustainable in terms of energy efficiency.

He advocates for a fundamental paradigm shift and exploring paths closer to how human intelligence is acquired, thereby achieving a 'dimensional reduction strike' on energy efficiency improvements to the 'scale-first' path.

References:

https://news.sina.cn/2026-01-07/detail-inhfpfxi7221085.d.html?vt=4

https://news.ifeng.com/c/8pk2i2YIKEE?ch=ttsearch

https://www.ckcest.cn/entry/news/detail/cb7f5fc794300001e748c5184dd0dc50

https://vip.stock.finance.sina.com.cn/q/go.php/vReport_Show/kind/lastest/rptid/819192141679/index.phtml

https://finance.china.com.cn/industry/20251110/6277091.shtml