Beijing exports full-stack AI: open weights, state diplomacy, and Ascend silicon

Beijing is coordinating its open-weight models, multilateral diplomacy, and domestic silicon to offer emerging markets an alternative to the American AI stack.

The strategy unfolds across three interconnected layers. At the software level, Chinese laboratories and technology conglomerates have embraced open-weight releases. Architectures such as Alibaba’s Qwen and DeepSeek provide developers across the Global South with high-capability base models. Unlike proprietary frontier systems hosted behind American cloud interfaces, open weights allow partner governments to deploy models on sovereign servers, fine-tune on local datasets, and avoid recurring foreign currency subscriptions.

Beijing reinforces this software footprint through institutional diplomacy. Chinese diplomats actively promote the Global AI Governance Initiative across multilateral forums, including the United Nations, BRICS, and the Shanghai Cooperation Organisation. State officials frame Western safety frameworks and export regimes as exclusionary measures designed to entrench technological dependency. In their place, Beijing advocates for capacity-building programmes and sovereign regulatory autonomy, aligning itself with developing nations that feel marginalised by transatlantic governance initiatives.

The final layer targets physical compute. Through the Digital Silk Road, Chinese vendors pair data centre construction with domestic hardware exports. Companies such as Huawei market Ascend compute clusters to foreign governments seeking turnkey artificial intelligence facilities. Although export restrictions on advanced semiconductor manufacturing equipment continue to constrain the raw efficiency of Chinese chips, aggregated domestic clusters provide adequate throughput for local fine-tuning and inference workloads.

By bundling foundation models, diplomatic pacts, and physical infrastructure, Beijing offers an integrated, state-backed pipeline. Partner states gain accessible computing capacity alongside insulation from unilateral American sanctions. However, adopting this architecture embeds recipient nations within Chinese telecommunications standards, proprietary compute frameworks, and supply chains.

Can partner states build genuine technological autonomy on Beijing's subsidised foundation, or will they simply exchange Western platform lock-in for Chinese infrastructural dependency?