Executive Summary
The launch of the Kimi K3 ahead of the Shanghai World AI Conference represents an important development in the China-US AI race, highlighting the rapid pace at which China is narrowing the technological gap. This brief argues that China’s AI development is not merely a story of technological catch-up: China’s open-weight ecosystem is positioning AI as the new frontier of China’s “making it cheaper and better” export-led growth model. This raises the possibility that AI could be the next sector in which China gains substantive market share, as has already happened in the electric vehicle and solar panel industries.
Key points
- On 16 July, Chinese AI start-up Moonshot unveiled the Kimi K3 model, the world’s first open-weight model in the three-trillion-parameter class.
- This launch has attracted attention in the US AI industry as it competes with leading closed models such as Anthropic’s Claude and OpenAI’s GPT at substantially lower cost.
- AI development is a top governmental priority in China. It is considered one of the future-oriented industries driving China’s “high-quality development” and one of the “new ‘new three’” export drivers, i.e. the main sectors driving its export competitiveness.
- In 2017, Beijing released the New Generation Artificial Intelligence Development Plan, setting the goal of making “China the world’s primary AI innovation center by 2030” and outlining a strategy combining aggressive adoption with open-source support.
Analysis
Leading Chinese AI labs rely on open-weight models, letting developers download, modify, and build on their underlying weights. This choice has helped many Chinese AI start-ups to survive despite the intense domestic competition: lower cost and accessibility of open-weight models facilitate foreign market expansion, while providers monetize through the infrastructure required to access and run the model, including API access and cloud services. This dynamic is already visible in Japan, across African markets, and in the United States, with one industry estimate suggesting that roughly 80% of US startups use Chinese models for their cost competitiveness and accessibility. Kimi K3 offers a significantly better price-to-performance ratio than leading US models. At launch it ranked third on the Intelligence Index, behind only leading models from Claude and GPT. Yet its cost per task is roughly 65% lower than Claude Opus 5 (max)’s and 53% lower than GPT-5.6 Sol (max).
The emergence of an increasingly capable and cost-competitive open-weight Chinese AI ecosystem resembles the dynamics that have driven China’s global expansion in solar panels and EVs, with China leveraging cost competitiveness and scale to accelerate global adoption. By making capable models available at lower cost, China’s open-weight strategy allows Chinese firms to scale internationally. Wider adoption generates direct revenue while creating a feedback loop in which a growing user base supports iteration and improvement. This could lead to the emergence of an ecosystem shaped by Chinese technology and standards, potentially raising switching costs and making alternatives progressively more difficult to adopt. A pressing question for US firms thus emerges: how to maintain market competitiveness as lower-cost and comparably capable alternatives gain market share?
Policy Implications
US policymakers should exercise caution in banning or restricting Chinese models. Such measures could reduce access to low-cost alternatives, damaging domestic developers while doing little to prevent Chinese models from spreading to other countries.
Policies should be directed to regulate the commercial access to Chinese models, by requiring developers to use US-based service providers that meet data-protection and security requirements.
US policymakers should design a strategy to strengthen domestic open-weight capacity and offer a cost-competitive alternative to Chinese models. This could include government procurement contracts and support for the security and software tools required to operate the models.
References
Artificial Analysis. (n.d.). Artificial Analysis Intelligence Index. Artificial Analysis. Retrieved July 31, 2026, from https://artificialanalysis.ai/evaluations/artificial-analysis-intelligence-index?eval-cost=cost-per-task
CGTN. (2026, July 27). China’s New “New Three” Exports Power Next Wave of Global Trade. CGTNOfficial. https://news.cgtn.com/news/2026-07-27/China-s-new-new-three-exports-power-next-wave-of-global-trade-1P7zEBqbtUk/p.html
Chen, T. (2026, April 21). China’s open-source bet. MIT Technology Review. https://www.technologyreview.com/2026/04/21/1135658/china-open-source-models-ai-artificial-intelligence/
Fan, Z. (2024, April 17). China’s emerging industrial vision: the significance and impact of ‘New Quality Productive Forces’. Cambridge Industrial Innovation Policy. https://www.ciip.group.cam.ac.uk/reports-and-articles/chinas-emerging-industrial-vision/
Lin, D. (2026, June 4). Forged in a Knife Fight: China’s Brutal Domestic AI Competition. War on the Rocks. https://warontherocks.com/cogs-of-war/forged-in-a-knife-fight-chinas-brutal-domestic-ai-competition/
Luong, N. (2026). Two Loops: How China’s Open AI Strategy Reinforces Its Industrial Dominance. U.S.-China Economic and Security Review Commission. https://www.uscc.gov/sites/default/files/2026-03/Two_Loops–How_Chinas_Open_AI_Strategy_Reinforces_Its_Industrial_Dominance.pdf
Pascual, M. G. (2026, July 26). Kimi K3, the Chinese AI That Has Put Silicon Valley on Alert. EL PAÍS English. https://english.elpais.com/technology/2026-07-26/kimi-k3-the-chinese-ai-that-has-put-silicon-valley-on-alert.html
Velasquez, F. (2026, July 17). China’s moonshot AI claims Kimi K3 can rival OpenAI and Anthropic. BBC. https://www.bbc.com/news/articles/cy9w4q8pgp0o
Webster, G., Creemers, R., Kania, E., & Triolo, P. (2017). Full translation: China’s “new generation artificial intelligence development plan” (2017). Stanford University. DigiChina. https://digichina.stanford.edu/work/full-translation-chinas-new-generation-artificial-intelligence-development-plan-2017/