Within days of each other, the world’s two artificial intelligence (AI) superpowers unveiled duelling blueprints for the future. The United States released its sweeping
AI Action Plan, calling for deregulation, semiconductor expansion and “full-stack” AI export packages to allies. Days later, China put forth
its proposal at the World Artificial Intelligence Conference in Shanghai: a global AI governance body open to the Global South, a push for open-source collaboration and a subtle rebuke of AI becoming “an exclusive game” dominated by a few nations.
The visions mirror one another in ambition at first glance, but beneath the surface lies a strategic divergence: the US aims to dominate through invention and control while China seeks to influence through adoption and distribution. In the unfolding race for AI leadership, the former focuses on capabilities and the latter on infrastructure.
The US playbook is familiar. For decades, American hegemony ran on a compounding engine of public research and development feeding private commercialisation, which in turn built platform monopolies such as Unix and iOS that scaled globally. Like steel and railways once powered industrial empires, these digital scaffolds became the data arteries of the modern world, quietly carrying the lifeblood of global infrastructure. American tools became global defaults, not by decree but by design.
AI, particularly large language models, is reshaping the formula. Today’s leading models, such as GPT-4, Claude 3 and Gemini 1.5, remain gated behind proprietary interfaces.
Even Llama, Meta’s so-called open model, carries usage restrictions. These are technical marvels, but in many parts of the world they remain inaccessible, unaffordable or inflexible.
Meanwhile, China is taking a different path. Last month, Chinese labs released two of the world’s most effective open-weight models in Moonshot’s
Kimi K2 and Alibaba’s
Qwen3, which both rival their Western peers across several benchmarks. Critically, they are optimised for use cases that matter most to governments and enterprises, such as document processing and financial summaries.
Chinese companies aren’t stopping at model release, either. Some of the biggest tech firms, such as Tencent, ByteDance and Alibaba, are racing to build agent platforms that make these models useful. At the Shanghai AI conference, Tencent unveiled a suite of
enterprise-ready agents that automate marketing, evaluate campaigns and write code.
