Cambricon Surges as Nvidia Export Curbs Reshape China's AI Chip Market

A 4,300% revenue jump puts Cambricon in the spotlight

Cambricon Technologies reported revenue of 2.88 billion yuan, about $402.7 million, for the first half of 2025, a rise of 4,300% that has placed the Beijing chipmaker at the center of China's effort to build its own artificial intelligence industry. Its net profit reached a record 1.04 billion yuan as restrictions on American chips pushed Chinese buyers toward domestic suppliers.

By August 30, 2025, the partially state-owned company's market value had more than doubled during the year to around $80 billion. Its shares had climbed 125% in one month and more than 1,000% in two years. Cambricon also overtook liquor producer Kweichow Moutai as China's most expensive stock by share price.

Nvidia, the world's leading maker of AI chips, faces a different problem in China: customers want computing power, but American export controls have blocked access to some of its products. Nvidia reported no H20 sales to China in the second quarter of 2025, losing access to billions of dollars in potential business.

The contrast captures a central tension in Washington's technology policy. Restrictions intended to slow China's progress in advanced computing are also giving Chinese chip designers a larger market at home. Beijing's support for domestic technology is helping turn that opening into orders, investment and rising expectations.

Cambricon remains far smaller than Nvidia, and China's manufacturers still lack access to some of the equipment needed to make the most advanced chips. Its sudden growth nevertheless shows how decisions about export licenses can change which companies supply the hardware behind AI.

Why Chinese buyers are turning to local chips

Cambricon's rise follows China's urgent push for technological self-sufficiency. Founded by two brothers and backed by state capital, the company has benefited as Chinese technology groups and startups search for alternatives to American AI hardware.

AI accelerators are specialized chips that perform the large numbers of calculations needed to train and run AI models. Training teaches a model to recognize patterns in data. Running the model means using that training to produce an answer or carry out a task. For companies building AI services, access to these chips directly affects the computing resources they can deploy.

Nvidia has dominated this market for years. Its position rests on powerful hardware, extensive research and software that developers already know how to use. Export restrictions have made access to that established system less dependable for Chinese customers, creating an opening for domestic competitors.

Beijing has encouraged companies to prioritize homegrown technology, citing national security and the risk of supply disruptions. Chinese officials have reportedly warned local businesses against relying on American components wherever alternatives are available.

Chinese companies are responding with a mixture of old and new equipment. They continue using Nvidia chips stockpiled before the restrictions while moving more work toward domestic products, including Cambricon's. This gives local suppliers an opportunity to win customers while those customers still operate existing Nvidia systems.

How the H20 restrictions changed Nvidia's business

American export controls began under the Trump administration and continued under subsequent leaders. Their purpose was to limit China's access to the most advanced AI hardware, including technology that could support military or surveillance capabilities.

Nvidia's H20, a product with lower performance than its most advanced chips, was also caught by the restrictions. The ban was later partially lifted, but the reopening came with an arrangement requiring Nvidia to share 15% of revenue from H20 sales in China with the United States government.

The policy sequence has changed both the availability of Nvidia products and the commercial terms attached to selling them:

• Under the Trump administration and subsequent leaders, Washington imposed and maintained controls on advanced chip exports to China.
• During the first half of 2025, Cambricon's revenue rose 4,300% to 2.88 billion yuan as demand for domestic AI chips increased.
• In the second quarter of 2025, Nvidia reported no H20 sales to China amid export restrictions.
• By August 30, 2025, the H20 ban had been partially lifted, with a 15% revenue sharing arrangement adding financial and legal uncertainty.

Chinese authorities have also told technology companies to stop using H20 chips after American officials said only less advanced versions would be available to China. Nvidia therefore faces pressure from both governments: Washington controls what it can export, while Beijing is steering buyers toward local alternatives.

A 15% payment adds another layer of uncertainty

Nvidia's proposed access to China now carries a direct financial cost. Under the revenue sharing arrangement, 15% of every dollar earned from H20 sales in China would go to the American government. That operates as a tax on the business, reducing what Nvidia retains from those sales.

The arrangement has no formal United States regulation setting out the tax. Its legal basis and durability remain uncertain, leaving Nvidia to plan around terms that could shift with political decisions.

Export controls have already cost Nvidia billions of dollars through lost revenue and unsold inventory. Even when a product is designed to satisfy restrictions, changes in policy can affect whether it reaches customers and whether those customers are willing to depend on it.

Nvidia's challenge extends beyond securing permission for individual shipments. Chinese buyers must decide whether an American chip will remain available for future purchases, while domestic suppliers are competing for the same spending with support from Beijing.

Cambricon's growth still starts from a small base

Cambricon's $402.7 million in revenue for the first half of 2025 is a sharp increase from its earlier business. Nvidia, however, reported revenue of $46.7 billion in the comparison presented alongside those results, more than 100 times Cambricon's total. The figures reflect different reporting periods: Cambricon's first half and Nvidia's second quarter.

Nvidia's global business is supported by years of research, a large community of developers and partnerships that make its chips a standard choice for AI research and deployment. Cambricon's rapid expansion is taking place against that much larger commercial and technical base.

An industry analyst described the scale of Cambricon's advance in these words:

from tiny to small

Chinese competitors are still improving. Huawei and Semiconductor Manufacturing International Corporation, known as SMIC, are catching up in product quality and manufacturing capacity. Some analysts say certain Chinese chips already outperform Nvidia's H20, while restrictions are speeding efforts to develop alternatives.

The H20 is a less advanced product than Nvidia's latest hardware. For Cambricon and its peers, winning business from customers who cannot reliably obtain American chips is one task; matching Nvidia's newest technology and its global software network is a larger one.

Software is as important as the chip itself

Cambricon is investing heavily in its next generation of hardware and in better software tools. Both are needed because an AI chip has to work with the programs developers use to build and operate models.

Nvidia's advantage comes partly from this surrounding software ecosystem. Developers rely on established tools, and companies have built their work around them. A competing chip becomes more useful when customers can move their models onto it without having to rebuild large parts of their software.

Chinese AI startups, including DeepSeek, are releasing models optimized for domestic chips. Optimization means adjusting the software so that it makes better use of the hardware it runs on. Those releases help local chipmakers compete by giving buyers more practical ways to use their products.

Cambricon's opportunity therefore extends beyond individual chip sales. As more models and development tools support Chinese hardware, customers gain more reasons to adopt it. Government support, new orders and software development are reinforcing the same shift toward domestic suppliers.

Nvidia still has the stronger combination of advanced hardware, developer tools and global partnerships. Cambricon is trying to narrow that gap while demand gives it room to expand.

Manufacturing limits complicate China's ambitions

Chinese chip designers face a major production constraint: Western export controls keep the most advanced manufacturing equipment out of reach. That includes extreme ultraviolet lithography machines, usually called EUV machines.

Lithography is the process used to create the tiny patterns that form a chip's circuits. EUV equipment helps manufacturers produce extremely small features. Without access to those tools, Chinese companies face a harder task turning competitive designs into chips made at the leading edge of manufacturing.

Cambricon and its peers can improve their designs, but production capacity and manufacturing technology affect how far and how quickly those designs can advance. China's chip industry is making progress while still trailing Nvidia's latest products.

Ed Yardeni, a prominent macro analyst, described the scale of China's production ambitions and the policies supporting them:

China plans to triple AI chip production by next year, supported by policies to embed AI across economic and social sectors and strong backing for domestic firms like Cambricon.

From the perspective of August 2025, that production target points to 2026. Meeting growing demand will require more chips as well as software that allows Chinese companies to use them effectively.

Investors are pricing in a larger domestic market

Cambricon's market value has risen by more than $40 billion during 2025. The surge has drawn attention from Chinese investors, global hedge funds and technology analysts watching the shift toward domestic AI infrastructure.

Eric Jackson of EMJ Capital has highlighted Cambricon's rapid growth and the strategic importance of its new product launches. Investors are placing substantial value on the company's ability to turn China's demand for local technology into a much larger business.

The stock's rise of 125% in a month and more than 1,000% over two years shows how quickly expectations have changed. Its roughly $80 billion valuation also places considerable weight on future growth beyond the revenue it currently generates.

Nvidia remains popular with investors because of its AI leadership and strong financial results. Its expanding exposure to geopolitical risk, however, means that global demand and Chinese market access are pulling its prospects in different directions.

For both companies, investor confidence is tied to more than the next chip launch. Export decisions, domestic purchasing policies and manufacturing capacity all influence how much of China's AI spending each company can capture.

Nvidia seeks a route back into China

Nvidia sees China as a $50 billion opportunity in 2025, with potential annual growth of 50% if access is secured. Chief executive Jensen Huang has repeatedly stressed the market's importance and described the United States and China as complementary in the global AI race.

Even a partial return of Chinese sales could help offset the losses caused by export restrictions. At the same time, worsening tensions could exclude Nvidia from a Chinese data center market valued at $120 billion, weakening a major source of potential future growth.

Nvidia is developing products tailored to American export rules, including the B30A, a version of its next generation Blackwell architecture intended for China. The company hopes regulatory approval will allow it to resume H20 sales and introduce a Blackwell chip to Chinese customers.

The strategy depends on permission from Washington and demand from buyers facing pressure to choose domestic equipment. Nvidia is trying to preserve access to a large market while Chinese competitors use the disruption to strengthen their position.

Huang, Nvidia's chief executive, has argued that restrictions can also stimulate the competition they are intended to restrain:

Export controls only encourage Chinese rivals to innovate.

A contest over technology, trade and national security

Washington's policy places national security concerns alongside the commercial interests of American chipmakers. Restricting advanced exports is intended to prevent sensitive technology from supporting Chinese military or surveillance capabilities. The same restrictions are encouraging China to invest more heavily in its own suppliers.

Beijing views semiconductor self-sufficiency as a matter of national survival. Its support for domestic companies, warnings about supply disruptions and push to use AI across economic and social sectors are creating demand for firms such as Cambricon.

An industry analyst summarized the broader force behind the company's rise:

The bigger story is China's push for self-sufficiency.

Semiconductors power devices ranging from smartphones to supercomputers. Control of advanced chip technology therefore affects economic strength, military capability and the supply chains on which technology companies depend.

Cambricon's growth shows that this contest is already changing where money and orders go. Chinese chipmakers could become global contenders within a few years if current trends continue, though they still face manufacturing limits and the challenge of building software systems that rival Nvidia's.

For Chinese AI businesses, the immediate change is concrete: more spending is moving toward domestic hardware while existing Nvidia chips remain in use. For Nvidia, the task is to regain access before local alternatives become more deeply established. The competition is reshaping investment, supply chains and the development of AI in the world's two largest economies.

Key Points

• Cambricon's revenue rose 4,300% to 2.88 billion yuan in the first half of 2025, with net profit of 1.04 billion yuan.
• Its market value reached about $80 billion, while its share price overtook Kweichow Moutai's as China's highest.
• Nvidia reported no H20 sales to China in the second quarter of 2025 as export restrictions disrupted its business.
• A 15% revenue sharing arrangement adds financial and legal uncertainty to a potential resumption of H20 sales.
• Beijing is encouraging domestic chips, and companies including DeepSeek are adapting AI models to use them.
• Chinese manufacturers still lack access to advanced EUV equipment, while Nvidia retains major hardware and software advantages.
• Nvidia is pursuing compliant products, including the B30A, as it seeks access to China's growing AI market.