Alibaba Cloud Cuts AI Inference Prices 70% in War for Developers
AI Price War Intensifies in China
Alibaba Cloud has slashed pricing for its AI model inference APIs by up to 70%, pricing its Qwen 2.5 model at 0.0008 yuan per 1,000 input tokens — less than $0.15 per million tokens. The aggressive move is designed to capture developer market share as Chinese tech giants compete to become the default AI platform for enterprise and startup customers.
Alibaba Cloud Intelligence CEO Eddie Wu announced the cuts at the Apsara Conference in Hangzhou. "AI inference is the new cloud workload. The platform that wins developers today wins the enterprise market tomorrow," Wu said.
Competitive Responses
The price cut triggered immediate responses. Baidu reduced ERNIE Speed pricing by 50% within 24 hours. ByteDance's Volcano Engine matched Alibaba's rates for its Doubao model. Tencent Cloud announced free tiers for its Hunyuan model for startups processing under 10 million tokens per month.
The pricing dynamics mirror US cloud AI competition between OpenAI, Google, and Anthropic, but at even lower price points. Chinese AI inference pricing is now roughly one-fifth of comparable US offerings, driven by lower labor costs, government subsidies, and intense domestic competition.
Developer Ecosystem Strategy
Alibaba Cloud's strategy extends beyond pricing. The company simultaneously launched ModelScope 2.0, an open model marketplace hosting over 4,000 pre-trained models including Qwen variants, image generators, and specialized vertical models. Developers can fine-tune and deploy models through Alibaba Cloud infrastructure with integrated billing.
The platform now serves over 3 million registered developers, up from 1.8 million a year ago. Enterprise API calls grew 300% in the past quarter, with financial services and e-commerce as the largest verticals.
Margin Pressure Concerns
Analysts have raised concerns about the sustainability of aggressive AI pricing. IDC analyst Tracy Woo said the current pricing "is below cost for all major providers. The question is who can sustain losses longest to build an ecosystem moat." She estimated that Chinese cloud providers collectively lost 2 to 3 billion yuan on AI inference services in 2025.
Alibaba reported total cloud revenue of 29.6 billion yuan in Q3 2025, up 7%. The cloud division's adjusted EBITDA margin was 5.4%, down from 7.1% in the prior quarter — a decline management attributed to AI infrastructure investment and promotional pricing.
