Alibaba's heavy investment fuels the strong rise of the AI video track
On the 14th, AISHE Technology, a company specializing in AI video generation and real-time world models, announced the completion of its overall Series C financing, with a total accumulated amount reaching 2.98 billion RMB.
Among this round, the latest closed Series C+ was led by Alibaba, with over ten institutions participating as follow-on investors, including Lollapalooza Capital (Wang Huiwen Family Office), Ivy Capital, Huiyuan Capital, Zhongding Capital, South Korea's Mirae Asset, the Lion X Fund under OCBC's ecosystem, BlueFocus, CloudAlpha, and iGlobe Partners.
A reporter from the Science and Technology Innovation Board Daily noticed that merely over three years after its founding, AISHE Technology, established and led by Wang Changhu, has raised nearly 3 billion RMB solely in its Series C financing. Compared to the financing scale of domestic large model and AI video peers, the company has already entered the first tier. This top-tier financing in the AI video sector this year also confirms the industry consensus: AI video has taken over the baton from AI coding, becoming one of the core tracks that are starting to see large-scale commercial implementation.
Founder is a core backbone of ByteDance AI Lab
AISHE Technology was founded in 2023, dedicated to the research and development of multimodal large video models and world models, while providing diverse native AI video products based on self-developed models. The company's product portfolio includes PixVerse for global users, "Paiwo AI" targeting the Chinese market, as well as a series of models, products, and capabilities for team production, developer integration, and future interaction.
In early 2026, AISHE Technology launched PixVerse R1, a general-purpose real-time world model supporting 1080P resolution. According to the company, its users are spread across 177 countries and regions worldwide.
The founder of AISHE Technology is Wang Changhu, who holds a PhD degree from the University of Science and Technology of China. Wang joined Microsoft Research Asia in 2004, being one of the earliest leading figures in China to promote the R&D and industrialization of AI video models. He joined ByteDance in 2017 as the director of AI Lab, and in March 2023, he embarked on his first entrepreneurial attempt by establishing AISHE Technology.
Wang Changhu's entrepreneurial mindset was to avoid the then-popular NLP large models, start from the vision and multimodal fields that he was familiar with and interested in, develop large AI video models, and build large model products targeting global users.
It is reported that many members of Wang Changhu's core team come from the ByteDance ecosystem. He first participated in the zero-to-one R&D of Douyin and TikTok starting in 2017, responsible for building the video artificial intelligence behind this product with hundreds of millions of users, accumulating practical hands-on experience. This work allowed his team to cover almost all fields related to video intelligence, including but not limited to data processing, content generation, security issue handling, precise understanding of video content, and comprehensive advertising scenarios.
In August 2023, AISHE Technology completed its angel round financing, raising tens of millions of RMB. This financing was rapidly closed four months after the company's establishment.
Regarding the Alibaba-led consortium that led this round of financing, apart from investing in AISHE Technology across multiple rounds, it has also invested in Shengshu Technology in this track, and in July 2026, Alibaba Cloud participated in Keling AI's nearly 3 billion USD financing.
Notably, Lollapalooza Capital, the family office of Wang Huiwen, also appeared in the list of investors in this round. Wang Huiwen has continuously made moves in the AI field in recent years, and his investments are often interpreted by the market as a personal expression of confidence in the next-generation technology platform. At the same time, the participation of marketing and communications groups such as BlueFocus provides industrial implementation implications.
Capital is accelerating into the AI video track
If 2025 was a big financing year belonging to AI Coding, then in 2026, the spotlight is rapidly shifting to AI video. Since 2025, the domestic AI video generation track has continued to receive increasing capital injections, with the total publicly disclosed financing amount exceeding the 10-billion-RMB level.
Among them, Keling AI under Kuaishou closed a single round of financing of nearly 3 billion USD (approximately 20.4 billion RMB) in July 2026, setting the record for the largest single financing in the global large video model track to date, with leading internet companies including Tencent, Alibaba, and Baidu participating in the investment, a rare case of joint participation;
During the same period, after AISHE Technology secured a 60-million-USD Series B financing led by Alibaba in September 2025, it completed the 2.98-billion-RMB accumulated Series C financing in July 2026. Together with other key industry events such as Yanyu Technology's nearly 300-million-USD Series B+ financing and Sand.ai's two rounds of financing totaling over 100 million USD, capital is continuously concentrating on leading technology-oriented enterprises.
At the same time, leading projects have successively announced capitalization developments: Keling is rumored to submit its formal IPO application form to the Hong Kong Stock Exchange as early as the beginning of 2027 to launch the IPO process; regarding Shengshu, the reporter previously exclusively learned from multiple sources close to Shengshu Technology that the company may launch its Hong Kong stock IPO as early as the first half of 2026.
It is worth noting that the technological core of the AI video track is being upgraded. In 2024, all players were competing on the "duration and clarity of generated videos", but from the second half of 2025 to the present, one of the core competitive focuses has shifted to "world models".
ByteDance's Seedance 2.0, relying on in-depth modeling of the dynamic laws of the physical world, has achieved breakthroughs in multi-shot editing, camera movement control, cross-scenario character consistency and other aspects, ranking among the global top tier.
Shengshu Technology takes the Foundation World Model as its core underlying architecture, building a World Generation Model (WGM) for the digital world and a World Action Model (WAM) for the physical world respectively.
Xie Xuzhang, co-founder of AISHE Technology, recently demonstrated PixVerse Game, the first real-time video game engine built on world model technology, at the UN AI for Good Summit. This engine combines real-time video models, game mechanism toolchains, and user-customizable worldviews. Xie Xuzhang told the Science and Technology Innovation Board Daily reporter that game scenarios are the rigorous testing ground for world models and real-time video technologies.
Wang Changhu judged to the Science and Technology Innovation Board Daily reporter that the world model itself is still in its early stage, different teams do not have exactly the same goals and technical paths, and he personally categorizes them into two types. The first type is constructing a digital world. The model does not just generate a segment of content, but can generate a sustainable, interactive, and mutable digital environment. For example, the exploration of Li Feifei's team in the direction of spatial intelligence, and Google's research on interactive environment generation in the Genie series.
The second type is influencing the physical world through the digital world. With the development of embodied intelligence, robots, and agents, this type of world model has also received increasing attention. Its purpose is not only to generate a viewable world, but to enable the model to form an understanding of physical spaces, action results, and environmental feedback, ultimately serving decision-making and execution in the real world.
"So I am reluctant to use a conceptual label to judge which one is real or fake. The key is not whether it is named a world model, but whether it has a practical system that allows users, agents, or devices to interact freely with this digital or physical world."
This article is from the WeChat official account "Venture Capital Daily", authored by Li Mingming, published with authorization from 36Kr.