China is adding nearly 200GW of nuclear and solar for AI

Signals Inbox·July 20, 2026·AI Infrastructure

Larry Fink says the AI race may be decided by electricity, and China is building it at a scale few countries can match. The viral 200GW number is loose, but the underlying gap is real: China has about 40GW of nuclear power under construction and added more than 430GW of wind and solar in 2025 alone.

Q1Where does the claim come from?

The claim comes from BlackRock CEO Larry Fink, who argues that countries need huge amounts of electricity to keep scaling AI. BlackRock makes the broader infrastructure argument in its official 2026 chairman letter. For the underlying numbers, the best checks are the IAEA reactor database and data published by China’s National Energy Administration.

Q2Is China really building nearly 200GW?

Not exactly in the way the clip suggests. The IAEA currently lists about 40GW of Chinese nuclear capacity physically under construction. The much larger 125GW nuclear figure includes reactors already operating, being built, and approved. The solar side is actually bigger: China added roughly 315GW of solar during 2025 alone.

Q3So what is the real signal?

China is building the whole power system AI will need: nuclear plants for steady electricity, huge solar farms for cheap daytime power, storage, transmission lines, and industrial supply chains. The important comparison is speed. China can approve and build major energy projects while the US and Europe are still debating permits, grid connections, and who should pay.

Q4Why does AI need this much electricity?

Large AI data centers can use as much power as small cities. Training models is demanding, but serving them to millions of users also runs around the clock. More chips therefore create demand for more power plants, substations, transformers, cooling equipment, and transmission capacity. A company can buy GPUs quickly. It cannot build a new grid nearly as quickly.

Q5Does solar alone solve the problem?

No. Solar is fast and relatively cheap to install, but it does not produce steady electricity at night or during poor weather. Data centers want power every second. That is why China’s nuclear construction matters. Nuclear can provide stable power while solar lowers costs during the day, with batteries, coal, gas, and the grid filling the remaining gaps.

Q6Could this actually shift the AI race?

Yes, but not overnight. The US still leads in advanced chips, cloud platforms, and many frontier models. China’s advantage is the speed and scale of physical construction. If electricity becomes scarce, expensive, or slow to connect in the West, more compute could move toward places that can offer reliable power quickly. The AI race would then be decided partly by reactors, grids, and electricians, not only researchers.

Q7What should we watch next?

Watch where the largest AI data centers are actually connected, not just announced. The useful numbers are completed gigawatts, grid connection dates, electricity prices, reactor completion times, and how much renewable power gets curtailed. China already has scale. The next test is whether it can turn that scale into reliable, cheap compute faster than its rivals.