Yang Zhilin was known as “Yang the genius” by his classmates at Tsinghua University, and the label has followed him across two continents, through doctoral research at Carnegie Mellon, internships at Google Brain and Meta’s FAIR, and back to Beijing, where he founded Moonshot AI in early 2023. The Financial Times profile published on July 25 traces the arc of a founder who has, in the space of three years, built China’s fastest-ever unicorn, delivered a model that rattled US semiconductor stocks, and is now preparing a Hong Kong IPO that would value the company at approximately $30 billion.
From Tsinghua to Carnegie Mellon: The Making of a Technical Founder
Born in 1992 in Shantou, Guangdong Province, Yang graduated from Tsinghua’s computer science department before earning his PhD at Carnegie Mellon under the supervision of Ruslan Salakhutdinov, one of the world’s leading deep learning researchers. During his doctoral studies, Yang published groundbreaking papers in natural language processing — including Transformer-XL and XLNet — that established him as a serious researcher rather than simply a talented engineer. Salakhutdinov described him in a phone interview as “a phenomenal student who loved listening to Pink Floyd” and “a unique talent who possesses not only business acumen but also the technical skill to drill down to the very bottom of programming.”
The Pink Floyd reference is not incidental. Moonshot AI’s Chinese name, Yue Zhi An Mian (月之暗面), translates as “The Dark Side of the Moon”, the title of the iconic 1973 album that Yang has long counted among his favorites. The choice reflects a founder who thinks in long arcs and is comfortable with ambiguity, qualities that have served him well in navigating the volatile landscape of Chinese AI development.
After his PhD, Yang interned at Google Brain and Meta’s FAIR before returning to China to work on Huawei’s PanGu language model and BAAI’s Wu Dao project, two of China’s earliest large-scale language model efforts. A senior Apple executive who reported directly to Tim Cook approached Yang about a recruitment opportunity, but Yang declined, telling his advisor: “I feel like I’ll regret it for the rest of my life if I don’t even try now.” He raised $60 million and built a 40-person team in three months, and by 2024 had raised over a billion dollars, with Alibaba among his investors.
The Kimi K3 Breakthrough: Moonshot’s DeepSeek Moment
The release of Kimi K3 on July 16 is framed by the Financial Times as Moonshot’s “DeepSeek moment”, the point at which a company that had faced persistent skepticism about its revenue model and technical capabilities delivered something that forced the global AI industry to take it seriously. Kimi K3 is a 2.8 trillion-parameter mixture-of-experts model with a 1-million-token context window, scoring 57 points on Artificial Analysis’s composite intelligence index, placing it near the top tier alongside Claude Opus 4.8 and GPT-5.5, trailing only Anthropic’s Fable 5 (60 points) and OpenAI’s GPT-5.6 Sol (59 points). This marks the first time a Chinese open-weight model has demonstrated performance approaching that of cutting-edge US frontier models.
The model’s API pricing undercuts US frontier models by up to 70%, challenging the “intelligence monopoly” business model that OpenAI and Anthropic have built. Vercel CEO Guillermo Rauch noted that “this is the first time that an open model is ahead of all proprietary ones for this comprehensive web engineering benchmark.” The model’s release triggered a sell-off in US and Asian semiconductor stocks, with Nvidia briefly relinquishing its position as the world’s most valuable company by market capitalization, a “Kimi Shock” that echoed the DeepSeek shock of early 2025 and demonstrated that Chinese AI companies are now capable of moving global markets.
Revenue Trajectory and the Hong Kong IPO Race
Moonshot’s financial trajectory has been equally dramatic. By early 2026, the company had crossed $240 million in annualized revenue. The Kimi K2.5 release exceeded the entire 2025 revenue total in under 20 days, a rate of commercial adoption that few AI companies anywhere in the world have matched. As Nikkei Asia reported, the company is now planning an IPO in Hong Kong within six months, with Yang Zhilin estimated to hold a 51.83% stake worth approximately $16.3 billion at the current $30 billion valuation.
The IPO timing is significant. Moonshot is moving to go public while the Kimi K3 momentum is at its peak, before the next generation of models from OpenAI, Anthropic, or domestic Chinese competitors can erode its competitive position. As Moonshot AI’s valuation and IPO ambitions signal a maturing industry, Yang’s trajectory from Tsinghua classroom to global AI disruptor is complete, at least for this chapter. The question now is whether Moonshot can sustain the pace of innovation that Kimi K3 represents or whether, like many AI companies before it, it will find that a single breakthrough model is easier to build than a durable competitive position.