https://en.people.cn/n3/2026/0109/c90000-20412189.html

According to Xinhua, Elon Musk Must said in an episode of the “Moonshots with Peter Diamandis” podcast that China will far exceed the rest of the world in AI compute, because China is on track to outpace every other country in the computing power needed to run artificial intelligence, and domestic chip capabilities are accelerating their catch-up.

Musk estimated that China could reach about three times the electricity output of the US by 2026, giving it the capacity to support energy-hungry AI data centers. Electricity generation is considered the limiting factor to scaling AI systems.

Power supply constraints in many parts of the US, combined with the concentration of major tech companies and the enormous electricity demands of AI data centers, have likely limited the further expansion of some projects in the USA.

Musk’s comments come as energy supply and data infrastructure emerge as key constraints in scaling AI, rather than chips or algorithms, according to Business Insider.

According to Wei Shaojun, vice chairman of China Semiconductor Industry Association, China must continue to strengthen its independent chip development. According to a Morgan Stanley research report released in 2025, China’s self-sufficiency rate in AI GPUs has risen from below 10 percent in 2020 to about 34 percent in 2024, and is expected to reach roughly 82 percent by 2027. According to a recent government action plan, issued by eight government agencies including the Ministry of Industry and Information Technology, China aims to achieve secure and reliable supply of key core AI technologies by 2027, with its industrial scale and empowerment level remaining among the world’s forefront. The plan aims to integrate AI more deeply into manufacturing and drive new industrial growth. By 2027, it targets

• the deep application of three to five general-purpose large AI models in manufacturing,
• the development of specialized, full-coverage industry-specific large models,
• the creation of 100 high-quality industrial datasets, and
• the promotion of 500 typical application scenarios.