Palo Alto-based startup Rhoda AI announced that it has raised US$450 million in a Series A funding round, valuing the company at about US$1.7 billion.
The company plans to use the capital to expand development and deployment of artificial intelligence systems designed to train robots using internet-scale video data.
The round was led by Premji Invest and included investors such as Khosla Ventures, Temasek Holdings, and venture capitalist John Doerr.
Rhoda emerged from 18 months in stealth while unveiling its robotics platform, designed to enable machines to operate more effectively in complex and unpredictable industrial environments.
Chief executive Jagdeep Singh said the method helps robots adapt to unfamiliar conditions that might otherwise disrupt traditional models.
“We believe the next era of robotics requires models that understand how the world moves — not just what it looks like or how it’s described in language,” Singh said in a recent press release. The goal is simple: robots that work in the real world, not just controlled lab settings.”
The company calls its system a Direct Video Action model, which continuously observes its surroundings, predicts how the environment may change, and translates those predictions into robotic actions in real time.
According to Rhoda, the system updates its decisions every few hundred milliseconds in a closed feedback loop, allowing robots to adjust as conditions evolve.
“We believe the next era of robotics requires models that understand how the world moves — not just what it looks like or how it’s described in language,” Singh said. “By learning from internet-scale video and operating in closed loop, our systems are designed to adapt to real-world variability in ways conventional approaches struggle to achieve.”
Traditional robotics AI models are typically trained using teleoperation data, where human operators remotely control robot movements to generate training examples.
While effective in controlled settings, this method can limit how much data is available for robots to learn from. Rhoda’s approach instead uses millions of publicly available internet videos to teach robots about physical movement, object interactions and real-world dynamics.
The company then combines this large-scale video training with smaller amounts of robotic data to refine how machines execute tasks.
The company said its technology has already been tested in industrial settings using off-the-shelf robotic hardware inside an automotive manufacturing facility.
Rhoda plans to license its software platform to industrial customers and is also developing its own robotics hardware, including humanoid robots, to ensure the technology can operate effectively in production environments.
Securities Disclosure: I, Giann Liguid, hold no direct investment interest in any company mentioned in this article.