signal/raiseTrack rounds
Series ARobotics & hardware · Palo Alto

Rhoda AI exits stealth with $450m Series A at a $1.7bn valuation to train robots from video

The Palo Alto company, led by Jagdeep Singh, says its video-predictive FutureVision system lets robots adapt to changing factory conditions; Premji Invest led the round.

s/rFunding deskNorth America · 2 min read

Rhoda AI, a Palo Alto robotics company, emerged from stealth on 10 March 2026 with a $450m Series A round that values it at $1.7bn, Reuters reported. SiliconANGLE said Premji Invest led the round.

Reuters named Khosla Ventures, Temasek, Mayfield, Premji Invest and Capricorn Investment Group among the backers. Pulse 2.0 listed a longer group that also includes Leitmotif, Matter Venture Partners, Prelude Ventures and Xora, together with Silicon Valley investor John Doerr.

“The goal is simple: robots that work in the real world, not just controlled lab settings.”Jagdeep Singh, cofounder and CEO, Rhoda AI

What Rhoda AI does

The company unveiled FutureVision, which it describes as the intelligence layer of its robotics systems. It is built around a model the company calls Direct Video Action, or DVA, which treats robot control as video prediction. According to The Robot Report, the system continuously observes its environment, predicts future states as video, converts the predictions into actions, executes them and then observes again, repeating the loop every few hundred milliseconds.

Rhoda says it first pre-trains the model on hundreds of millions of internet videos to learn motion and physics, then post-trains it on a smaller amount of robot data. It claims new tasks can often be learned from as little as ten hours of teleoperation data, a figure the company has not had independently verified in the sources reviewed.

Jagdeep Singh, the cofounder and chief executive, told Bloomberg, as quoted by SiliconANGLE, that robots trained by teleoperation can fail if something such as a phone's orientation changes, whereas Rhoda's model has seen many examples of objects in different orientations and can generalise.

Customers and early results

The Robot Report said Rhoda works with industrial partners in manufacturing and logistics. The company said that in a recent high-volume manufacturing evaluation a robot completed a component-processing workflow in under two minutes per cycle without human intervention, meeting or exceeding the customer's performance targets. SiliconANGLE reported tests using off-the-shelf parts with an automotive company.

The company expects FutureVision to power its own robotics platforms and, over time, to be licensed to partners using different hardware and software. SiliconANGLE said Singh also plans to build humanoid-style robots and to make Rhoda's own hardware to ensure quality for real-world work.

What the money is for

Rhoda said the Series A will fund continued research and engineering, expansion of industrial deployments and customer pilots, and growth of a team spanning generative AI, computer vision and robotics. Pulse 2.0 named Eric Ryan Chan, formerly a generative model architect at World Labs, as chief science officer, and Stanford professor Gordon Wetzstein as a contributor to the leadership team.

Singh said the goal is robots that work in the real world, not only in controlled laboratories. Reuters noted that industry experts caution that reliability, safety certification and cost remain key hurdles for large-scale deployment of general-purpose robots. The sources did not report Rhoda's founding year, headcount or earlier funding.

Company website: rhoda.ai