内容来源：https://aibusiness.com/generative-ai/-1-billion-funding-for-spatial-intelligence-startup

内容总结：

由知名AI科学家李飞飞于2023年创立的World Labs公司近日宣布完成新一轮融资，估值或达50亿美元。该公司专注于空间智能技术，致力于通过生成式AI构建可理解物理规律的世界模型，其首款产品Marble已于2025年11月发布，支持用户通过文本或图像提示生成可编辑的3D虚拟场景。

本轮融资吸引了英伟达、AMD及3D设计软件巨头欧特克等重量级投资者参与，其中欧特克投资2亿美元并宣布将成为World Labs的技术顾问。公司自2024年9月以10亿美元估值亮相后，目前已累计获得2.3亿美元融资。

Marble平台提供免费及每月20至95美元不等的付费服务，其生成的3D环境可应用于游戏、沉浸式媒体、机器人仿真、建筑设计等多个领域。欧特克总裁安德鲁·安纳格诺斯特在声明中强调，该投资将推动"物理世界AI"发展，让数字智能真正理解空间、材料与物理规律。

李飞飞指出，下一代AI的关键突破在于让系统不仅能理解语义，更能认知几何、物理与动态规则构成的三维世界。此前她因创建ImageNet图像数据库对深度学习发展产生重要影响，此次创业再度引发行业高度关注。

中文翻译：

由谷歌云赞助
选择您的首批生成式AI应用场景
要开启生成式AI之旅，首先应关注能够提升人类信息交互体验的领域。

这家公司已引发广泛关注并承载着巨大期望。
Startup World Labs的新一轮融资是其推进空间智能技术发展的重要举措，该技术可使用户通过文本或图像提示生成3D虚拟世界。该公司由知名AI计算机科学家李飞飞于2023年创立，已吸引英伟达、AMD及软件设计巨头欧特克等重量级投资者，其中欧特克注资2亿美元。

World Labs于2024年9月以2.3亿美元融资结束隐匿模式，当时估值约10亿美元。
今年早些时候，彭博社曾推测该公司后续融资估值可能达到50亿美元。

据World Labs官网公告称：“我们致力于通过构建世界模型加速推进空间智能发展，这些模型将为叙事创作、创意产业、机器人技术、科学探索等领域带来革命性变革。”

该公司的首款产品“Marble”于2025年11月发布，支持用户生成3D虚拟世界。这些3D环境可编辑、可下载，并能应用于多个领域。World Labs指出其潜在应用场景包括游戏与沉浸式媒体、机器人技术与仿真模拟、建筑与设计等领域。

Marble提供免费版本及三档付费方案：标准版（每月20美元）、专业版（35美元）和高级版（95美元）。

欧特克作为3D计算机辅助设计软件领域的领军开发商，其总裁兼首席执行官安德鲁·阿纳格诺斯特在博客中阐释了投资World Labs的缘由：“对于欧特克所服务的行业而言，真正产生影响力需要AI理解空间、结构、材料、物理和时间。通过对World Labs的投资，我们正在推进物理世界AI的发展，这是将数字智能应用于世界设计与建造的下一个重大突破。”

李飞飞在声明中强调：“AI要真正发挥作用，就必须理解世界而不仅是文字。世界由几何、物理和动力学规律支配，实现语义、空间与物理的融合将是AI领域的下一个重要前沿。”

除投资外，欧特克还将担任World Labs的顾问，在研究与新模型开发方面展开紧密合作。

李飞飞是AI领域最具影响力的人物之一，其创建的ImageNet开源数据库包含数百万张带标签图像，为深度学习模型的训练奠定了重要基础。

英文来源：

Sponsored by Google Cloud
Choosing Your First Generative AI Use Cases
To get started with generative AI, first focus on areas that can improve human experiences with information.
The company has generated substantial buzz and sizeable expectations.
Startup World Labs' new funding round is part of its quest to advance spatial intelligence, a technology that enables users to generate 3D virtual worlds from text or image prompts
The company, founded in 2023 by renowned AI computer scientist Fei-Fei Li, has attracted some significant investors, including Nvidia, AMD and software design giant Autodesk, which contributed $200 million.
World Labs emerged from stealth in September 2024 with funding of $230 million, when it was valued at about $1 billion.
Earlier this year, Bloomberg speculated that any new investment World Labs attracted would be based on a $5 billion valuation.
"We are focused on accelerating our mission to advance spatial intelligence by building world models that revolutionize storytelling, creativity, robotics, scientific discovery and beyond," according to a post on the World Labs website.
The vendor's first product, Marble, was released in November 2025 and lets users generate 3D virtual worlds.
These 3D environments are editable and downloadable, and can be deployed across a number of areas, with World Labs citing gaming and immersive media, robotics and simulation, and architecture and design as potential applications.
Marble is available for free, and in three paid tiers -- Standard ($20 a month), Pro ($35) and Max ($95).
Autodesk is one of the most prominent developers of 3D computer-aided design software. President and CEO Andrew Anagnost explained the company’s backing for World Labs in a blog post.
"For the industries Autodesk serves, real impact requires AI that understands space, structure, materials, physics, and time," he said. "Through our investment in World Labs, Autodesk is advancing physical-world AI, the next major step in applying digital intelligence to how the world is designed and built."
“If AI is to be truly useful, it must understand worlds, not just words. Worlds are governed by geometry, physics, and dynamics, and reconciling the semantic, spatial, and physical is the next great frontier of AI," Li said in a statement.
Alongside its investment, Autodesk said it will work as an advisor to World Labs, collaborating closely on research and new models.
Li is among the most notable names in AI, particularly for her work on ImageNet, an open source database of millions of labeled images used in training deep learning models.