The fight over AI chips is no longer just a business story or a defense story. Huawei’s rotating chairman Xu Zhijun has now said out loud what many in Washington worried might happen, that U.S. export pressure pushed China’s semiconductor industry to grow faster and build more of its own technology stack.
That matters far beyond boardrooms and government briefings. AI chips sit inside data centers, and data centers sit on real land, draw real power, and use real water. The International Energy Agency projects global data center electricity use will roughly double to about 945 terawatt-hours by 2030, with the United States and China accounting for nearly 80% of that growth.
Huawei’s unexpected thank you
Xu’s message was striking because it turned a punishment into a talking point. According to reports based on his recent interview, the Huawei executive said the company was “grateful” for U.S. pressure because it forced China’s chip industry to grow its own supply chain.
“If the United States hadn’t forced our country, our companies, and our industry, we wouldn’t have done something like this,” Xu said, referring to Huawei’s new chip ideas. “But we are also grateful to the U.S. for enabling our country’s semiconductor industry chain to truly grow.”
That is not a small claim. Huawei was added to the U.S. Entity List in 2019, when the Commerce Department said the company and dozens of affiliates posed national security or foreign policy concerns.
From sanctions to self-reliance
The bigger shift came in 2022, when U.S. rules targeted advanced computing chips, supercomputer uses, and semiconductor manufacturing equipment tied to China. Washington said the controls were aimed at national security risks, including military modernization and advanced AI systems.
In practical terms, that made it harder for Chinese companies to buy the most advanced Nvidia and AMD AI hardware. Later restrictions hit China-focused chips too. Nvidia disclosed a possible $5.5 billion charge tied to H20 products, while AMD said export controls on MI308 products could lead to about $800 million in charges.
But pressure cuts both ways. When imported chips are harder to get, local alternatives become more attractive, even when they are not as powerful or efficient. That is where the story moves from geopolitics to the grid.
The energy catch
AI is often described as software, but it runs on a physical machine. Servers heat up. Cooling systems kick in. Power plants and grids carry the load.
United Nations researchers said data centers consumed 448 terawatt-hours of electricity in 2025, with AI accounting for about one-fifth of that total.
They also estimated data centers used about 1.2 trillion gallons of water and generated about 208 million U.S. tons of carbon dioxide emissions. By 2030, annual power use is projected to reach 945 terawatt-hours, while water consumption could approach 2.5 trillion gallons.
So, what happens if China builds a larger domestic AI chip industry? The answer is not automatically cleaner or dirtier. More domestic chips could make AI cheaper and easier to deploy, but if those chips use more electricity for the same job, the environmental bill may rise quickly.
LogicFolding enters the picture
Huawei’s answer is not simply to copy the old playbook. In May, the company presented its Tau Scaling Law at the IEEE International Symposium on Circuits and Systems, describing it as a path beyond traditional transistor shrinking.
The company says technologies such as LogicFolding can reduce signal delay and improve transistor density. Huawei also says the method is meant to improve performance and energy efficiency across devices, circuits, chips, and full systems.
Its first Kirin chips using LogicFolding are scheduled for fall 2026, while high-end chips based on the approach are expected to reach density equivalent to a 1.4 nanometer process, about 55 billionths of an inch, by 2031.
That sounds promising. Still, outside observers are cautious. Reuters noted that Huawei hopes the approach can reduce the time and energy spent moving data between and within chips, but it remains unclear how challenges such as heat will be solved at scale.
Nvidia’s warning
Nvidia CEO Jensen Huang has long argued that cutting China off from U.S. chips could backfire. His view, in plain English, is that if Chinese companies cannot buy American hardware, they will have a stronger reason to build their own.
Recent market data suggests the picture is messy. Huang has warned that Nvidia’s China position was being wiped out, but IDC data reviewed by Reuters showed Nvidia still shipped 2.2 million AI accelerators to China in 2025, equal to a 55% market share. That is far from the near-monopoly it once had, but it is not zero either.
Washington has also changed course more than once. In January 2026, the Bureau of Industry and Security said it would review licenses for Nvidia H200, AMD MI325X, and similar chips on a case-by-case basis for China, provided security conditions were met.
A greener chip race?
Here is the tricky part. A chip race can push efficiency forward, but it can also push everyone to build more, faster. That is great for AI capacity, but not always great for local power grids, water systems, or carbon targets.
For the most part, the environmental outcome will depend on two things. First, whether new Chinese chips can deliver useful performance without wasting too much power. Second, whether data centers running those chips use cleaner electricity and smarter cooling systems.
At the end of the day, export controls may have done more than reshape the semiconductor market. They may have helped create a second AI hardware ecosystem, one that will now compete with the U.S. not only on speed and price, but also on energy use.
The official statement was published on Huawei.
