China's artificial intelligence chip industry is experiencing a dramatic shift from defensive strategy to offensive infrastructure buildout, with domestic chipmakers now winning substantial orders as Beijing pushes local tech firms to abandon American silicon. The country's AI chip designers are expecting record earnings this year, driven by government mandates for homegrown components and a massive five-year, $295 billion data center investment plan that will rely on domestic suppliers for at least 80 percent of technology.

Why Are Chinese Companies Suddenly Switching to Domestic AI Chips?

For years, China's AI labs relied on Nvidia's graphics processing units (GPUs), which are specialized computer chips designed for processing large amounts of data simultaneously. But after the United States restricted access to Nvidia's most advanced processors, Beijing took action. In response to these export controls, the Chinese government ordered local technology firms to find alternatives to Nvidia's cutting-edge products. This mandate has created an immediate surge in demand for domestic silicon.

The timing is significant. Although the U.S. recently cleared sales of Nvidia's H200 processor, the most powerful AI chip ever allowed to be sold to China, U.S. officials reported that the actual number of chips shipped to the country has been "trivial". This scarcity has forced Chinese companies to accelerate their reliance on homegrown alternatives.

According to a Bloomberg Intelligence survey of Chinese technology executives conducted in June 2026, companies expect to spend 46 percent of their AI accelerator budget in the next 12 months acquiring locally-made chips, compared to just 30 percent currently. This represents a dramatic 53 percent increase in domestic chip spending within a single year.

Which Chinese Chipmakers Are Winning the Most Orders?

Several domestic chip designers are positioned to benefit from this shift. Cambricon Technologies is projected to post surging revenue for the first half of 2026, while Shanghai Iluvatar CoreX Semiconductor is expected to see sales triple over the same period. Beijing-based Moore Threads Technology reported in July that it expected revenue for the first six months of 2026 to grow as much as 149 percent.

Huawei, the privately-held telecommunications giant, is sharply ramping up production of its Ascend 910C chips, aiming to produce approximately 600,000 units in 2026. Alibaba Group's chip unit, T-Head, is also expected to boost its domestic market share. These companies are now competing directly with Nvidia and Advanced Micro Devices (AMD), which are virtually locked out of the Chinese market due to export restrictions.

The competitive landscape includes several key players working to capture market share:

• Huawei Ascend: Planning to produce 600,000 of its 910C chips in 2026, representing a major scaling effort for the company's flagship AI accelerator.
• Cambricon Technologies: Expecting surging revenue in the first half of 2026 as Chinese firms evaluate its AI accelerators for deployment.
• Moore Threads Technology: Projecting revenue growth of up to 149 percent for the first six months of 2026, driven by increased domestic demand.
• Shanghai Iluvatar CoreX Semiconductor: Anticipating sales to triple over the first half of 2026 as orders accelerate.
• Kunlunxin (Baidu's chip unit): Targeting a $50 billion Hong Kong IPO, representing a 17-fold valuation increase from December 2025.

How Are Chinese Labs Building Gigawatt-Scale Infrastructure Around Domestic Chips?

Beyond individual chip sales, Chinese AI labs are now constructing massive data centers powered entirely by domestic silicon. Z.AI, the Beijing company formerly known as Zhipu, has completed construction of a one-gigawatt AI data center powered exclusively by Chinese-made chips, with partial operations already underway. To put this in perspective, one gigawatt is roughly the electricity draw of 750,000 homes at any given moment.

The facility is built around several computing clusters, each containing more than 10,000 chips, making it one of the largest server hubs ever assembled by a Chinese AI lab. This represents a critical milestone: Z.AI is no longer treating domestic chips as a fallback option but is building frontier AI infrastructure around them. The facility will train next-generation versions of Z.AI's GLM language model, a large language model (LLM) that powers conversational AI applications.

This infrastructure shift signals a broader change in how Chinese AI companies approach hardware. Rather than viewing domestic chips as inferior alternatives, companies are now designing entire data center architectures around them. Chinese chipmakers have concluded that they can compensate for performance gaps in individual chips by deploying them at massive scale, linking tens of thousands of chips together through optimized networking and software.

"What they are trying to reassure is that we have enough chips and we have the infrastructure to link all these chips together so we can outnumber you. It also demonstrates how China, given enough time, will eventually narrow the gap to much narrower levels than today," said Phelix Lee, analyst at Morningstar.

Phelix Lee, Analyst at Morningstar

What Does This Mean for China's Long-Term AI Chip Independence?

The implications are substantial. Beijing is preparing to spend roughly 2 trillion yuan, approximately $295 billion, on data centers over the next five years, with Alibaba and China Telecom among the biggest builders. This massive investment will rely on local suppliers including Huawei for at least 80 percent of technology such as AI chips.

According to estimates from Morgan Stanley, this buildout will push China's self-sufficiency level in AI chips to 70 percent by the end of this decade, up from 42 percent in 2025. This represents a 67 percent increase in domestic chip self-sufficiency within five years.

However, Chinese chipmakers still face significant technical challenges. China-made chips lag behind Nvidia's offerings in single-unit performance, software maturity, and cluster stability at the frontier of AI development. This performance gap is partly the result of a years-long U.S.-led campaign that restricts the most advanced chipmaking equipment, including extreme ultraviolet (EUV) lithography systems made by Dutch company ASML, from flowing to China. These tools are essential for manufacturing the world's most powerful AI accelerators.

Despite these limitations, Chinese chipmakers are winning orders by focusing on deployment economics rather than absolute peak performance. As customers shift from using chips for training AI models to integrating them into their businesses, Chinese chipmakers are winning orders from their ability to generate tokens, the small units of text that language models process, at competitive pricing.

"Purchasing decisions are increasingly driven by deployment economics rather than absolute peak silicon performance," wrote Charlie Chan and other Morgan Stanley analysts in a research note at the end of July.

Charlie Chan, Analyst at Morgan Stanley

The race is now officially underway. Z.AI has already completed its gigawatt-scale facility and begun partial operations, while competitors like DeepSeek are still building out their infrastructure plans. Chinese chipmakers are no longer asking whether they can replace Nvidia; they are demonstrating that they can build the infrastructure to do so at scale.