China's push toward technological independence is moving from policy ambition to business reality. Companies across the country are now actively developing and deploying alternatives to Nvidia semiconductors, driven less by geopolitical pressure and more by practical economics and competitive advantage. This shift signals a fundamental change in how the US-China AI race is unfolding, with implications that extend far beyond chip supply chains.

Why Are Chinese Companies Abandoning Nvidia?

The transition away from Nvidia chips is happening across multiple sectors simultaneously. Robovan startup Zelostech plans to use multiple chip suppliers from China and elsewhere over the next one to two years, moving away from relying solely on Nvidia for its self-driving systems. The primary driver is cost; using China-made chips would be significantly cheaper than the two Nvidia Orin chipsets currently used in each vehicle, according to Shi Yunjian, director of finance and investment at Zelostech.

This matters because scale creates competitive advantage. The more autonomous vehicles a company can deploy, the more operating data it collects, and the easier it becomes to convince regulators that the technology is ready for wider use. Zelostech already operates more than 25,000 vehicles across over 20 countries, far outpacing competitors like Alphabet-backed Waymo, which has just under 4,000 vehicles on the road.

The shift extends well beyond autonomous vehicles. Chinese electric car manufacturers are moving aggressively into chip development. BYD, Nio, and Xpeng have all recently revealed their own semiconductors for driver-assist systems. Nio CEO William Li stated that the company is no longer buying Nvidia chips but instead renting compute power powered by a variety of processors, while also planning a fivefold increase in spending on computing power.

How Are Chinese AI Models Adapting to Homegrown Chips?

The transition extends to the software layer as well. Chinese AI developers have increasingly optimized their models to run on homegrown hardware rather than relying on Nvidia's widely used CUDA ecosystem, which has long been the industry standard. The latest MiniMax and Kimi models, along with DeepSeek's V4, are now compatible with local Chinese semiconductors. DeepSeek V4 specifically works with eight China-made chips, including those from Huawei and Alibaba's T-head chip unit.

Goldman Sachs analysts predicted in May that the pivot to domestic chips will accelerate through 2026 and into 2028. Huawei, which has faced years of US restrictions, recently revealed that it has been using a new scientific approach to developing its chips and plans to incorporate them in upcoming products, signaling a potential comeback for the Chinese telecom giant.

Steps China Is Taking to Build Semiconductor Independence

• Vehicle Integration: Autonomous vehicle makers like Zelostech, BYD, Nio, and Xpeng are deploying Chinese-made chips in production vehicles, generating real-world performance data that improves future iterations.
• AI Model Optimization: Chinese AI developers are rewriting software to run efficiently on domestic semiconductors, reducing dependence on Nvidia's CUDA ecosystem and creating a parallel software ecosystem.
• Chip Development Acceleration: Companies like Huawei and Alibaba are investing in new chip architectures and manufacturing approaches, with plans to integrate them into commercial products within the next year or two.

What Does This Mean for US Policy Coherence?

The irony is that US export controls, intended to slow China's technological advancement, may have inadvertently accelerated the very self-sufficiency efforts they sought to prevent. Mieke Eoyang, a national security policy expert who previously served as deputy assistant secretary of defense for cyber policy, highlighted a critical problem: the United States lacks a unified strategy on China's technology development.

"We have a chaotic policymaking process in the United States right now. There seems to be a real disconnect between President Trump and other parts of his administration, as well as between the executive branch and the legislative branch. There doesn't seem to be a shared understanding of what is in the strategic interest of the United States," said Mieke Eoyang.

Mieke Eoyang, National Security Policy Expert, former Deputy Assistant Secretary of Defense for Cyber Policy

Congress has advanced the MATCH Act to restrict China's access to advanced US chip technology, while President Trump has pursued a more transactional approach. This inconsistency creates uncertainty for global customers and allies trying to plan long-term business strategies. Without clear, consistent policy direction, companies struggle to know whether to invest in US technology partnerships or pursue alternatives.

Could This Accelerate Global Economic Fragmentation?

Eoyang raised a deeper concern: if China successfully builds a completely independent semiconductor ecosystem, it could undermine the global supply chain integration that has historically served as a deterrent to armed conflict. Global chip production is inherently integrated; no single country can produce everything needed for modern electronics. A conflict that disrupts any critical piece, whether Taiwan, Japan, wafer production, or fabrication facilities, would ripple throughout the entire global economy.

"Integration of global supply chains is a deterrent to armed conflict because we all need each other for continued economic development. If international trade is no longer one of the core global underpinnings for peace, then we could wind up in a much more vulnerable place," explained Eoyang.

Mieke Eoyang, National Security Policy Expert, former Deputy Assistant Secretary of Defense for Cyber Policy

Kevin Xu, founder of hedge fund Interconnected Capital, expects Chinese companies to continue needing Nvidia chips for the next three to five years. However, he argues that Beijing has a strong incentive to limit that dependence sooner rather than later. China-made chips can only improve if companies use them in real-world scenarios, generating the feedback needed to make the technology useful for businesses. The more Nvidia chips remain in the ecosystem, the less incentive companies have to adopt and refine domestic alternatives.

Meanwhile, Nvidia's revenue from mainland China and Hong Kong is already shrinking, even as the company doubles down on Taiwan with plans to spend as much as 150 billion dollars per year there. The US chipmaker is trying to maintain a foothold in China's emerging "physical AI" sector by collaborating with Chinese humanoid startup Unitree on a research robot sold globally, and CEO Jensen Huang has reportedly joined a Tsinghua University board.

The broader message is clear: China's technological ambitions are no longer defined by access to Nvidia, but by the companies that can build without it. The question now is whether the United States can develop a coherent strategy to compete in a world where that independence becomes reality.