China still relies on Nvidia for its most advanced AI

In the middle of 2026, China continues to rely on Nvidia graphics cards to train its most advanced AI models, despite the Chinese government’s efforts to achieve technological self-sufficiency. This dependency highlights the persistent technical and economic gap that hinders the transition to local components, emphasizing the challenges China faces on its path to digital sovereignty.

Persistent Dependency

James Wang, an AI model developer at a research institute affiliated with a university in Shanghai, explains that migrating to local semiconductors such as Huawei’s Ascend chips presents a prohibitive cost. According to estimates, switching existing workflows to these local technologies could add at least an additional 50% in time and costs. This dependency reflects the complexity of the technological infrastructure and the need for compatible software, such as Nvidia’s Compute Unified Device Architecture (CUDA), which still surpasses the efficiency of local alternatives like Huawei’s Compute Architecture for Neural Networks (CANN).

Technical and Economic Challenges

Advanced AI models, known as Large Language Models (LLMs), require high-performance infrastructure for their training. Nvidia has been a leader in this field, offering robust and efficient solutions that have facilitated the development of these models. However, this dependency also means that China is limited by U.S. export restrictions and policies, further complicating its strategy for technological self-sufficiency.

Huawei Technologies and Alibaba Group, two of China’s leading tech companies, have been working on local solutions to reduce this dependency. However, progress has been slow due to technical and economic challenges. The costs associated with transitioning to new technologies and the need to adapt existing software represent significant barriers.

Future Prospects

China’s continued reliance on Nvidia graphics cards for its most advanced AI underscores the complexity of achieving digital sovereignty. Although the Chinese government has launched numerous initiatives to promote local technological innovation, the technical and economic gap remains evident. To overcome this dependency, a concerted effort from both businesses and the government will be necessary to develop local solutions that are both efficient and accessible.

In the future, China’s ability to reduce its dependence on foreign technology in the AI sector will be crucial for its digital sovereignty strategy. This will require not only technological advancements but also coherent policies that encourage innovation and adaptation by local enterprises. The transition to autonomous technological infrastructure will not be easy, but it is an essential step to ensure China’s independence and security in the realm of AI.

In conclusion, China’s reliance on Nvidia graphics cards for its most advanced AI reflects the technical and economic challenges it faces on its path to technological self-sufficiency. As it moves forward, it will be crucial for the country to continue working on the innovation and adaptation of its own technological solutions to ensure its independence in the AI sector.