On September 1, 2026, AI 3D unicorn VAST announced the completion of its Series B and B+ financing rounds, raising approximately 3 billion yuan in total, led by Matrix Partners China.

This marks another major funding milestone for VAST within the year, following its over 1 billion yuan Series A3 round completed in July 2026. In less than six months, the company has accumulated approximately 5 billion yuan in total funding, setting a new record in the 3D-native intelligence sector.

The roster of strategic investors in this round is particularly noteworthy. Industry capital including Perfect World, BlueFocus, XinDongNeng Investment, Yanqu Games, Thundersoft, Guangdian Investment, and 37 Interactive Entertainment jointly participated. Financial investors such as CDH VGC, CICC Capital, CMC Capital, HongTai Fund, SanZheng Health Investment, Zhongping Capital, Fujian Industrial Investment Fund, Fujian Venture Capital, Zhuoyuan Asia, and Jiangxi Financial Holdings also joined the round. Existing shareholders including Dachen Capital, Primavera Capital, 4399, Muhua Kechuang, Oasis Capital, Ince Capital, and Huakong Fund continued to oversubscribe.

Looking back at the Series A3 round, game developers such as 4399, Tanwan Games, and Giant Network had already taken the lead. With Perfect World, 37 Interactive Entertainment, and Yanqu Games joining in this round, VAST's shareholder list now spans "half the gaming circle." Such concentrated betting on AI 3D technology by gaming companies is rarely seen in the primary market in recent years.

Founded by Song Yachen in November 2022 and headquartered in Beijing, VAST focuses on developing AI 3D large models and general world models. The company has built two core technology systems.

The first is the Tripo series of 3D-native foundation models, with its core product Tripo Studio positioned as an AI 3D workbench covering creative workflows including concept generation, modeling, texturing, segmentation, topology, rigging, animation, and export. Concurrently with this funding round, VAST released Tripo P2.0, the world's first 3D-native foundation model supporting native quad-mesh topology, enabling 3D generation to simultaneously meet three industrial-grade thresholds: "engine-ready, animation-ready, and editable."

The second is the Project Eden world model, whose core approach separates world state from visual rendering, allowing multiple users to act within the same persistent environment. For instance, when a player extinguishes a fire, the environment retains that change; two players driving vehicles on the same track share a synchronized world state even when viewing from different angles. For gaming, this addresses the most fundamental interaction requirement—actions taken by players must leave lasting results, and other players should see and participate in the same world.

On the commercialization front, VAST's AI 3D generation technology has deeply integrated into real production workflows across 3D printing, film and animation, industrial design, XR, gaming, and spatial design. In gaming, Tripo powers the 3D model segmentation used in NetEase's Eggy Party workshop, and has collaborated with Where Winds Meet on gameplay features. In 3D printing, Tripo has been integrated into Bambu Lab's MakerWorld. In film and animation, overseas creative platform Intangible AI and leading domestic video platforms have integrated Tripo models.

The wave of game companies taking equity stakes in VAST represents a collective industry response to the "industrialization bottleneck" in 3D content production. In game development, 3D assets such as characters, props, and environments require substantial production time. In traditional workflows, a single 3D character—from high-poly sculpting, topology optimization to UV unwrapping and weight rigging—often demands weeks or even months of manual labor. AI-generated 3D tools have long struggled with the "looks good but hard to use" problem—generated high-poly models appear visually impressive but contain excessive face counts and messy edge flow, requiring significant manual cleanup before they can enter a game engine.

Tripo P2.0 directly targets this pain point: it can generate 3D objects with quad-dominant faces, more accurate shape reconstruction, controlled face counts, and more reasonable part separation in seconds, making generated assets easier to carry through subsequent production stages. For game teams, even if a prop's appearance is close to expectations, if hours are still needed to clean up the mesh, the time saved in modeling could be consumed by rework.

From Perfect World and 37 Interactive Entertainment to 4399, Tanwan Games, and Giant Network, the collective equity participation by game developers is essentially a play to secure "tickets" to the next-generation 3D content production infrastructure. When AI 3D large models can increase game asset production efficiency by a hundredfold, any game company without this capability will find itself at a cost disadvantage over the next 3 to 5 years. When "half the gaming circle" simultaneously places its bets on the same AI 3D company, the positioning battle for next-generation 3D content production infrastructure has officially begun.