On July 23, after Dark Side of the Moon's K3 model caused a sensation, the company's founder Yang Zhilin left the United States and returned to China to start a business, sparking discussion. "Financial Times" Asia technology correspondent Zijing Wu (transliteration) wrote,More and more Chinese entrepreneurs believe that the development opportunities in China are greater than those in Silicon Valley, and the entrepreneurial environment is also better.

Yang Zhilin

The following is the main content of the article:

In 2019, Dark Side of the Moon founder Yang Zhilin completed his PhD from Carnegie Mellon University in just four years, about two years less than the usual time required. Apple wanted to recruit him. Google and Meta also let him intern at the company. Stanford and MIT also opened their postdoctoral doors to him. But he chose none and returned to China.

His mentor at Carnegie Mellon University, Russ Salakhutdinov, later recalled Yang Zhilin’s thoughts at the time:"If he doesn't try to start his own business at least once, he'll regret it for the rest of his life."

This decision, which was seen as a contrarian choice at the time, now seems like a foresight. Dark Side of the Moon last week released the large language model Kimi K3, which is already considered to be on par with flagship models from OpenAI and Anthropic. Musk called it an "impressive result."

Silicon Valley suddenly realized a question: Why did the United States let him go in the first place? However, there is a theory that Yang Zhilin was forced to leave by the US immigration policy. This explanation is incomplete. Salakhtinov emphasized that Yang Zhilin has many ways to stay in the United States. An Apple executive even offered Yang Zhilin a position in Apple's Beijing office to accommodate his desire to return to China for development. However, Yang Zhilin chose to start his own AI company and named it after his favorite rock band Pink Floyd's album "Dark Side of the Moon."

The attractiveness of China’s AI ecosystem

The deeper reason lies in a changing trade-off: the attractiveness of China’s AI ecosystem is increasing. Yang Zhilin’s development trajectory reflects this change. He graduated from Tsinghua University and participated in writing top papers such as Transformer-XL while at Carnegie Mellon University. These research results have promoted the development of the Transformer architecture that currently supports large language models. Within three years of founding Dark Side of the Moon, his team of 300 people had built a product that could compete with the world's top models.

Kimi K3 causes a stir

Would he have achieved the same success if he had founded Dark Side of the Moon in the United States? Maybe. But several Chinese entrepreneurs with similar backgrounds told me that they found it easier to build a startup in China. This is partly because they have wider networks and a more familiar cultural environment at home. Some entrepreneurs also mentioned that China has a unified regulatory system and deeper and more efficient talent resources.

One Chinese entrepreneur told me that although his startup would likely receive a valuation at least ten times higher if developed in the United States, in China he has a better chance of developing AI that surpasses human capabilities in tasks such as autonomous driving within five years.

In the past, most Chinese students studying in the fields of science, mathematics and technology chose to stay in the United States after receiving their PhDs in the United States. But according to reports, the number of people returning home is now increasing. Even among those who choose to stay, an increasing number are beginning to return home mid-career. One reason is that some people are worried about the political and social environment in the United States. A Stanford University survey of 1,304 Chinese scientists working in the United States in 2024 showed that 73% of respondents said they felt unsafe as academic researchers. 65% of them stated that they were worried about "hatred and violence against Asian Americans."

The discussion surrounding Yang Zhilin’s departure from the United States has become a microcosm of the United States’ broader concerns about the brain drain of AI. Some people in Silicon Valley call for a green card to be directly issued to every AI doctor. But this view ignores a deeper structural reality.

China is no longer just a net importer of talents trained by the United States. It has established its own talent training system. Stanford University's Hoover Institution tracked 356 researchers behind DeepSeek's core model and found that 53.5% of them had never studied or worked outside of China. Of those researchers with U.S. experience, 70% eventually returned to China.

Yang Zhilin belongs to such a group of pioneers. For them, returning home is not a last resort but a first choice. His decision was driven not by visa uncertainty but by entrepreneurial conviction. However, the message sent by the United States to the next batch of "Yang Zhilin" is very important. The U.S. visa system is primarily built around employment, not entrepreneurs. This doesn’t work so well for founders.

For decades, America’s technological dominance has been built on a simple formula: attract the world’s brightest minds and keep them. Yang Zhilin's story shows that this model is being tested, not only because the United States has become less friendly, but also because China has become more attractive.

The question is not whether Yang Zhilin made the right choice, but how many people like him will make the same decision in the future. They firmly believe that the future they want to create lies elsewhere.