Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
Following the breakout global success of Chinese AI lab DeepSeek’s R1 model early last year, Yang Zhilin faced what many believed would be an existential test for his start-up Moonshot AI.
The bespectacled 34-year-old founder had failed to capitalise on the early popularity of Kimi, his company’s chatbot, which was plagued by outages and rapidly overtaken by larger rivals. At the same time, he was contending with a lawsuit from an investor in a previous venture and reports that he and a co-founder sold shares in Moonshot during an early fundraising round. The company denied the reports, which brought unwelcome scrutiny.
Many in China’s AI industry questioned whether Moonshot could survive the brutal and expensive race to build frontier large language models. In 2024, the start-up had made just $2.4mn in revenue, while its costs from model training and hiring researchers were exploding, according to one investor briefed on the numbers. Rivals including Baichuan and 01.ai had already abandoned the contest to focus on applications, and some wondered whether Yang would follow suit.
But after DeepSeek’s R1 release, Yang concluded that Moonshot had moved into commercialisation too early and decided to focus purely on model training, ditching efforts to market the Kimi chatbot. In another pivotal decision, he committed to open-sourcing Moonshot’s leading AI models, allowing users to download the systems and run them on their own hardware.
“If the top leader can’t make decisive calls, that won’t work,” Yang said in a 2024 interview with the Chinese press.
It was the start of a journey that culminated last week in the release of the company’s K3 model. The new arrival closes the gap with leading US models and has prompted panic about the rising capabilities of China’s AI, in an echo of last year’s “DeepSeek moment”.
Investors are now clamouring to get access to Moonshot. The company, which is gearing up for an initial public offering, is currently doing two rounds of consecutive fundraising, the second of which would value it at around $50bn, according to investors with knowledge of the matter.
A boom in AI coding assistants and autonomous agents has created surging demand from developers, turning Moonshot into one of China’s fastest-growing model providers. Annual recurring revenue reached about $300mn in June, they added.
Wang Tiezhen, an independent consultant and engineer, said Yang’s status as a “leading AI researcher” has enabled him to “build a team of talented engineers all working towards the same goal”. Before founding Moonshot, he was known for research extending models’ context windows, allowing chatbots to retain more information over long conversations.
China’s AI race has been a battle for talent, with deep-pocketed tech groups aggressively poaching from smaller start-ups. Throughout this, Yang has kept one of China’s strongest research teams together. Many of Moonshot’s founding researchers studied alongside him either at Tsinghua University, where he graduated top of his computer science class, or at Carnegie Mellon University, where he completed his PhD. During his CMU years, Yang had stints at Google Brain and Meta.
That cohesion has distinguished Moonshot from Chinese AI groups including Alibaba and Baidu, which have seen prominent researchers depart amid disagreements over balancing frontier research with commercial demands.
Moonshot’s decision to pivot to open source reinforced that advantage. Researchers are attracted by the opportunity to publish papers and receive individual recognition. “Researchers want to be remembered for breakthroughs,” Wang said. “Open source gives them that.”
Employees describe an intense culture of long hours in Moonshot’s utilitarian headquarters in Beijing’s Haidian district. But Yang has also infused the company with his own particular obsession with rock music.
The company’s name comes from Pink Floyd, while meeting rooms are named after rock bands. A white piano stands prominently in the office.
Former classmates recall Yang, born in the southern province of Guangdong, as unusually gifted. He began coding in high school and won first prize in the National Olympiad in Informatics — an achievement that earned him direct admission to university. He also balanced elite academic performance with music. A drummer in a campus band called Splay, he organised university music competitions alongside future Moonshot co-founder Zhou Xinyu.
“The fact that Yang could get top grades while doing all that made him famous at Tsinghua,” one former classmate said. “People called him Yang Shen” — which translates to “Yang the genius”. “Friends would describe him as romantic and idealistic,” they added.
That mythology has helped distinguish Moonshot from China’s otherwise austere AI industry. But Kevin Xu, an investor and China AI specialist, argues Yang’s interest in music and the humanities is more than branding.
“Culture that embraces not only hardcore technology but also music, art and the humanities gives Moonshot an edge,” he said. “So much of frontier AI development has to do with research taste and combining the humanities with technology in a way that we have not seen before.”
[email protected], [email protected]
Credit: Source link
