The 2026 World Internet Conference opened in Wuzhen on November 3. At the opening ceremony, Joseph Tsai, co-founder and chairman of Alibaba, presented a set of figures: Alibaba Qwen has open-sourced more than 470 models, with cumulative global downloads exceeding 4 billion and more than 330,000 derivative models. In his view, as AI demand keeps climbing and computing power remains in short supply, "only open source can make it widespread"—open source can bring down R&D costs for small and medium-sized enterprises, which is especially critical for developing countries.

"Opening up like a top university": the open-source philosophy of Chinese companies

Tsai compared Alibaba's approach to a top university that opens its doors to the whole world and even offers financial aid: AI should also open-source its advances so more people can benefit. This comes from a philosophical choice by Chinese companies. His earlier view also carries through: according to an October Bloomberg report, he urged Europe that if it wants to break its dependence on China and the United States for computing power, the best path is to embrace open-source AI, and he called on Europe to build its own computing infrastructure to train and run models.

Taking a longer view, he predicts AI will eventually fade from being an explicit topic to becoming invisible infrastructure. Five years from now, people may no longer talk about AI as a separate subject, because it will seep into every part of business just like the internet does today—the terminology will change, but technology's fundamental reshaping of society and business will not stop. He repeatedly stresses that "only open source can make it widespread." In essence, this is about releasing AI capabilities from the data centers of a few giants: when weights, training paradigms, and toolchains can all be shared, teams that lack computing power can also stand on the shoulders of giants to build applications.