Dream Labs Raises $90M to Give Any Robot a Brain

Dream-1 uses zero-shot learning to control unfamiliar robots via plain-English commands — no task-specific programming required.

By The Robotics Desk

Dream Labs has raised $90 million in seed funding and shipped Dream API, a public interface that converts a robot's camera feed and plain-English instructions into motor commands — even for hardware the model has never encountered before. [1]

The problem it is targeting is structural: factories worldwide run a record 5 million industrial robots, yet nearly every one is programmed for a single task, and teaching one robot one task can take specialists months. [1] Low-cost robots now sell for a few thousand dollars but ship with no built-in intelligence. [1]

[1] Because it learns from video rather than from a single machine, its capabilities transfer across hardware from different manufacturers — zero-shot, out of the box. [1]

Dream API is live today. Developers connect any robot with grippers, type a command, and the model handles execution via Dream Labs' global data-center network. [1] For unsupported hardware, a small batch of demonstration data returns a tailored model within hours. [1] The company also runs a browser-accessible robot playground at its San Francisco office — live, unedited, no robot purchase required. [1]

B Capital and Kindred Ventures led the round, with participation from A.Capital, Unusual Ventures, SV Angel, Wing VC, Scribble VC, and Adrian Aoun, among others. [1]

The bet is that intelligence, not mechanics, is the bottleneck — and that a model trained on human movement can dissolve the wall between one machine and the next.

Dream Labs plans to spend the new capital primarily on compute as it expands the tasks and robot types Dream-1 can handle. [1]