Chinese AI Models Are Crushing It on Price—And the World Is Buying

Open-weight AI—models whose internal code is public for anyone to download and modify—is turning into the new battleground in the US-China tech race. Chinese developers have moved aggressively in this space, releasing systems that match the performance of leading US models while undercutting them on price. In late 2024, DeepSeek released an open-weight model that matched top US systems at a fraction of the cost, surprising Western experts and proving that Chinese firms can innovate even under strict US export controls on advanced chips [222267]. Those controls forced Chinese developers to write more efficient code, which became a competitive advantage [222267].

The shift is drawing new users, especially businesses that want to run AI on their own servers rather than relying on cloud-based services from US tech giants [225390]. For many everyday tasks, Chinese models now perform nearly as well as US ones—but at significantly lower prices, a combination that is proving hard to ignore for cost-conscious firms [225390].

The trend is now reaching the highest levels of Western tech. San Francisco-based legal tech company Harvey, backed by OpenAI, Sequoia Capital, and Andreessen Horowitz, announced it built its first in-house AI model using Kimi K3, an open-weight base model developed by Chinese lab Moonshot AI [224857]. The move was driven by soaring development costs for building models from scratch; by using Kimi K3 as a foundation, Harvey aims to cut expenses while tailoring the model for legal work [224857]. Harvey emphasized that the model remains fully owned and controlled by the company, with all client data processed in the US [224857].

African businesses are taking a similar pragmatic approach. Nigeria-based insurtech company Curacel, which uses AI to process insurance claims, added Chinese models to its toolkit alongside American ones [225870]. This reflects a broader trend across the continent: firms are focusing on what works best for their operations, not on political loyalties, mixing technologies from both superpowers to cut costs and improve services [225870].

The geopolitical stakes are high. The competition is not just about technology—it is about setting global standards. Countries that adopt a particular open-weight model will build their infrastructure around it, giving the originating nation long-term geopolitical influence [222267]. China is courting the Global South with free, capable models, while the US is leaning on its allies to adopt its own open-source ecosystems [222267].

At a recent APEC digital technology and AI forum in Chengdu, ministers from across the Asia-Pacific Economic Cooperation issued a joint statement supporting open-source AI models [225879]. The statement points toward an AI order that defies the assumption of a single winner—an assumption that still drives Washington’s alarm over every Chinese AI breakthrough [225879]. For many regional economies, open-source models offer a cheaper, more flexible alternative to proprietary systems, reducing the risk of being locked out of advanced technology by political tensions [225879].

The next few years will be decisive. If the US restricts open-weight exports too heavily, it may push neutral nations toward Chinese systems. If it opens up completely, it risks losing its edge in safety and control. For now, the battle lines are drawn, and the winner will shape how the world uses AI for decades [222267].