Dream Labs, a San Francisco startup founded in 2026 by former NVIDIA robotics researchers, has raised US$90 million in Seed funding to scale a model that teaches new robots useful tasks in minutes rather than months. The round was led by B Capital and Kindred Ventures, with participation from A.Capital, Unusual Ventures, SV Angel, and Adrian Aoun, among others. Announced on October 19, the financing included Dream API, which turns a robot's camera feed and plain-English instructions into motor commands.
How Dream-1 Works
Dream-1 is trained on video of people performing everyday tasks, then predicts what happens next and determines the robot's resulting action. The company says learning from video rather than one machine allows skills to transfer across hardware and lets Dream-1 control any robot with grippers zero-shot. The approach builds on work by co-founder and CEO Joel Jang, who was a senior research scientist at NVIDIA's GEAR Lab and led the world-model team for Project GR00T.
Developer Access and Testing
Developers can connect a robot through Dream API, which runs Dream Labs' models in data centers worldwide, or send a small batch of demonstration data to a fine-tuning service that returns a tailored model within hours. A browser-based Robot Playground lets anyone type commands to a robot in the company's San Francisco office and watch it respond live and unedited. The company is intentionally offering a working model that people can test themselves, mistakes included.
Competitive Landscape
Dream Labs is not the only company selling intelligence for other people's hardware. Skild AI raised about US$1.4 billion in January 2026, while Rhoda AI raised US$450 million at a US$1.7 billion valuation in March and Genesis AI emerged from stealth with a US$105 million Seed in July 2025. Dream Labs is targeting a different buyer by supporting robots priced from about US$2,000 to US$10,000 and working with more than 20 hardware partners.
Pricing, Capabilities, and Hardware Plans
Dream-1 is priced from US$3 per hour, with a cheaper model available at 75 cents per hour. The system currently handles tabletop tasks such as picking and placing objects and folding laundry, with support for wheeled mobile robots and dexterous hands planned. Jang told Forbes that capable dexterous hands cost about US$100,000 today, and he expects that figure to fall to US$5,000 to US$10,000 by 2027.
Use of Funds and Industry Outlook
Dream Labs will spend the money primarily on compute to expand the tasks and robots Dream-1 handles and to support more users. Humanoid startups had raised US$8.6 billion by late July 2026, 1.8 times the 2025 total, according to Dealroom data cited by TFN. The International Federation of Robotics counted a record 5 million industrial robots in factories in 2025, up 9%, with more than 600,000 installed that year, up 11%.
The company is betting that a working, testable model will win developers even when rivals own their own hardware, data, or both. Jang says the hardware is ready and the missing piece is a brain any robot can use, similar to how a driver adapts to a rental car without relearning to drive. The next test is whether its mistakes become rarer as more robots connect and whether renting out intelligence is enough to compete against deeper-integrated rivals.