29-Year-Old Entrepreneur Raises $50 Million, Betting on World Models and User-Generated Content

VAST, a general artificial intelligence company, recently closed a $50 million Series A funding round co-led by Alibaba and Hengxu Capital, with participation from Yuanhe Puhua, BV Baidu Ventures, and Dongfang Jiafu. Existing investors Primavera Capital and the Beijing Artificial Intelligence Industry Investment Fund also increased their investments.

Founded in early 2023 by Song Yachen, an early co-founder of MiniMax, VAST positions itself as a general artificial intelligence company, with a core focus on AI 3D large models and world models. Its client roster includes leading companies such as Alibaba, Tencent, ByteDance, NetEase, SAIC, and Tuozhu. Its Tripo Studio platform has attracted over 6.5 million creators and serves more than 90,000 developers.

Earlier this year, Song Yachen revealed in a conversation with Tea House that the "World Model" is a key area they are currently focusing on.

According to Tea House, VAST plans to allocate the funds raised in this round primarily to the development of world models, the iteration of core algorithms, and the construction of a user-generated content (UGC) platform. Their first world model is set to be released in the near future.

The Intersection of AI 3D Generation and the Gaming Industry

The industry context for VAST’s latest funding round is as follows: AI-driven 3D generation and world modeling are emerging as key areas of focus in the global AI sector, with the gaming industry serving as the primary driver of demand.

According to a report by Meticulous Research, the global AI-powered 3D asset generation and texturing market is projected to grow from $1.95 billion in 2026 to $12.84 billion in 2036, at a compound annual growth rate of 20.8%. The gaming industry will be the primary driver of this growth.

According to a report by Third Point Ventures, nearly half of the budget in game development is directly related to asset production, making art production one of the most capital-intensive aspects of the industry. AI-driven 3D generation technology may become a key factor in reshaping the industry landscape.

Meanwhile, world models have become a new battleground for global tech giants. Last year, Li Feifei’s World Labs announced that it had developed the ability to generate “persistent, navigable, and customizable 3D worlds.” Musk’s xAI plans to release a world-model-driven AI-generated game by the end of 2026.

Companies such as Google DeepMind, Meta, and NVIDIA are also accelerating their development of world models. In China, Tencent has released the Hunyuan 3D World Model, while Ant Lingbo has open-sourced LingBot-World. Fields such as autonomous driving, embodied intelligence, and gaming are all seeking world model solutions tailored to their specific needs.

In a previous conversation with us, Song Yachen expressed the view that the ultimate form of world modeling should be based on 3D spatial data, rather than frame-by-frame generated video.

In his view, there is a progressive relationship between 3D large models and world models: the former addresses “creating objects,” while the latter addresses “creating worlds.” A world model must not only understand the geometric shapes of objects but also grasp physical laws and interaction logic—for example, a door should be able to open, and the direction in which it opens depends on the structure of its hinges and springs. Song Yachen believes that with the development of world models, the creation of many RPG-style games will become much easier.

VAST’s extensive collection of 3D data and system-level engineering expertise are considered its core competitive advantages in the field of world modeling.

Games are the entry point, but the goal goes beyond gaming

VAST's connection to the gaming industry runs through every stage of the company's journey, from its founding vision to the launch of its products.

Song Yachen is an avid gamer who enjoys strategy and role-playing games, and he once served as an alliance leader in *Rateshu Zhibin*. In 2019, while working on the “AI + Gaming” strategy at SenseTime, he frequently interacted with small and medium-sized game studios and noticed that the high costs and long lead times associated with art production were the biggest challenges facing independent developers.

His relevant experience has led him to a deeper understanding of the industry: the high barriers to entry not only limit the diversity of games but also distort business models—when development costs are extremely high, products are forced to prioritize the most profitable avenues, and the maturation of AI-driven 3D generation technology is expected to address this industry pain point.

On the product side, VAST has built a thriving ecosystem centered around game developers. The Game Hub community has attracted over 100,000 active developers and has produced more than 2,000 AI-interactive content projects to date. The inaugural Tripo AI 3D Game Jam received the highest number of submissions among similar global competitions, and the shortlisted entries will be showcased at this year’s GDC.

Along with the funding announcement, VAST also unveiled two core members of its next-generation AI 3D large-model family.

Tripo H3.1 is an updated flagship high-precision model that offers significant improvements in key metrics such as input alignment, structural accuracy, and texture quality. It can accurately reproduce the internal structures and textures of complex machinery, producing results that rival the work of professional 3D modelers.

Of particular note is the all-new Tripo P1.0 architecture. This model employs a brand-new algorithmic framework that moves away from traditional sequential generation methods, instead enabling the simultaneous generation of 3D structures within a unified probabilistic space. In practice, this results in the direct output of topologically clean, engine-ready 3D assets within 2 seconds—a speed increase of over 100 times compared to existing solutions. The Smart Mesh feature, based on this model, is now available in Tripo Studio.

The model’s power can significantly boost game production efficiency. However, Song Yachen repeatedly emphasizes that VAST’s goal is not to become a game production tool. He believes that as creation costs approach zero, a flood of interactive, expressive UGC content will emerge—content that isn’t “games” in the traditional sense, just as short videos aren’t “UGC movies.” This could give rise to platform-level opportunities.In the future, traditional games may account for only a small fraction of all interactive content.

The company plans to officially launch a UGC interactive content product for general users in 2026, allowing users to create, browse, and share interactive 3D content, integrating content generation capabilities, distribution mechanisms, and interactive systems into a complete closed-loop ecosystem.

This approach is not a new concept. From its inception, VAST sought to directly enter the interactive content platform space, but failed due to a lack of user-friendly creation tools, so it shifted its focus to developing large-scale AI 3D models. Now, as the models have matured and the user base has grown, the window of opportunity to resume platform development is opening.