I just finished watching the robot performances on the Spring Festival Gala, and I was truly shocked. This isn’t simple dance-and-mechanical routines—it’s a collective surge from the entire industrial supply chain. Unitree’s G1 and H2 onstage no longer look like they’re “performing.” They look like coordinated bodies in real life. Magic flute’s atom robot dances together with Chen Xiaochun and Yi Yangqianxi—their moves match so perfectly it feels seamless. When Galaxy General’s robot performs the action of cracking nuts, the precision tactile-feedback control is especially striking. This isn’t a stage show of chip performance anymore—it’s a display of how mature the entire robotics industry has become.

But what’s interesting is that just as we’re cheering for these robots, the U.S. is falling into a kind of silent panic. They suddenly realize a problem: the “blood” that powers these AIs—electricity—is running out.

America’s energy crisis is no longer just talk. Earlier this year, U.S. residential electricity prices jumped by 36%, reaching $0.18 per kilowatt-hour. But that’s only the surface. The real issue is a breakdown on the supply side. Training a GPT-4-level model consumes electricity equivalent to the total electricity consumption over one year of 100,000 households. By 2028, U.S. data centers’ expected annual electricity consumption will reach 600,000 GWh. What does that mean? The U.S. power grid faces a double blow: old fossil-fuel and nuclear power facilities are being shut down at large scale, and while new energy is growing, it’s far from keeping up with demand.

To make matters worse, the infrastructure of the U.S. power grid itself is also a bottleneck. The entire country is divided into three nearly independent grid “islands”—the East, the West, and Texas—with interconnections that are painfully weak. Approving a single interstate transmission line takes 15 years. Wind power in the Midwest simply can’t be shipped to data centers on the East Coast. The CEOs in Silicon Valley are now most worried about not chip quotas, but this: where on earth can I find enough electricity to run these chips?

China’s situation is completely different. More than a decade of early planning has already built strategic advantages that the U.S. can’t replicate in the short term. By 2025, China has completed 45 ultra-high-voltage direct current (UHVDC) transmission projects, with the total length of transmission lines exceeding 40,000 kilometers. This “electricity highway” can deliver clean energy from the West to East Coast data centers at millisecond-level speed, or directly support “East Data West Computing” nodes. Among the world’s largest 37 direct-current systems, China accounts for 35. With this kind of technical gap in infrastructure, the U.S. can’t catch up in the short term.

Even more crucial is that AI’s inherently high power-consumption characteristics require that the energy itself must be clean. In 2025, China’s share of installed capacity from renewable energy first surpassed 60%. New wind and solar capacity additions exceeded 430 million kilowatts. In total electricity consumption, nearly 40% comes from green energy. By contrast, in the U.S., debates over delays in building nuclear power plants continue endlessly, while China’s solar PV and wind power have already achieved grid parity—providing an economical and sustainable solution for energy-intensive AI data centers.

There’s another detail that’s easy to overlook. China is the global manufacturing hub for transformers, with production capacity accounting for over 60% worldwide. What is the biggest bottleneck in modernizing the U.S. power grid? A transformer shortage—lead times have been stretched to 3 to 4 years. U.S. power grid maintenance, whether via Mexico as a transit point or through direct procurement, relies heavily on Chinese manufacturing. When U.S. data centers are forced to shut down due to a lack of transformers, Chinese electrical equipment companies are operating at full capacity—supporting the rapid expansion of domestic computing infrastructure.

Behind every precise move by those agile robots on the Spring Festival Gala—including those impressive atom robots—there isn’t just advanced algorithms. There is also stable current transmitted from thousands of kilometers away through ultra-high-voltage lines. This isn’t only a feast of robots—it’s a concentrated demonstration of China’s industrial capability.

From an investment logic perspective, the U.S. has the most advanced algorithms, but China controls the strongest energy conversion and transmission systems. In this AI gold rush, if NVIDIA is selling picks and shovels, then China’s infrastructure builders—UHV transmission, electrical equipment, and green energy—are the ones who truly control the water source. This difference will become increasingly obvious in the coming years.