Qinghua-Stanford team raises $15M angel round to build an AI research "superfactory
What's the deal? Chinese startup Qingxi (清析智研) has raised nearly US$13.8M in an angel round, the company said. Investors include SMIC Capital, GL VenturesDealroom has a profile for this one. Try Dealroom →, Zhiyuan Robotics, Fengwu Capital, and Oriental Jiafu. The company is building what it calls an "AI Science SuperFactory" to automate materials research.
What's the endgame? Qingxi wants AI to run physical experiments rather than just assist scientists. Its system embeds AI into lab instruments to autonomously set parameters and collect data, then feeds that data back to improve its models. The aim is a self-running research factory spanning materials design, experimental validation, and industrialisation.
Who's behind it? Founder Dong Siming holds a bachelor's from Tsinghua University and a materials PhD from Stanford, and led an AI process-development project at a major US semiconductor firm. Co-founder Zhao Zhenlin brings Tesla battery and rare-earth magnet production experience; chief scientist Wang Huan works at the intersection of AI and complex scientific systems.
Why now? AI for Science has drawn heavy funding globally, with players such as Lila Sciences and CuspAI abroad. But models that top benchmarks often fail in the lab, where experimental data is costly, sparse, and noisy. Qingxi argues the real bottleneck is securing continuous feedback data from the physical world, not model size.
By the numbers: Qingxi says it has signed tens of millions of yuan in orders within six months of founding, with customers across consumer electronics, semiconductors, energy, and aerospace. It claims to have accumulated tens of millions of multimodal experimental data points. It is also developing rare-earth permanent magnet materials for robotics and electric vehicles, targeting mass production in 2027.
What could go wrong? These claims come largely from the company and remain unverified independently. The "superfactory" is a long-term goal, the capital raised is modest, and materials IP faces heavy, long-cycle engineering risk before any volume production.
The signal: The investor mix — spanning semiconductors, financial capital, and robotics — reads like a business map for a cross-industry bet on research automation. At this stage, it is best seen as an early wager on AI moving from digital tools toward running physical experiments, not a proven, scalable model.
Read more: sohu.com
Image credit: ESA_events