Analysis
Dream Labs has raised $90 million in seed funding to build foundation models that can operate many different kinds of robots from a single AI system, according to Pulse 2.0. 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's flagship model, Dream-1, is trained largely on video of people performing everyday tasks rather than on data collected from one specific robot. An accompanying API turns camera feeds and plain-language instructions into motor commands, so the pitch is that any robot body -- an arm, a wheeled platform, a humanoid -- can run on the same underlying model instead of needing its own bespoke training pipeline.
Why a $90M seed, not a Series A
Dream Labs was founded in 2026 by CEO Joel Jang, a former Nvidia robotics researcher, along with colleagues from the same team. Rounds this large at the seed stage are rare outside of founders with a research pedigree investors are willing to underwrite before a product exists -- in this case, the Nvidia robotics lineage appears to be doing that work. The company has not disclosed a valuation, revenue, a single paying customer, or how many robots its model has actually controlled outside a lab.
The competitive field already has a head start
Dream Labs is entering a category where the capital bar is already high. Figure AI, Physical Intelligence and Skild AI have each raised large rounds since 2024, with valuations that now range from roughly $1.5 billion to the low double-digit billions, detailed below. Our funding tracker has logged the broader run of nine- and ten-figure robotics checks this cycle, and Dream Labs' seed joins that list at the earliest stage yet.
- Figure AI -- $675M Series B near a $2B valuation (Feb. 2024), backed by Nvidia and Bezos Expeditions; pairs its own humanoid hardware with a model co-built with OpenAI and Microsoft.
- Physical Intelligence -- $400M at roughly $2B (2024), no single hardware partner; reportedly sought $1B at an $11B valuation this year.
- Skild AI -- $300M Series A at $1.5B (2024), backed by SoftBank, Coatue and Lightspeed; aims for one model across industrial and humanoid robots.
What the headline misses
None of Dream Labs' investors or the company itself have disclosed a deployed customer, a pilot partner, or a timeline for commercial availability. The harder, unresolved question for every company on this list, Dream Labs included, is whether a model trained mostly on human demonstration video actually transfers to robots operating in unstructured environments -- warehouses, homes, construction sites -- where failure means a dropped object or a stalled line, not a wrong chatbot answer. That gap between a well-funded thesis and a working product in the field is the central risk the whole robotics-foundation-model category carries right now.
For founders, the round is another sign that seed-stage capital is concentrating in a handful of categories -- robotics foundation models, agentic coding, inference infrastructure -- at check sizes that used to require a Series A track record. For LPs, it raises a blunter question: how many multibillion-dollar robotics bets can the market actually support before some of them consolidate or fold.
