China AI Startups in 2025

China's AI startup scene has exploded in 2025, with dozens of well-funded companies racing to develop large language models (LLMs) and commercialize generative AI applications. Leading the pack are Zhipu AI (backed by Alibaba and Tsinghua University), MiniMax (valued at over $2.5 billion), Moonshot AI (creator of the popular Kimi assistant), and Baichuan AI. These startups compete not only with each other but also with tech giants like Baidu (Ernie Bot), Alibaba (Tongyi Qianwen), and Tencent (Hunyuan). The market has seen intense price wars, with AI API costs dropping by over 90 percent in 2024 alone. This report profiles China's leading AI startups, their technologies, funding, and competitive positioning in the global AI race.

TL;DR

Zhipu AI raised $400M+ at $3B+ valuation. MiniMax reached $2.5B valuation. Moonshot AI's Kimi surpassed 30M monthly active users. Baichuan AI launched open-source LLMs rivaling Llama. China had 200+ LLM developers. AI API prices dropped 90 percent in 2024.

Key Insights

Zhipu AI Funding Round

$400M+

Zhipu AI, spun out of Tsinghua University's KEG lab, raised over $400 million in its latest round at a $3 billion+ valuation, with backing from Alibaba, Tencent, Meituan, and several state-backed funds.

MiniMax Valuation

$2.5B

MiniMax, founded by former SenseTime executives, achieved a $2.5 billion valuation after raising over $600 million total, offering LLMs, AI companions, and content generation tools to consumer and enterprise clients.

Moonshot Kimi Users

30M MAU

Moonshot AI's Kimi chatbot surpassed 30 million monthly active users, becoming one of China's most popular AI assistants, competing directly with Baidu's Ernie Bot and Kimi's strength in long-context document processing.

Baichuan Open-Source LLMs

Open Source

Baichuan AI released multiple open-source LLMs that benchmark competitively against Meta's Llama models on Chinese language tasks, attracting over 5 million developer downloads on platforms like ModelScope and Hugging Face.

AI Price War Impact

-90% Cost

Fierce competition among Chinese AI companies drove API pricing down by over 90 percent in 2024, making AI capabilities accessible to startups and enterprises at unprecedented low costs, accelerating AI adoption across industries.

Total LLM Developers

200+

Over 200 organizations in China have developed or are developing large language models, including tech giants, AI startups, universities, and research institutes, creating the world's most competitive LLM ecosystem.

Side-by-Side Comparison

StartupValuationKey ProductUsers/CustomersDifferentiator
Zhipu AI$3B+GLM-4, ChatGLMEnterprise + ConsumerTsinghua research, enterprise focus
MiniMax$2.5Babab LLM, AI companions10M+ usersCharacter AI, social applications
Moonshot AI$2B+Kimi chatbot30M MAULong-context processing (2M tokens)
Baichuan AI$1.5B+Baichuan LLMs5M+ developersOpen-source strategy, multilingual
01.AI$1B+Yi LLM seriesEnterprise + APIKai-Fu Lee led, international focus
StepFun$1B+Step-1/2 LLMsGrowingMulti-modal AI, image generation

Frequently Asked Questions

How do Chinese AI startups compare to OpenAI and Google?

Chinese AI startups generally trail OpenAI and Google in frontier model capabilities but excel in Chinese language processing, price-performance for enterprise applications, and deployment speed within China's market. Zhipu AI's GLM-4 and Baichuan's models perform competitively on Chinese benchmarks, while Moonshot AI's Kimi leads in long-context processing. However, gaps remain in reasoning, coding, and multi-modal capabilities.

Why are AI API prices so low in China?

Intense competition among 200+ LLM developers, combined with abundant computing resources from government-subsidized data centers and aggressive market-share strategies, has driven prices to rock-bottom levels. Some companies offer free tiers or charge as little as 0.001 RMB per 1,000 tokens, compared to $0.01-0.03 for comparable international APIs.

What regulatory challenges do Chinese AI startups face?

Chinese AI companies must comply with regulations requiring algorithm registration, content safety reviews, deep synthesis labeling, and generative AI service licenses. These requirements add compliance costs but have also created barriers to entry that benefit established players. Companies must also ensure their models align with government content guidelines.

Which Chinese AI startup is most likely to succeed internationally?

01.AI, founded by Kai-Fu Lee with explicit international ambitions, has the strongest global positioning. Its Yi model series performs well on English benchmarks and the company has built partnerships with international cloud platforms. MiniMax also has international character AI products. However, US export controls on AI chips and geopolitical tensions present significant challenges for all Chinese AI companies seeking global expansion.