Dream Labs, a San Francisco startup founded by former leaders of NVIDIA’s GR00T robotics research program, has raised $90 million in seed funding to build what it calls “a brain any robot can use.”
The round was led by B Capital and Kindred Ventures. The company, founded in 2026, is betting that robotics hardware is ready — what is missing is shared intelligence software.
Its flagship model, Dream-1, is a foundation model trained on videos of people performing everyday activities. It turns visual information into action, controlling robots equipped with grippers. The key claim: Dream-1 can run compatible hardware it has never seen before, without extra training — sparing manufacturers from writing separate software for every machine.
Alongside the financing, Dream Labs launched Dream API, a developer platform that turns camera inputs and natural-language instructions into robot movement commands, plus a fine-tuning service for developers who want customized models. The company also opened Robot Playground, a browser-based demo where anyone can send instructions to a robot in its San Francisco office and watch it respond.
Dream Labs says it already works with more than 20 hardware partners on machines priced between roughly $2,000 and $10,000. Its robots currently handle tabletop tasks like pick-and-place and folding laundry, with wheeled mobile robots on the roadmap.
“The hardware is ready. What’s been missing is a brain any robot can use,” said Joel Jang, co-founder and CEO. Daisy Cai, general partner at B Capital, pointed to falling hardware costs as the opening for software platforms that make robots easier to deploy.
Most of the new funding will go toward computing infrastructure to expand the model’s capabilities. To understand the technology, read our guides on foundation models and computer vision.
Source: citybiz
