ANALYSIS Dream Labs' $90 million seed round, one of the largest on record for a robotics startup, tests a specific thesis: that the value in robotics sits not in the hardware but in a shared software brain sold as a metered service. The bet arrives just as broader embodied-AI financing shows signs of contraction, making the deal both a vote of confidence in foundation-model robotics and a marker of how selective that confidence has become.
Why it matters
Dream Labs raised $90 million in seed funding announced October 19, 2026, led by B Capital and Kindred Ventures, with participation from A.Capital, Unusual Ventures, SV Angel, Wing VC, Scribble VC, and angel investor Adrian Aoun1,4,8. The company simultaneously launched Dream API, which it describes as the first publicly available API for a robot foundation model10,6. Dream Labs was founded in 2026 by former NVIDIA robotics researchers who worked on the company's GR00T research program7. CEO Joel Jang framed the opportunity bluntly: "The hardware is ready. What's been missing is a brain any robot can use"5. ◆ The round's size and its investor roster suggest that at least one cohort of venture capital still sees a massive opening in robotics software, even as the broader sector's deal pace slows.
The big picture
Dream Labs estimates more than five million industrial robots are working inside factories and warehouses globally, and nearly all of them are designed to perform a single task. The startup's flagship model, Dream-1, is trained on video of people performing everyday activities and uses that data to predict what should happen next, then translates predictions into motor commands. Dream Labs calls Dream-1 a "World Action Model" that predicts video and actions together: anticipating how a scene will change is linked to deciding what a robot should do next. Rather than issuing joint-level commands tied to a specific arm, Dream-1 predicts end-effector poses, the desired position and orientation of each gripper, along with whether it should be open or closed. The company says Dream-1 can control bimanual robots fitted with parallel grippers without robot-specific training beforehand. Dream Labs currently works with more than 20 hardware partners whose supported robots range in price from approximately $2,000 to $10,000.
Alongside the API, Dream Labs launched a fine-tuning service and a browser-based Robot Playground that lets users send instructions to physical robots in the company's San Francisco office. The API itself takes a robot's camera feed plus a plain-English instruction and converts both into motor commands, including on hardware the model has never seen before. Dream Labs says the funding will primarily support compute as it expands model capabilities and capacity.
ANALYSIS Dream Labs is making the opposite bet from the hardware-first robotics companies that have spent recent years chasing custom silicon for AI inference. As one report put it, Dream Labs argues "the bottleneck in robotics is not the chip inside the machine, it is the software brain that has to generalize across thousands of different chassis, grippers, and sensor rigs". The $3-per-hour API price point9 positions the service as an operating expense rather than a capital outlay, a model familiar from cloud computing but novel in physical robotics.
That positioning matters against the backdrop of a cooling embodied-intelligence funding market. Data from IT Juzi cited by 36Kr showed that as of October 15, only 19 financing deals had taken place in the embodied intelligence sector that month, totaling 49.34 billion yuan3. In contrast, August saw 91 deals with a total of 313.23 billion yuan. ◆ Dream Labs' ability to close a $90 million seed in this environment suggests investors are differentiating between hardware-heavy robotics plays and software-layer bets that can scale across existing installed bases. The company is not building a robot; it is selling intelligence to robots that already exist.
Dream Labs says it is working with more than 20 hardware partners. Dream Labs says Dream-1 can already handle tasks such as picking, placing, and folding laundry. ◆ Those partnerships and that task range point to an ecosystem strategy from day one, positioning Dream Labs as a middleware layer before any single robotics OEM locks in its own foundation model. The gap between tabletop demonstrations and the complex manipulation tasks industrial customers require remains the central technical question for the company's next phase.
What's next
Dream Labs said much of the $90 million will go toward compute infrastructure. The company's public API is now live, and the Robot Playground offers a real-time demonstration channel. With the embodied-AI financing environment tightening, Dream Labs' next milestone will likely be measured not in follow-on valuation but in how many of those 20-plus hardware partners convert from integration to production deployments.