China's UniPat hits $1B valuation on just $30M in orders
What's the deal? UniPatDealroom has a profile for this one. Try Dealroom →, a Beijing-based AI data and model evaluation startup, has closed a late-stage funding round at a $1 billion pre-money valuation, according to a person with knowledge of the matter. Founded in December, it produces specialised benchmarks used by developers and researchers to evaluate American and Chinese AI models.
Why now? Access to high-end data is increasingly critical for building advanced models capable of expert-level tasks. UniPat pays human experts from law, finance, healthcare, and science to produce specialised data that refines models during the reinforcement learning, or post-training, phase.
What's the endgame? Beyond expert data, UniPat generates synthetic data with its own algorithms and builds evaluation benchmarks in areas such as coding, biomedical research, and visual intelligence. Most major Chinese model makers — including Alibaba, ByteDance, MoonshotDealroom has a profile for this one. Try Dealroom →, and DeepSeek — have bought data or services from UniPat and its main rival, HumanlayaDealroom has a profile for this one. Try Dealroom →.
What could go wrong? The valuation looks steep against the numbers. As of the end of May, UniPat had $30 million worth of customer orders, making $1 billion an extreme multiple of forward revenue. By comparison, US peer Mercor crossed a $2 billion gross revenue run rate in June at a roughly $10 billion valuation.
The context: Investors say the premium is justified because China's domestic market for high-end data and post-training services is still relatively new. Humanlaya has just closed its own round at a post-money valuation of nearly $1 billion, according to two people with knowledge of it.
The signal: Silicon Valley venture firm Dimension CapitalDealroom has a profile for this one. Try Dealroom → flagged UniPat after a recent China research trip, telling limited partners it saw "a nascent cohort of Chinese upstarts building evaluation and verification infrastructure" rather than basic data annotation. The rise of local answers to Surge, Mercor, and Scale marks a new phase in China's AI push — one where domestic labs increasingly shape model quality with home-grown expert data rather than distilling American models.
Read more: The Information
Image credit: Generated with Gemini
