Physix Frontier · News Briefing Card (enbrief · Nov 9, 2026)

Chinese open-source models take 89% of enterprise open-source AI spend; Kimi leads at 61%

KEY FACTS

Menlo Ventures released a report this week tallying the global distribution of enterprise spending on open-source models.
Chinese AI models together account for 89% of global enterprise spending on open-source models.
Kimi leads with a 61% share, DeepSeek holds 15%, and the Zhipu GLM series holds 11%.
Meta's Llama series accounts for only 7% of open-source enterprise spending.
Open-source models overall make up just 4% of total enterprise AI spending.

KEY DATA

89%Chinese models' share of open-source enterprise spend

61%Kimi share

15%DeepSeek share

11%GLM series share

PHYSIX OBSERVATION

The open-source track is almost entirely swept by Chinese models, yet open source accounts for only 4% of total enterprise AI spending, which shows the real money still flows to closed source. Kimi's lead rests on developer word of mouth rather than big-vendor channels, and that is a signal for domestic vendors: open source can win share but is hard to monetize directly, and the next step is to see who can convert open-source users into paying enterprise customers.

Source: enbrief original report ↗