^{Photo: South China Morning Post}

China is doubling down on artificial intelligence — and its strategy is as vast as it is unconventional. While American companies dominate cutting-edge semiconductor design, China is leaning on an entirely different advantage: scale, energy, and self-reliance.

With Huawei’s powerful Ascend chips, massive high-performance clusters, and an abundance of cheap, government-subsidized energy, Beijing is crafting an alternative AI ecosystem designed to compete head-on with the United States despite ongoing sanctions.

Building an AI Empire Without Nvidia

For years, Nvidia’s GPUs have been the industry gold standard for training and running AI models. But after U.S. export controls restricted the sale of advanced Nvidia chips to China, Beijing was forced to look inward.

The result is a homegrown solution spearheaded by Huawei, whose Ascend series has become the cornerstone of China’s AI infrastructure. While Huawei’s chips don’t match Nvidia’s raw efficiency on a one-to-one basis, the company has mastered the art of scaling — linking hundreds of its processors into massive computing clusters that collectively rival U.S. systems.

A prime example is Huawei’s CloudMatrix 384, which connects 384 Ascend 910C chips to achieve performance comparable to Nvidia’s GB200 NVL72 system, which uses just 72 GPUs. This design requires far more hardware and energy — but it works, thanks to China’s unique advantages in energy production and industrial policy.

The Secret Weapon: Cheap Power and Generous Subsidies

The trade-off for Huawei’s “scale over sophistication” approach is high power consumption. Each additional chip adds to the energy load — but that’s where China’s energy policy becomes a strategic weapon.

China has spent years investing in renewable and nuclear energy, allowing it to produce electricity at some of the world’s lowest industrial rates. Cities such as Shanghai, Shenzhen, and Chengdu now offer energy subsidies and “computing vouchers” to offset electricity costs for AI-focused data centers, especially those using domestic chips.

Some local governments have gone even further, cutting electricity bills by as much as 20–30% for companies running domestic hardware, according to recent financial reports. This policy not only helps sustain Huawei’s energy-hungry clusters but also strengthens the national push for technological self-sufficiency in the face of U.S. sanctions.

“China has built the perfect ecosystem to support large-scale AI infrastructure,” said one analyst. “Even if their chips are less efficient, cheap power and government support make up the difference.”

A Race Against U.S. Tech Dominance

China’s AI strategy reflects a broader goal: independence from Western technology. Despite heavy U.S. export restrictions, Chinese giants such as Alibaba, Baidu, and DeepSeek have continued to roll out competitive AI models trained on domestic chips.

Huawei’s chips are produced by Semiconductor Manufacturing International Corp. (SMIC), China’s leading foundry. Although SMIC is several generations behind Taiwan’s TSMC, it has managed to manufacture 7-nanometer Ascend 910 chips using older lithography tools. The process is costly and less efficient — but it allows Huawei to maintain production volumes large enough to compete through scale.

Still, China’s long-term challenge remains clear: SMIC’s technological limitations stem from its inability to acquire extreme ultraviolet (EUV) machines made by Dutch company ASML, which are crucial for producing advanced chips. This gap could widen as Nvidia and TSMC move toward 3-nanometer and even 2-nanometer technologies in the coming years.

Self-Sufficiency Over Perfection

While China may not be able to match the precision of American chipmaking soon, it’s building something equally powerful — an ecosystem that doesn’t rely on the U.S. at all.

The country’s AI superclusters, supported by cheap energy, state funding, and domestic manufacturing, ensure that Chinese AI labs and startups can continue to train large models without Western hardware. Analysts estimate that China’s AI compute capacity, measured in petaflops, has grown more than 40% year-on-year, making it one of the world’s largest markets for high-performance computing.

However, as AI becomes increasingly data- and compute-intensive, questions remain about whether this energy-heavy model is sustainable. Environmental pressures and rising power demands could push Beijing to innovate in chip efficiency and energy management sooner than expected.

The Road Ahead: Competing on Scale and Strategy

For now, Huawei’s approach — combining massive chip clusters with abundant energy — keeps China competitive. But the global AI race isn’t just about who can compute more; it’s about who can innovate faster.

To bridge the gap with the U.S., China is accelerating investment across its semiconductor ecosystem, from advanced chip design and optical interconnects to AI software optimization. Beijing has already earmarked tens of billions of dollars in its upcoming Five-Year Plan to support AI infrastructure, semiconductor R&D, and power-efficient computing technologies.

As one industry observer put it: “Nvidia is winning the efficiency battle. But Huawei — and by extension, China — is winning the scale war.”

In essence, China’s AI strategy is a story of adaptation. By leveraging cheap energy, government support, and massive parallel chip systems, Beijing has found a way to stay in the global AI race even without access to the world’s best chips. While the U.S. continues to lead in chip design, China’s ability to build, scale, and sustain AI infrastructure at national scale may ultimately prove to be its most powerful weapon.