Dream Labs $90M Seed Builds the Brain for Everyday Robots

The Dream Labs $90M seed round announced October 19 is one of the largest seed financings in robotics this year: $90 million, led by B Capital alongside Kindred Ventures, with participation from A.Capital, Unusual Ventures, SV Angel, Wing VC, Scribble VC, and Adrian Aoun, backing a startup building what it calls the brain for everyday robots. Alongside the financing, the company launched the first publicly available API for a robot foundation model, according to its Business Wire release — a bet that robot intelligence will follow the same playbook as software AI: one general model, distributed by API.

The San Francisco company was founded in 2026 by former leads of NVIDIA's GR00T robotics research, including co-founder and CEO Joel Jang. Its flagship model, Dream-1, learns how the physical world works from video of people performing everyday tasks, then converts a robot's camera feed and a plain-English instruction into motor commands — even for robots it has never seen before. The Dream Labs $90M round exists to scale exactly that transfer.

Dream-1: video in, motor commands out

Instead of training each robot on its own sensor data, Dream-1 is trained on video of humans doing ordinary things. The model predicts what happens next in a task and determines the robot's resulting action, according to the company. Because it learns from video rather than from a single machine, its capabilities transfer across hardware: out of the box, Dream-1 can control any robot with grippers, zero-shot.

The launch products are aimed squarely at developers. Dream API translates visual inputs and natural-language prompts directly into robot actions — developers sign up, connect any robot with grippers, and the robot carries out the task, with Dream Labs running the models in data centers around the world so robots respond quickly wherever they are. For unsupported hardware or new capabilities, a fine-tuning-as-a-service offering lets developers and robot makers submit a small batch of demonstration data and receive a tailored model within hours. And for the merely curious, a browser-based Robot Playground lets anyone type commands to a real robot in Dream Labs' San Francisco office and watch it respond live, unedited, with no robot purchase required.

The baseline is solid but narrow: Dream-1 currently handles tabletop tasks such as picking and placing objects and folding laundry, and needs no extra data to control a two-armed robot with parallel grippers, according to the company. Support for wheeled mobile robots and dexterous hands is planned next. The company works with more than 20 hardware partners, and the robots it supports today cost between about $2,000 and $10,000.

The hardware got cheap. The brain didn't exist.

Dream Labs is arriving at a moment when the economics of robot bodies have flipped. Factories worldwide run a record 5 million industrial robots, yet nearly every one is programmed for a single task, and what it learns can't transfer to another machine, according to the company. Teaching one robot one task can take specialists months, and a new task or new hardware often means starting from zero. That brittleness is the opening the Dream Labs $90M financing is meant to exploit.

Meanwhile, cheap arms are flooding in from below. A capable robot arm has gone from the price of a luxury car to the price of a high-end laptop, according to B Capital general partner Daisy Cai — but it ships with no built-in intelligence. That gap between cheap bodies and missing brains is why B Capital led the Dream Labs $90M round. Jang's pitch, per the release, is that Dream-1 approaches a new robot the way a driver approaches a rental car: past experience transfers, the controls get found quickly, and useful work starts in minutes rather than months.

The hardware makers seem to agree. Ryan Chan, co-founder and CEO of MakerMods, which sells $2,000 bimanual robot arms, said his company's robot was up and running in minutes on Dream-1 and attempting tasks it was never programmed for — letting the company focus on building better hardware while the intelligence improves centrally, according to the release. That division of labor is the whole thesis: commoditize the body, centralize the brain.

Seed money with a compute appetite

Ninety million dollars is an enormous sum for a seed round, and the Dream Labs $90M raise tells you what the money is for: compute. The company says the funding will go primarily to compute as it expands the tasks and robots Dream-1 can handle and scales to support more users. Cai put the bet simply: the team did some of the defining work in robot learning at NVIDIA, and now anyone with an API key can build on it, according to the release.

The raise fits a wider pattern in agentic AI funding this month. AI Agent Store's agentic AI weekly roundup notes sustained investor appetite for agent-first companies, including TrueFoundry's $19 million Series A for agent and LLM-operations tooling and Goodweek's €10.69 million seed for an "AI Manager" aimed at small businesses. The same week also saw Town raise $108M to turn AI assistants into teammates — the software-agent side of the same argument Dream Labs is making for robots: assistants should be general, shared, and available on demand rather than custom-built per machine or per team. For more on this funding wave, see genznewz.com's AI News coverage.

What could still go wrong

The honest caveats are in the product's current limits. Tabletop tasks are the friendliest possible proving ground for robot learning; homes, warehouses, and streets are harder in ways that don't show up in a live demo. Twenty-plus hardware partners is a start, not an ecosystem. And a video-trained model that predicts "what happens next" can fail in ways no camera catches — which makes the Robot Playground's live, unedited stance as much a marketing bet as a technical one.

Still, the milestone is real. Robot foundation models have been demoed for years; the Dream Labs $90M launch puts the first public API for one in developers' hands, with zero-shot gripper control that works on hardware from other makers, according to the company. If the model improves centrally the way software models do, the value of every cheap robot body keeps rising after purchase — which would make the $2,000 arm the most interesting computer peripheral of the decade. The Dream Labs $90M bet, in short, is that the next platform war will be won by whoever owns the brain, not the body.