In a conversation at the 2026 Future Science Awards, Zhang Yaqin, a foreign academician of the Chinese Academy of Engineering, raised a sharp commercial question to Tencent's chief AI scientist Yao Shunyu. What exactly is the model company's moat? Yao Shunyu's answer is that AI moats exist more in the product layer than in the model layer. What is really difficult to replace is not necessarily the smartest model, but AI products that can understand the user's background, use proprietary data and context, and gain the user's trust. After hearing this answer, Zhang Yaqin followed the trend and made a joke. He ridiculed that according to Yao Shunyu's logic, Anthropic may not be worth 1 trillion or even 2 trillion dollars in the future, because the model may one day be commercialized. Afterwards, he smiled and reminded the students at the scene to never listen to his suggestions to buy stocks. In addition to the commercial moat for AI companies, on-site communication also involved agent technology routes, model open source and closed source choices, and how AI can actually participate in scientific discovery.

In a conversation at the 2026 Future Science Awards, Zhang Yaqin, a foreign academician of the Chinese Academy of Engineering, raised a sharp commercial question to Tencent's chief AI scientist Yao Shunyu. What exactly is the model company's moat? Yao Shunyu's answer is that AI moats exist more in the product layer than in the model layer. What is really difficult to replace is not necessarily the smartest model, but AI products that can understand the user's background, use proprietary data and context, and gain the user's trust. After hearing this answer, Zhang Yaqin followed the trend and made a joke. He ridiculed that according to Yao Shunyu's logic, Anthropic may not be worth 1 trillion or even 2 trillion dollars in the future, because the model may one day be commercialized. Afterwards, he smiled and reminded the students at the scene to never listen to his suggestions to buy stocks. In addition to the commercial moat for AI companies, on-site communication also involved agent technology routes, model open source and closed source choices, and how AI can actually participate in scientific discovery.