China's Kimi AI Model Leads 90% of Global Open-Source Enterprise Spending, New Study Finds

The Quick Take

• Chinese AI models led by Moonshot AI's Kimi accounted for nearly 90% of global enterprise spending on open-source AI systems
• China's AI sector is demonstrating commercial traction that rivals closed proprietary models from US firms
• The enterprise adoption data signals China's AI ecosystem is winning on cost and performance in price-sensitive markets

The finding that Chinese AI models—led by Moonshot AI's Kimi family—capture roughly 90% of global open-source enterprise AI spending is a landmark data point in the ongoing US-China AI competition narrative. This is not a theoretical research benchmark but a commercial deployment metric: enterprises are choosing Chinese open-source models for real workloads. The economic logic is straightforward—Chinese models offer competitive performance at dramatically lower inference costs, particularly for Asian-language tasks, making them the rational choice for cost-conscious enterprise IT buyers in markets without regulatory mandates to use US models.

For global AI and semiconductor markets, this data point has several second-order implications. It validates that the open-source model distribution strategy (releasing weights publicly) is a viable commercial moat, not just an academic exercise—Moonshot AI and others are building enterprise relationships and integration stacks on top of open-source releases that create switching costs. For US AI firms, it narrows the addressable enterprise market outside North America and Europe where regulatory barriers to Chinese models don't exist. NVIDIA's China revenue, already under export control pressure, faces a structural market where Chinese AI firms are increasingly self-sufficient on model development.

Forward signals include Kimi's enterprise customer count disclosures, any regulatory responses from USTR or Commerce Department on Chinese AI model access in US enterprise markets, and the upcoming DeepSeek and Baidu AI model benchmark releases which will show whether Chinese open-source capabilities continue to close the gap with US closed models. The macro variable: US AI export controls expansion to cover model weights (not just chips) would fundamentally change this dynamic by restricting Chinese model distribution to international markets.

Synthesized from 1 source.