Kimi K3 Shocks Silicon Valley, and the U.S. Begins Reviewing Why It Couldn’t Keep AI Genius Yang Zhilin

Moonshot AI founder Yang Zhilin sent shockwaves through Silicon Valley with Kimi K3, an open-weight model with 2.8 trillion parameters, turning this 34-year-old Tsinghua and CMU alumnus into a global talking point overnight. By choosing to return to China to start a business rather than stay in a U.S. AI lab, he also prompted Silicon Valley figures like Vinod Khosla to reconsider the Trump administration’s immigration policies—are they actually pushing top talent toward the rival camp.
(Background: Moonshot AI is about to launch Kimi K3! With 2.8 trillion parameters and million-length context, performance nearly matches Anthropic Opus 4.8.)
(Additional context: Is Kimi K3 pressuring U.S. AI giants? Experts predict: Anthropic will urgently push Opus 5, and GPT-6 may be released early.)

At 34 years old, Yang Zhilin stunned Silicon Valley with an open-weight model. His company, Moonshot AI, released Kimi K3, its first open-weight model with 2.8 trillion parameters, with coding and agent task capabilities nearing the top models from Anthropic and OpenAI. Vercel CEO Guillermo Rauch described on X that this marks “the first time an open model has surpassed all closed models in such a comprehensive web engineering benchmark test.”

From Tsinghua to CMU—researchers targeted by big tech all the way through

Born in 1992 in Shantou, Guangdong, Yang Zhilin first studied at Tsinghua University, then went to Carnegie Mellon University (CMU) for a PhD. His advisors were AI scholars Ruslan Salakhutdinov and William Cohen, and he earned his degree in just 4 years. On X, Salakhutdinov said Friday: “He’s absolutely a genius.” He also noted that after Yang Zhilin graduated, major tech companies actively recruited him, but his mind had already been set on starting a business.

I’ve been asked several times whether Zhilin Yang, the founder of @Kimi_Moonshot was my PhD student. The answer is yes and he is absolutely brilliant.

But I’ve been incredibly fortunate to work with so many outstanding PhD students over the years. So I thought I’d brag a little…

— Russ Salakhutdinov (@rsalakhu) July 17, 2026

During his time at CMU, Yang Zhilin interned at Google Brain and Meta, and co-authored multiple papers. Transformer-XL and XLNet are standout works. Back then, XLNet outperformed Google’s BERT on 20 benchmark tasks. After returning to China, he worked on Huawei’s PanGu model, the multi-modal model “悟道” from the Beijing Academy of Artificial Intelligence, and co-founded Recurrent AI, a startup that uses AI to analyze sales conversations.

In early 2023, Yang Zhilin and his team founded Moonshot AI. The company name comes from Pink Floyd’s album The Dark Side of the Moon, and he himself is a fan of the band. Moonshot AI first made a name with Kimi K2, an open-weight model with 1 trillion parameters known for its ability to handle long-form text. The company later expanded into programming, research, and autonomous AI agents, receiving funding from major investors including Alibaba and Tencent. This year, Moonshot AI raised about $2 billion in fundraising at an estimated valuation of about $20 billion, and is also preparing to list in Hong Kong.

One of Yang Zhilin’s principles for building AI is to win primarily through scale. In an interview with reporters, he said: “If you can solve it through scale, don’t solve it with a new algorithm. The value of a new algorithm is letting scaling be done better.” He also judged in the same interview: “In the end, the scale of AGI companies will far exceed today’s giants—double, even triple. It may not be OpenAI, but companies like that will definitely emerge.”

Can’t keep talent in the U.S.?

After Kimi K3 was released, some in the U.S. tech community turned to questioning why Yang Zhilin didn’t stay and work in U.S. AI labs. Venture capitalist Vinod Khosla pointed to the Trump administration’s immigration policies. He said: “The bigger issue is that we’re using immigration policies that aren’t friendly to top talent, pushing excellent talent out of other countries.”

In recent times, the Trump administration has continued to tighten immigration restrictions, including student visas. Last year, the government added a $100k fee for employers to sponsor certain H-1B applicants for new hires, but it was blocked by a federal judge and is still tied up in ongoing litigation. In May this year, a memorandum from the U.S. Citizenship and Immigration Services showed that those who could originally apply for a green card from within the country may in the future have to leave first; this week, the government also set limits on the initial length of stay for student visas.

However, his mentor Salakhutdinov offered a different interpretation. He said the U.S. immigration process is indeed daunting and full of uncertainty, but Yang Zhilin had already decided from the start to return to China to start a business: “I remember him telling me that if he didn’t try starting his own company, he would regret it for the rest of his life. I respect that decision, and he’s right.”