China's AI strategy built for world beyond America, emphasizing domestic silicon and export-controlled alternatives.

**Note:** The source text appears truncated at the end (cuts off mid-sentence: "Developers outside China can change syst"). I've cleaned the complete portion below.

---

China may not own the world's best AI model for long stretches of time. It may not need to. Its stronger wager is that the model used by a factory, ministry or start-up will count for more than the model topping a benchmark. Beijing is therefore backing systems that are capable, inexpensive and available for others to modify. The United States sells access to intelligence. China increasingly distributes the machinery.

Stanford University's 2026 AI Index found that the performance gap between leading American and Chinese models had narrowed to 2.7% by March. American laboratories still produced more top-tier models; China led in publications, citations, patents and industrial robot installations. The frontier remains largely American; the ground beneath it is becoming Chinese.

Alibaba's Qwen, DeepSeek, Moonshot AI's Kimi and models from Z.ai are now principal building blocks of the open-weight market. Chinese model downloads overtook American ones in 2025. Though Qwen and DeepSeek account for much of this increase, the breadth of China's field matters more than any single laboratory's success. Baidu, once committed to closed models, has also released weights from some of its systems.

An open model attracts developers, derivative models, cloud customers and applications. Each adoption strengthens surrounding tools, documentation and technical habits. Chinese providers may not collect fees per download, but Chinese cloud platforms, chips and engineering services can follow. Android demonstrated how software given freely at one layer could secure commercial power elsewhere. Beijing has studied such precedents.

American companies entered the generative AI boom with the strongest models, most advanced chips and enormous capital. They chose a business model consistent with those advantages. OpenAI, Anthropic and Google sell controlled access through subscriptions and APIs. Customers receive a service while providers retain model weights and can adjust prices, conditions or availability.

That arrangement suits companies accepting dollar-denominated charges and dependence on an American supplier. Governments are less comfortable. A ministry handling tax records or medical data may want the model on its own servers. A business in Indonesia, Brazil or Kenya may prefer a system adapted to local languages without sending every query abroad. Chinese open-weight models address these objections at prices few domestic alternatives can match.

US export controls helped produce this outcome. Denied unrestricted access to Nvidia's most advanced chips, Chinese laboratories used computing power more frugally. DeepSeek's early 2025 success demonstrated that efficiency in training and inference could narrow part of the hardware disadvantage. Chinese developers subsequently competed through sparse architectures, smaller specialised models and lower token prices.

Washington raised China's cost of obtaining frontier chips while giving Chinese firms incentive to build models running on cheaper hardware. That quality travels well.

The Chinese state supplied much supporting structure. Local governments subsidised computing centres. State-owned enterprises purchased domestic technology. Universities and public laboratories provided researchers while venture funds absorbed risks private capital might reject. RAND describes this as "full stack" industrial policy covering research, talent, compute and applications.

Results included waste. Some local governments built computing centres unsuited to current workloads; 2025 reports suggested high idle capacity at several facilities. Subsidies financed duplication and weak firms. China's advantage does not arise from flawless planning but from tolerance for costly experiments until a few achieve scale.

China's policy differs most sharply from America's in industry's assigned role. The State Council's "AI Plus" programme directs technology towards scientific research, manufacturing, consumer products, public services and government. The 15th Five-Year Plan for 2026–30 adds AI agents, multimodal systems, embodied AI and swarm intelligence. Beijing expects AI-related industries to exceed 10 trillion yuan by 2030.

This target should be treated as official ambition rather than forecast. The industrial connection is nevertheless substantial. China already has the world's largest manufacturing base and installs more industrial robots than any other country. Its companies produce drones, electric vehicles, batteries, machine tools and telecommunications equipment. AI can be trained and tested inside these systems, generating data from machines operating outside laboratories.

A March 2026 US-China Economic and Security Review Commission report described two reinforcing loops. Open models encourage wider industrial use. Factories and robots then generate specialised data improving the models used in production. The loop is difficult to reproduce in an economy that has surrendered much manufacturing depth.

American frontier laboratories attempt increasingly general and autonomous systems. China attempts placing adequate intelligence inside millions of machines. The first course may produce larger scientific breakthrough. The second could generate earlier gains in productivity and exports.

Energy also favours China less obviously. Training and running AI systems require dependable electricity, transmission capacity and cooling infrastructure. China builds power plants, grids and data centres faster than most democracies. Its dominance in solar modules, batteries and power equipment gives it control over much of the physical supply chain behind new computing capacity. Restrictions on advanced chips still hurt. Electricity shortages are less likely.

Beijing carries its domestic model abroad. At the 2026 World Artificial Intelligence Conference, Xi Jinping offered 5,000 AI training and seminar places to developing countries over five years. China proposes cooperation centres with BRICS, ASEAN, the African Union and Latin American countries. Its Global AI Governance Action Plan calls for greater infrastructure access, open-source technology and training.

The offer addresses developing countries' complaints well. Most supplied data and users to the digital economy but own little compute, few large models and no global platforms. Western AI safety arguments can sound like attempts to preserve technological hierarchy. China offers usable models and engineers.

There echoes the Belt and Road Initiative, though AI travels more easily than railways or power stations. A government can download a model before signing financing agreements. Chinese companies can provide cloud capacity, fine-tuning, cybersecurity tools and training afterwards. Dependence develops through compatibility and accumulated technical choices rather than sovereign debt.

Beijing also understands attendance's value. Chinese officials and technology executives have become prominent in UN discussions on standards and AI governance. China previously influenced international work on facial recognition, surveillance technology and automated vehicles. When senior American officials and frontier laboratories stay away, China needs not defeat opposing proposals. It provides the draft.

Standards are rarely neutral once products are built around them. Technical definitions decide which systems can be certified, which data must be retained and whose equipment can connect. A country supplying both model and standard gains advantage that does not appear on benchmark tables.

Concerns about Chinese models cannot be dismissed as American protectionism. China's 2023 interim rules require public generative AI services to uphold "core socialist values". They prohibit content subverting state power, damaging national unity, encouraging separatism or harming the country's image. Tests of DeepSeek found evasions and official formulations on Tiananmen Square, Xinjiang, Taiwan and Liu Xiaobo.

The claim that every downloaded Chinese model is immutable propaganda goes too far. Open weights can be fine-tuned. [*Source text ends here.*]