VAST raises 3 billion yuan to scale AI 3D foundation models

The scale of this investment indicates that 3D generative AI is moving from experimental use cases to industrial-standard production. By increasing triangle counts and adding native quad topology, VAST is addressing the technical barriers that previously prevented AI-generated assets from entering professional gaming and film pipelines. For the streaming ecosystem, this suggests a rapid reduction in the cost and time required to build high-fidelity interactive environments and virtual sets. As ByteDance and Tencent integrate these models, the industry should watch for a surge in user-generated 3D content that could challenge traditional high-budget production cycles.

VAST faces intensifying competition in the AI 3D generation space, where multiple startups and established platforms are racing to capture industrial demand. Nvidia and Meshy AI announced a partnership in May 2026 to integrate Meshy's text-to-3D pipeline directly into Nvidia Omniverse, targeting game studios and industrial simulation workflows that require production-ready mesh output. Meanwhile, Stability AI, which previously dominated open-source image generation, launched its Stable 3D 2.0 model in April 2026 with native quad topology and PBR material output, positioning itself as a direct competitor to VAST's Tripo P2.0 in professional content pipelines. The competitive density in this segment reflects broader investor conviction that 3D generative AI will become a foundational layer for gaming, film, and spatial computing platforms.

On the business and integration front, VAST's partnerships with ByteDance and Tencent place it at the center of China's platform economy. ByteDance integrated Tripo into its Jianying video editing suite in July 2026, enabling creators to generate 3D assets directly within short-form video workflows, a move that could expose the technology to hundreds of millions of content creators. Tencent, meanwhile, has been deploying Tripo models within its internal game development pipelines for prototyping, according to a report by LatePost in August 2026 that detailed Tencent's AI content strategy across its gaming division. Microsoft's involvement, referenced in the funding announcement, aligns with its broader push to embed generative 3D tools into Azure AI services and its Mesh collaboration platform. Matrix Partners China, which led the round, has previously backed companies in the AI infrastructure layer including Moonshot AI and Zhipu AI, signaling a portfolio strategy focused on foundational model companies across modalities.

Technical benchmarks for AI 3D generation have advanced rapidly in 2026. A comparative study published by researchers at Tsinghua University in June 2026 evaluated six leading 3D generation models on geometric accuracy, texture fidelity, and topology quality, finding that VAST's Tripo P2.0 achieved the highest scores on quad mesh quality and watertight surface generation, two metrics critical for downstream use in game engines like Unreal and Unity. The study noted that most competing models still produced triangle-dominant meshes requiring manual retopology, a step that adds significant production overhead. Separately, Epic Games announced in August 2026 that Unreal Engine 5.6 would include native support for importing AI-generated quad meshes with automatic LOD generation, a pipeline enhancement that directly benefits tools like Tripo that output production-ready topology. This convergence of model quality and engine-level support suggests that AI-generated 3D assets are approaching the threshold of professional viability at scale.