Liang Hao judges AI hardware by strict, practical standards—whether AI is truly required for the product to exist and whether it captures real-world context, like emotion or environment, that a phone can't—and he pays close attention to whether a team's whole organization, not just its product, keeps adapting alongside AI. He's drawn to founders who can explain user retention and monetization from first principles, favors ideas that turn old, low-trust, low-frequency human needs into high-frequency AI-driven relationships built on real context rather than one-off outputs, and applies a similar structural lens to cross-border e-commerce, where he tracks sellers moving from basic listing toward branding and more careful, software-supported operations.
About their Fund
Yunqi Partners is a venture capital fund that works closely with universities and researchers—running a joint fund with Shanghai Jiao Tong University to invest before companies are even formed—and organizes its focus into three layers: core AI models and applications, expansion-stage infrastructure like compute and semiconductors, and future bets like commercial space, nuclear energy, and quantum computing, paired with unusually deep technical diligence, as one founder recalled that Yunqi's managing partner was the first investor to actually review their source code. The firm also stands out for staying invested over many years, joining six rounds in MiniMax and backing Manycore before its first fund had even closed after spotting its GPU cloud rendering potential in 2013, while staying hands-on enough to help shape company strategy and products, and building ties with MIT and its own programs to connect with new generations of technical founders.
















.jpeg)







.jpeg)











































































.png)
























































































































.jpeg)









.jpeg)




















































































