The round in brief
Dream Labs, a San Francisco AI startup founded in 2026 by former NVIDIA robotics researchers, announced $90 million in seed funding on October 19, 2026. The round was led by B Capital and Kindred Ventures, with participation from A.Capital, Unusual Ventures, SV Angel, Wing VC, Scribble VC, and angel investor Adrian Aoun. The company also launched Dream API — described as the first publicly available API that turns a robot's camera feed and a plain-English instruction into motor commands, even for robots the model has never seen.
The team and its lineage
Co-founder and CEO Joel Jang was a senior research scientist at NVIDIA's GEAR Lab, where he led the world-model team for Project GR00T — NVIDIA's flagship humanoid-robotics foundation-model program. His team's DreamZero paper, published in February 2026, described a 14-billion-parameter world-action model that adapted to an entirely new robot with just 30 minutes of play data. The Dream Labs founding team is drawn from that same GR00T-era research lineage, one of the deepest talent pools in robot learning.
The problem: hardware is ready, software isn't
The company's thesis starts from a stark industry fact: 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. Meanwhile, low-cost robots now sell for a few thousand dollars but ship with no built-in intelligence. Teaching one robot a single task can take specialists months; a new task or new hardware means starting from zero.
"The hardware is ready. What's been missing is a brain any robot can use," said Jang. His analogy: "When you get into a rental car, you don't relearn how to drive. You find the controls and your past driving experience does the rest. Dream-1 approaches a new robot the same way, so a new robot can do useful work in minutes, not months."
Dream-1, Dream API, and the product suite
The flagship model, Dream-1, is a foundation model trained on video of people performing everyday tasks rather than on any single robot. It predicts what will happen next in a scene and translates that into the action a robot should take — which is why its capabilities transfer across hardware from different manufacturers. Out of the box, Dream-1 can control any robot with grippers in a zero-shot setting, no additional training required.
The launch included three developer-facing products:
- Dream API — camera feed + plain-English instruction in, motor commands out;
- Fine-tuning as a Service — developers submit a small batch of demonstration data and receive a model tailored to unsupported hardware or new tasks within hours;
- Robot Playground — a browser-based tool letting anyone send commands to a live robot in Dream Labs' San Francisco office and watch it respond in real time.
Dream Labs works with more than 20 hardware partners, supporting machines priced between roughly $2,000 and $10,000. Dream-1 currently handles tabletop tasks — picking, placing, folding laundry — with wheeled mobile robots and more dexterous hands planned. Much of the funding will go to computing infrastructure, according to TechFundingNews.
The strategy: sell the model, not the robot
Dream Labs is explicitly avoiding the capital-intensive path of building its own robots. By positioning Dream-1 as a horizontal intelligence layer across manufacturers, the company sidesteps hardware margins and hardware risk, collecting instead on API usage and fine-tuning services — a "foundation model for robots" playbook that mirrors how LLMs scaled across industries. Notably, Jang is shipping a working model people can test themselves — "instead of edited demos" — an unusually transparent move in robotics.
Why it matters for founders
The Dream Labs playbook is a template for competing in hardware-adjacent AI without building hardware: own the intelligence layer, partner with everyone making the machines, and let 20+ hardware partners become your distribution channel. The Robot Playground also demonstrates a powerful trust-building tactic — let skeptics touch the live product. Founders in robotics, industrial IoT, or any physical-world AI should ask whether an API model gets them to revenue faster than a device.
Why it matters for investors
A $90M seed — larger than many Series B rounds — reflects how crowded the race for robot foundation models has become and how scarce GR00T-caliber teams are. The bet is that transferability across hardware is the scarce asset, and Dream-1's zero-shot gripper control plus video-based training is a genuinely different technical path from imitation-learning startups. The risks to watch: whether "works on any gripper" survives contact with real-world factory floors, and whether compute costs scale faster than API revenue as users grow.
Angeled.com connects founders with investors.
Sources: TechFundingNews (October 19, 2026), citybiz (October 19, 2026), FinSMEs (October 20, 2026).
Read Next
General Translation Raises $27 Million Series A Led by a16z to Build Agent-First Localization Infrastructure
LangChain Raises $300 Million Led by CapitalG at a $3 Billion Valuation to Build the Agent Stack
Mount Thor Raises $45 Million to Run Cloud Fleets of Macs for AI Agents
OpenAI Reportedly Raises $15 Billion at a $1.25 Trillion Valuation
Q-CTRL Raises $290 Million Series C Led by Salesforce Ventures, Reaching Unicorn Status
SE3 Raises $30 Million Series A Led by Accel



