Alibaba to Charge Major Commercial Users of Next Open-Source Qwen AI Model Through Revenue-Sharing

Aug 7, 2026

Alibaba plans to pivot its open-source AI strategy by introducing a revenue-sharing model for major commercial users of its upcoming Qwen3.8-Max model. This shift mirrors a precedent set by rival Moonshot AI, which requires up to a 30% revenue share from partners earning over $20 million annually using its Kimi K3 model. The strategy signals a broader move by Chinese AI firms to monetize highly capable open-weight models and recoup massive computing costs amid intense domestic price wars and geopolitical competition with US rivals like OpenAI and Anthropic.

Alibaba Qwen revenue-sharing model

▪Alibaba has historically allowed customers to deploy most of its open-source models in their own data centers for free, while charging only for hosting on Alibaba Cloud.
▪Alibaba is expected to announce the revenue-sharing measures for its Qwen3.8-Max model as early as the week of August 10, 2026.
▪Alibaba plans to introduce a revenue-sharing mechanism for large commercial users in the licensing terms of its next-generation Qwen open-source AI model, Qwen3.8-Max.
▪Alibaba's planned revenue-sharing rate for the Qwen3.8-Max model remains unclear as internal discussions are still ongoing.

Moonshot Kimi K3 licensing precedent

▪Chinese AI startup Moonshot AI established a precedent by requiring companies generating over $20 million annually from its Kimi K3 model to sign separate commercial agreements.
▪Chinese IT services provider Chinasoft International disclosed in a July 2026 regulatory filing that it had reached a revenue-sharing agreement with Moonshot AI.
▪Moonshot AI requires a revenue share of up to 30% in its commercial licensing agreements for the Kimi K3 model.
▪DigitalOcean Holdings CEO Paddy Srinivasan confirmed his company has a commercial agreement with Moonshot AI but declined to disclose the specific terms.

Chinese AI commercialization strategy

▪A brutal price war has pressured Chinese AI firms, highlighted by DeepSeek making a 75% price discount permanent for its model access.
▪Chinese AI firms are shifting toward sustainable monetization to offset high infrastructure costs after years of aggressive price competition and heavy training investments.
▪Chinese AI companies are adopting a classic Silicon Valley freemium playbook by offering software at low cost to gain market share before charging heavy users.

Open-source versus closed-source AI models

▪San Francisco-based Thinking Machines Lab, founded in 2025 by former OpenAI CTO Mira Murati, released its first open-source model in July 2026.
▪Major US artificial intelligence models from OpenAI, Anthropic, and Google primarily remain closed-source, requiring users to pay for access via their data centers.
▪Alibaba's Qwen3.8-Max and Moonshot's Kimi K3 are open-weight models, allowing developers to download weights and run them locally, unlike closed-source US models.

US-China AI market competition

▪Chinese open-source AI models are approaching the capabilities of US models from OpenAI and Anthropic, intensifying international competition.
▪The cost of using Moonshot AI's Kimi K3 is approximately one-third the cost of Anthropic's Claude Fable model based on standard token pricing.
▪The White House has accused Moonshot AI of stealing technology from Anthropic, an allegation that Chinese officials have dismissed as baseless.