In a single week in March 2026, Mind Robotics ($500M), Rhoda AI ($450M), Sunday ($165M unicorn), and Oxa ($103M) collectively raised over $1.2 billion. Combined with Skild AI’s $1.4B round in January and Physical Intelligence’s ongoing $1B raise, robotics is on pace for over $20 billion in funding this year.
The thesis driving this capital: foundation models that worked for language and vision will work for physical manipulation.
The Billion-Dollar Bets
Skild AI — $1.4 billion at $14 billion valuation (January 2026) Led by SoftBank with NVIDIA, Bezos Expeditions, and Macquarie Capital. The Pittsburgh company builds general-purpose robotic software that can be retrofitted across different robots and tasks. Valuation tripled in seven months from $4.5B to $14B. Already generating ~$30M in live revenue.
Physical Intelligence — ~$1 billion at $11 billion valuation (in talks, March 2026) Founders Fund leading, with Lightspeed, Thrive, and Lux participating. The San Francisco company’s π0 model trained on 10,000 hours of real-world robot data across 68 tasks. Would double valuation in four months.
Figure AI — $1 billion+ at $39 billion valuation (September 2025) The BMW deployment milestone—11 months of daily 10-hour shifts, 90,000+ parts loaded, 30,000+ vehicles contributed to—gave investors confidence to push valuation from $2.6B to $39B in 18 months.
The Sub-Billion Heavyweights
Mind Robotics — $500 million Series A Rivian spin-out led by founder RJ Scaringe. Training AI systems on Rivian’s manufacturing data, deploying within Rivian facilities. Total raised: $615M since late 2025.
Rhoda AI — $450 million Series A at $1.7 billion valuation Palo Alto startup with FutureVision robotics intelligence platform built on video-predictive control. Emerged from stealth in March. Premji Invest led.
Sunday — $165 million Series B at $1.15 billion valuation Coatue-led round for the humanoid startup building household robots. Their Memo robot targets laundry and table-clearing. Beta deliveries planned for late 2026.
Oxa — $103 million Series D first close UK autonomous driving for industrial vehicles. NVIDIA and UK National Wealth Fund participating. Focus on ports, airports, manufacturing sites.
Why Now?
Three factors converged:
Foundation models proven elsewhere. GPT showed scaling works for language. Vision transformers showed it works for images. Investors are betting the same dynamics apply to robotics—collect enough diverse physical data, train large enough models, and robots generalize.
Data flywheels starting. Figure’s BMW deployment generated real operational data. Mind Robotics can train on Rivian’s manufacturing. Physical Intelligence open-sourced π0 to accelerate community contribution. The chicken-and-egg problem of needing deployments to get data to improve deployments is starting to break.
Hardware-software decoupling. Skild AI’s retrofit approach means robot hardware becomes commoditized while the foundation model layer captures value. This shifts the investable thesis from “who builds the best robot arm” to “who builds the best robot brain.”
The Acquisition Angle
Amazon isn’t just watching. In March, they acquired both Fauna Robotics (humanoid maker of the 3.5-foot Sprout) and Rivr (Swiss stair-climbing delivery robots). Terms undisclosed, but the dual acquisition signals consumer robotics ambitions beyond warehouse automation.
Who Wins, Who Loses
Winners:
- Foundation model robotics companies that achieve real deployment scale
- NVIDIA, which appears in nearly every major round as a strategic investor
- Companies with proprietary data flywheels (Rivian/Mind, Figure/BMW)
- Open-source contributors if Physical Intelligence’s π0 release accelerates the field
Losers:
- Traditional robotics companies without foundation model capabilities
- Late entrants who can’t match the data advantage of early deployers
- Potentially everyone, if the foundation model thesis doesn’t transfer to physical manipulation the way investors assume
The Reality Check
Despite the funding frenzy, commercial robotics remains hard. Figure’s BMW success is notable precisely because it’s rare. Most humanoid demos remain demos. The gap between folding laundry in a controlled environment and folding laundry reliably across millions of households is enormous.
At $20B+ annual funding pace, investors are betting that gap will close. The next 18 months will reveal whether robotics’ ChatGPT moment is imminent or whether the physical world remains stubbornly harder than language.