China tipped to dominate AI models by 2030
The forecast identifies large language models, which underpin generative AI applications, as an area where China is particularly well positioned to establish leadership. It also anticipates that Chinese companies and research institutions will become at least competitive across other critical technology sectors, potentially reshaping the balance of technological capabilities between Beijing and Washington.
The assessment comes as Chinese AI developers narrow the performance gap with American competitors despite restrictions on access to advanced semiconductors. Improvements in model efficiency, expanding domestic computing infrastructure and competitive pricing have strengthened China’s position in the global market.
Bloomberg Intelligence senior analyst Robert Lea identified a sharp improvement in Chinese model performance earlier this month. Leading Chinese systems trailed their American counterparts by approximately 3% on benchmark scores, compared with around 9% in May and 15% earlier this year.
That narrowing gap followed the September introduction of DeepSeek’s V4.1 Flash model, which demonstrated further advances in reasoning and computational efficiency. The findings suggest that restrictions on sophisticated American chips have not prevented Chinese developers from improving their systems.
The projections nevertheless remain conditional on technological progress, investment and commercial adoption over the next four years. Competitive performance in individual benchmarks does not necessarily establish overall leadership, particularly when computing capacity, profitability and enterprise deployment are considered.
China’s expanding capabilities are already influencing international adoption patterns. Models developed by DeepSeek, Alibaba and Moonshot AI have attracted developers seeking alternatives to comparatively expensive proprietary systems offered by American technology companies.
Data from OpenRouter, a platform providing access to multiple AI models, showed Chinese systems gaining substantial shares of usage during 2026. However, the figures represent activity through that platform rather than the entire global market, where direct subscriptions and enterprise contracts remain significant.
Chinese developers have also benefited from distributing models with openly available weights, allowing businesses and programmers to customise systems without depending entirely on proprietary services. This approach has encouraged international experimentation and strengthened adoption among cost-conscious users.
American companies retain important advantages, particularly in advanced semiconductor access, large-scale computing infrastructure and commercial relationships with enterprise customers. OpenAI, Anthropic and Google continue investing heavily in frontier models and the infrastructure required to train increasingly sophisticated systems.
Financial performance illustrates the distinction between technological competitiveness and commercial success. Anthropic disclosed revenue of $16.3 billion for the first half of 2026, compared with $1.2 billion during the corresponding period a year earlier, reflecting strong demand for its enterprise services.
Chinese developers face greater pressure on pricing, with intense domestic competition limiting opportunities to convert growing usage into sustainable earnings. Lower-cost models can accelerate adoption while simultaneously reducing margins available to finance future development.
Semiconductor manufacturing remains another constraint. Washington’s export controls restrict Chinese access to certain advanced processors and manufacturing equipment, although domestic suppliers including Huawei are developing alternatives intended to support national computing requirements.
China’s response increasingly combines domestic chip development with software optimisation. Researchers have demonstrated that improvements in training methods, architecture and inference efficiency can reduce dependence on the most powerful available hardware.
Beijing is also pursuing greater integration between artificial intelligence and telecommunications infrastructure. A three-year programme issued by the Ministry of Industry and Information Technology in June established targets for autonomous networks, computing access and specialised AI applications through 2028.
The programme envisages significant breakthroughs in core technologies by 2030, alongside improved coordination between communications networks, computing resources and intelligent services. Its objectives complement broader efforts to establish domestic technological capabilities.
Regulatory developments could influence the international availability of Chinese technology. Authorities proposed export restrictions this month covering 23 categories of AI technology, including language models, autonomous agents, chip designs and computing clusters.
The draft measures would potentially subject transfers of certain model weights, source code and advanced technologies to government approval. Their final scope and implementation remain undetermined.
China’s Ministry of Commerce has presented the proposals as part of a wider effort to regulate strategically sensitive technology transfers. The consultation also covers technologies beyond artificial intelligence, reflecting the government’s expanding attention to intellectual property and national security.
For overseas developers using Chinese models, the eventual rules could affect access to advanced systems and associated technical resources, depending on the restrictions adopted and the thresholds applied to particular technologies.
