Data centers are electricity devouring entities, consuming vast amounts of power to support their operations. A single training cluster can expend more energy than 100,000 homes . While the spotlight is often on the ongoing semiconductor war , the real issue persists quietly in the background: the relentless demand for electricity .
Why is it important? Training the most powerful AI models requires dedicated supercomputing resources running continuously for weeks on end. Your AI capabilities are limited if you have the best chips available—but lack the electricity necessary to operate them. It’s akin to owning a high-performance Ferrari without sufficient fuel; no matter how powerful the machine, it will remain dormant without the energy to drive it.
In figures:
The context: The United States still maintains an undeniable dominance in the field of Artificial Intelligence (AI), controlling 75% of the world’s computing capacity. However, American companies face significant delays—sometimes stretching into years—when trying to connect new data centers to the electricity network . In stark contrast, China can construct and connect new plants in just a few months. This disparity in speed is a major bottleneck for the U.S. in scaling up its AI capabilities .
Tech giants like Google , Microsoft , and Amazon are experiencing prolonged delays while waiting to connect their data centers to the electrical grid. The sluggishness in expanding electrical infrastructure could hinder America’s competitive edge in AI. Conversely, China is actively developing its “National Integrated Computer Network,” which integrates both public and private data centers. Their ambitious “East Data, West Computing” plan aims to build eight major hubs in provinces rich with cheap renewable energy. This resembles trends seen in cryptocurrency mining where data centers are positioned where energy resources are most abundant.
Between the lines: Recent restrictions on chips by the U.S. have challenged China to become more efficient, as evidenced by the Deepseek earthquake earlier this year, where they successfully demonstrated rapid adaptability. While China has made significant strides, it still relies on coal for 58% of its energy mix. Despite an accelerating transition to renewable sources, the reliance on fossil fuels raises questions about the long-term sustainability of its energy strategy.
And now what? The ongoing AI race will be won not just in laboratories but also in pivotal power plants . China is laying the groundwork to support expansive AI models with adequate energy resources. Though the U.S. may have superior chips and technology, its lagging electrical infrastructure could turn into a significant drawback. For AI development, two essential components are required:
- Silicon — representing the hardware side of technology.
- Electric muscle — denoting the power needed to drive AI operations.
China is doubling down on energy production and management, positioning it as a potential victor in this high-stakes global competition.
As the landscape evolves, it is becoming increasingly clear that the fundamental resource, electricity, is as significant as technological advancements in hardware. The available energy capacity will ultimately dictate who can execute their ambitious AI plans most effectively.
With the drastic differences in approach between the U.S. and China regarding energy infrastructure, the implications are profound not only for technological advancements but also for geopolitical dynamics. As both nations move forward, the ability to harness energy for the next generation of AI capabilities will play an increasingly pivotal role. The race is on, and the *power grid* is at the heart of this high-stakes competition.
Outstanding image | ダモ リ in Unsplash
For a deeper understanding of China’s strategic maneuvers in technology independence, see “China’s three master moves to ‘independently’ focus on raw materials, chips, and AI” on Xataka.