Why is China’s AI push gaining ground? | Explained
Chinese AI firms have narrowed the gap with leading U.S. models even though Washington restricts export of advanced AI chips to China.
Models such as Moonshot's Kimi, DeepSeek, Alibaba's Qwen, and Zhipu's GLM series are gaining global adoption, with Chinese systems ranking among the most-used models on international platforms tracking model downloads and API usage.
A defining feature of this push is that most flagship Chinese models are released as open-weight (their trained parameters are published for anyone to download and run), unlike the closed, subscription-gated models from leading U.S. labs.
Chinese authorities have publicly promoted open-source AI as a tool for technology access in developing countries, positioning it as an alternative to costlier proprietary Western systems.
Despite rapid revenue growth, several Chinese AI companies are posting steep losses, raising questions about the financial sustainability of the current expansion.
Open-Weight (Open-Source) AI Models
An open-weight AI model publishes its trained parameters publicly, letting any developer download, run, fine-tune, or build products on top of it without paying licensing fees to the original creator — as distinct from closed models accessed only through a paid API. This lowers the cost of entry for developers and countries with limited compute budgets, since they can self-host the model instead of depending on a foreign provider's servers and pricing.
Key Details
- Open-weight release does not necessarily mean the full training data or code is disclosed — "open" here refers mainly to model weights, not full reproducibility.
- Chinese large language models from firms including Alibaba, Moonshot, and Zhipu have topped global usage rankings on model-hosting platforms, competing directly with closed U.S. systems on cost and latency.
- Enterprises increasingly weigh cost, response speed, and control over data alongside raw benchmark performance, a shift that favors cheaper open-weight alternatives.
China's AI gains are attributed largely to its firms embracing the open-weight route at a moment when leading U.S. labs keep their frontier models closed, letting Chinese models fill the gap for cost-sensitive developers and governments worldwide.
U.S. Semiconductor Export Controls on China
Since October 2022, the United States has progressively tightened export controls on advanced AI chips and chip-making equipment to China under the Export Administration Regulations, aiming to slow China's access to the high-end processors (such as Nvidia's most advanced GPUs) used to train frontier AI models. These controls have been revised multiple times, including further curbs on lower-spec chips originally designed to comply with earlier rules.
Key Details
- The restrictions target both finished advanced chips and the equipment/software needed to manufacture them domestically, aiming to constrain China's semiconductor self-sufficiency.
- Chinese firms have responded by optimizing AI models to run efficiently on less powerful, domestically available hardware, and by prioritizing algorithmic efficiency over raw compute scale.
- The controls are one instance of "chokepoint" technology strategy, similar to controls on EUV lithography equipment exports.
The article's central paradox is that chip restrictions, meant to slow China's AI progress, may have pushed Chinese developers toward leaner, more efficient, and more shareable open models — indirectly helping Chinese AI expand its global footprint even as it faces a hardware disadvantage.
Technology Diplomacy and the Global South
Major powers increasingly use technology standards and access as instruments of soft power and strategic influence, extending an established pattern (seen earlier in telecom infrastructure and digital payment systems) into artificial intelligence. Offering free, self-hostable AI models to developing economies builds dependency on a given country's technology ecosystem and standards.
Key Details
- Chinese officials have framed open-source AI at international forums as a means for developing nations to access advanced technology without dependence on expensive, restrictive Western vendors.
- This mirrors broader Chinese technology-diplomacy patterns of offering infrastructure and digital public goods to expand geopolitical influence.
- For India, this raises questions about balancing adoption of low-cost open AI tools against strategic and data-security considerations, alongside its own IndiaAI Mission push for domestic AI capacity.
China's promotion of open-weight models at global forums is not just a technology strategy but a geopolitical one, aiming to build long-term technological influence over countries seeking affordable AI infrastructure.
- U.S. chip export controls on advanced AI processors to China have been in place and progressively tightened since October 2022.
- Chinese AI firms' models have ranked among the most-downloaded/most-used on major global model-hosting and API-routing platforms in 2026.
- At least one major Chinese AI firm reported triple-digit percentage revenue growth in 2026 alongside a sharp rise in net losses, reflecting continued heavy investment in the sector.
- Chinese leadership has publicly championed open-source AI at international technology forums as a strategy for global technology access.