IMPLICATOR
Naive AI has raised $400 million at a $1.42 billion valuation. The figures come from one unnamed person with direct knowledge cited in The Information’s exclusive, and Naive AI has not confirmed the financing. It plans to release its first open-weight language model as early as this month.
What Changed
- Naive AI has raised $400 million across three rounds at a $1.42 billion valuation, according to one unnamed person cited by The Information. The company has not confirmed the figures.
- Dai Jifeng, an associate professor at Tsinghua University, founded the Beijing startup in February 2026. April reports in China put it at roughly $800 million after about $300 million raised.
- The first model, due as early as this month, is built on an existing Chinese open-weight system and will be released as open weights.
- MiroMind, where Dai was technical adviser until January, said in an April 23 notice that its technology and intellectual property remain its own and reserved the right to take legal action.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Three rounds in seven months
Dai Jifeng, an associate professor in Tsinghua University’s Department of Electronic Engineering, founded Naive AI in February 2026. The company has fewer than 100 employees.
The first funding round brought in $100 million, followed by $180 million and then $120 million in the round that closed recently. Investors named in the report include Tencent, IDG Capital, MPCi and HSG, the firm formerly known as Sequoia Capital China.
The valuation has risen since reports in China on April 22, 2026 put Naive AI at roughly $800 million after about $300 million had been raised. Lin Junyang, the former lead of Alibaba’s Qwen models, closed a first round in June 2026 at a roughly $2 billion post-money valuation before releasing a product. HSG and Gaorong invested $100 million each, while Tencent put in $20 million.
FREE WEEKDAY MORNING BRIEFING
Follow China's AI labs as the money moves.
The Implicator Morning Briefing filters the AI news cycle to the stories worth your attention and explains their consequences. From San Francisco, every weekday at 4:45 a.m. Pacific, 7:45 a.m. Eastern.
About five minutes. No hype. No spam.
A model built on an existing base
Naive AI is building its model on an existing Chinese open-weight system. Its researchers are changing the pretrained model’s structure, then refining it with reinforcement learning and other processes. The planned release would let users download the weights without charge and customize them.
The company is also researching recursive self-improvement, in which a model attempts to upgrade its own capabilities.
An unresolved ownership dispute
Dai served as technical adviser to MiroMind, the AI lab founded by Shanda founder Chen Tianqiao, until a joint Shanda and MiroMind statement on January 18, 2026 said he was stepping down. Dai told the Washington Post on April 22 that MiroMind tried to force him to relocate overseas and that this triggered his departure.
MiroMind denied that account in an internal notice dated April 23, producing an O-1 visa approval in Dai’s name. It said the two sides never reached an agreement and that its core technology and intellectual property remained entirely its own, with no third party licensed.
Know someone who'd find this useful? ✉️ Email it to a friend in one click, or they can subscribe free here.
The notice said Dai initially offered 15% of his new company for an intellectual-property license and permission to take key staff. It said his investors reduced the offer to 5% on March 2 while seeking a free, perpetual, irrevocable worldwide license. MiroMind also said it had introduced Dai to IDG and Sequoia, reported the matter to Chinese authorities and reserved the right to pursue criminal and civil action.
Dai declined to comment to Bloomberg in April 2026. No lawsuit filing, settlement or resolution of the dispute has been reported publicly since the April notice.
Valuations ahead of revenue
Rhodium Group estimates published this month put the combined annual recurring revenue of Chinese AI models at about 10% of the annual recurring revenue generated by OpenAI and Anthropic. The figures use data available through summer 2026.
Rhodium estimated valuation-to-annual-recurring-revenue ratios of about 50 times for Moonshot and 163 times for DeepSeek, compared with 34 times for OpenAI and 21 times for Anthropic. “The financing gap means it will be far more difficult for Chinese frontier AI labs to scale sustainably,” Rhodium partner Logan Wright said. “They will be heavily dependent upon a favorable climate in the equity market. Historically that’s not an easy bet in China.”
Frequently Asked Questions
What is Naive AI?
Naive AI is a Beijing-based large language model startup founded in February 2026 by Dai Jifeng, an associate professor in Tsinghua University's Department of Electronic Engineering. It has fewer than 100 employees and has not yet released a model.
How much has Naive AI raised and who invested?
A person with direct knowledge told The Information the company raised $100 million, then $180 million, then $120 million, for $400 million in total at a $1.42 billion valuation. Investors named in the report include Tencent, IDG Capital, MPCi and HSG, formerly Sequoia Capital China. Naive AI has not confirmed the financing.
What kind of model is Naive AI building?
The company is building on an existing Chinese open-weight model instead of pretraining from scratch. Its researchers change the pretrained model's structure and refine it with reinforcement learning and other processes. The planned release would let users download and customize the weights free of charge.
What is the dispute between Dai Jifeng and MiroMind?
Dai told the Washington Post in April that MiroMind tried to force him to relocate overseas. MiroMind denied that in an April 23 internal notice, said its core technology and intellectual property remain its own with no third party licensed, and reserved the right to pursue criminal and civil action. No lawsuit or settlement has been reported publicly since.
How do Chinese AI lab valuations compare with revenue?
Rhodium Group estimated this month that Chinese AI models combined generate about 10% of the annual recurring revenue of OpenAI and Anthropic. It put valuation-to-ARR ratios at about 50 times for Moonshot and 163 times for DeepSeek, against 34 times for OpenAI and 21 times for Anthropic.
AI-generated summary, reviewed by an editor. More on our AI guidelines.
Marcus Schuler
Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai
