Rhoda AI

Rhoda AI is an American robotics-AI foundation-model company that exited stealth in March 2026 after approximately 18 months of confidential operation. The company develops FutureVision, a video-predictive-control foundation model that observes its environment, predicts future states, and converts those predictions into robot actions in a closed feedback loop running every few hundred milliseconds. Rhoda AI exited stealth concurrently with a $450 million Series A funding round at a $1.7 billion post-money valuation, led by Khosla Ventures with Temasek, Mayfield, Capricorn Investment Group, Premji Invest, and technology investor John Doerr among the participating investors. Industry coverage has characterized the round as one of the largest early-stage investments in humanoid-robot intelligence to date.

At a glance

• Founded: Approximately mid-2024 (entered stealth); exited stealth March 2026.
• Status: Private. Series A in March 2026 at $1.7 billion post-money valuation.
• Funding: $450 million Series A in March 2026, led by Khosla Ventures with Temasek, Mayfield, Capricorn Investment Group, Premji Invest, and John Doerr participating.
• CEO: Not publicly disclosed as of stealth-exit reporting.
• Other notable leadership: Founders not publicly disclosed in the stealth-exit announcement; the company has been comparatively quiet on senior-leadership identification.
• Open weights: No. FutureVision is positioned as a commercial robot-intelligence foundation model rather than as an open-weights research release.
• Flagship outputs: FutureVision, the video-predictive-control robot-intelligence foundation model demonstrated across humanoid and industrial-robot platforms.

Origins

Rhoda AI was founded approximately in mid-2024 and operated in stealth for roughly 18 months before its public launch in March 2026. The stealth-exit announcement focused on the FutureVision platform and the $450 million Series A funding round rather than on founder identification, which is uncommon among well-funded robotics-AI insurgents but consistent with the company's industrial-customer-first commercial focus.

The founding research thesis was that robotic intelligence should be grounded in video-prediction rather than in language-model-style next-token-prediction. Where peer robotics-AI foundation models such as Physical Intelligence π₀ and Skild AI Skild Brain combine vision-language-model architectures with robotic-action conditioning, Rhoda AI's FutureVision approach treats the prediction of future visual states from current observations as the primary modeling target, with robot actions emerging as the means by which the model navigates from the current observed state to the predicted future state.

The technical approach reportedly leverages large-scale autoregressive video pretraining as the foundation, with the resulting motion priors enabling the model to learn new tasks efficiently. Industry coverage has reported that new tasks can require as little as ten hours of teleoperation data after pretraining, which would compare favorably with peer robotics-AI foundation-model data-efficiency claims. Rhoda has reported that in a high-volume manufacturing evaluation, FutureVision completed a component-processing workflow in under two minutes per cycle without human intervention, exceeding the customer's key-performance-indicator targets.

The March 2026 stealth-exit and Series A announcement positioned Rhoda AI as one of the principal new robotics-AI foundation-model insurgents to emerge in the post-Physical-Intelligence-and-Skild cohort, with the $1.7 billion post-money valuation establishing the company in the top tier of robotics-AI insurgents alongside the earlier-cohort peers.

Mission and strategy

Rhoda AI's stated mission is to bring robots out of the laboratory and into real-world environments by building a robot-intelligence foundation model that operates beyond controlled demonstrations. The strategic premise is that the gap between laboratory robot demos and production-quality robot behavior in real-world environments has been the principal blocker for the robotics industry, and that a video-predictive-control foundation model trained on internet-scale video data has structural advantages in handling the long tail of real-world conditions that controlled-demonstration approaches cannot capture.

The strategy combines two threads. First, the FutureVision foundation-model layer, with the closed-feedback-loop video-prediction-and-action runtime as the technical premise. Second, hardware-partner licensing, with FutureVision positioned to be licensed to partners across diverse robotic hardware platforms over time, while initially powering Rhoda's own robotic systems.

The competitive premise positions Rhoda AI within the robotics-AI foundation-model insurgent cohort alongside Physical Intelligence, Skild AI, and Genesis AI, with the video-predictive-control technical bet providing distinctive positioning relative to vision-language-model-and-action-head peer architectures.

Models and products

• FutureVision. Video-predictive-control robot-intelligence foundation model. Demonstrated across humanoid and industrial-robot platforms. Closed-feedback-loop runtime updating actions every few hundred milliseconds.
• Rhoda robotic systems. Initial deployment platform powered by FutureVision; specifics on hardware partnerships and product variants have been comparatively under-disclosed.
• Hardware-partner licensing. Stated forward strategy of licensing FutureVision to partners across diverse robotic hardware platforms.

Distribution channels have been disclosed as direct industrial-customer relationships, with the high-volume manufacturing evaluation and the component-processing workflow demonstration as the principal public references.

Benchmarks and standing

Rhoda AI's evaluation framework focuses on production-quality robot-task completion metrics rather than horizontal foundation-model leaderboards. The principal disclosed benchmark to date is the high-volume manufacturing evaluation, in which FutureVision completed a component-processing workflow in under two minutes per cycle without human intervention, exceeding the customer's key-performance-indicator targets.

Industry coverage has consistently grouped Rhoda AI with the principal robotics-AI foundation-model insurgent cohort alongside Physical Intelligence and Skild AI, with the $1.7 billion post-money valuation putting Rhoda in the same valuation tier as the earlier-cohort peers despite the comparatively short public operating history. The video-predictive-control technical bet has been characterized as a distinctive technical positioning relative to peer architectures, though the production-deployment evidence base remains narrow.

Leadership

As of stealth-exit reporting in March 2026:

• Founders: Not publicly disclosed in the stealth-exit announcement; the company has been comparatively quiet on senior-leadership identification.
• Senior research and engineering leadership across the FutureVision program; specific named individuals have been comparatively under-disclosed.

The senior-leadership disclosure pattern is unusual among well-funded robotics-AI insurgents but consistent with the company's commercial-customer-first stealth-period posture. Subsequent public disclosures may identify the founder team and senior leadership as the company moves further from stealth-exit posture into normal commercial operating cadence.

Funding and backers

• Series A (March 2026): $450 million at $1.7 billion post-money valuation. Led by Khosla Ventures with Temasek, Mayfield, Capricorn Investment Group, Premji Invest, and John Doerr participating.

Cumulative disclosed private capital is the $450 million Series A. Earlier seed and bridge financing has not been publicly disclosed, and prior capital is presumed but not confirmed.

The Khosla Ventures lead is structurally distinctive: Vinod Khosla has been a longstanding senior investor in robotics-AI ventures, and the Khosla portfolio has historically taken concentrated positions in robotics-AI insurgents at the Series A stage. Temasek, the Singapore sovereign-investment vehicle, brings sovereign-backed institutional-investor visibility. Mayfield and Capricorn Investment Group bring senior US venture-capital relationships. Premji Invest, the family office of the Wipro founder Azim Premji, brings strategic enterprise-and-Indian-industrial relationships. John Doerr, the senior Kleiner Perkins partner, brings additional senior-US-venture-capital visibility. The investor cohort is comparable in seniority to the cohorts that backed Physical Intelligence and Skild AI in their respective Series A rounds, though the round size of $450 million is larger than either Physical Intelligence's $400 million Series A or Skild AI's $300 million Series A.

Industry position

Rhoda AI occupies a distinctive position among robotics-AI foundation-model insurgents: the video-predictive-control technical bet, the $450 million Series A round size, and the Khosla-Temasek-Mayfield-Capricorn-Premji-Doerr investor cohort combine into one of the highest-profile stealth-exits in the post-Physical-Intelligence-and-Skild cohort. Industry coverage has consistently characterized the company as one of the structurally consequential robotics-AI insurgents of the 2024 to 2026 cohort.

The structural risks are characteristic of foundation-model robotics insurgents. First, the video-predictive-control technical bet is unproven at broad commercial deployment scale: the manufacturing evaluation provides one production data point, but whether the approach generalizes across the diverse robot-task surface that hardware partners would license FutureVision for is the principal commercial question. Second, the founder-team and senior-leadership opacity is unusual among well-funded peers and may complicate research-talent recruiting against Physical Intelligence (with the Hausman-Levine-Finn-Groom founder cohort) and Skild AI (with the Pathak-Gupta CMU founder pair).

Competitive landscape

• Physical Intelligence, Skild AI. Direct robotics-AI foundation-model peers from the 2023 to 2024 insurgent cohort.
• Genesis AI. 2024-cohort robotics-AI foundation-model peer with full-stack hardware-and-AI approach.
• 1X, Figure AI, Apptronik. Humanoid-robotics hardware peers and potential FutureVision-licensing partners.
• Boston Dynamics, Tesla Optimus. Vertically integrated humanoid-robotics manufacturers with in-house AI; competitive dynamics depend on whether external AI licensing displaces in-house development.
• Google DeepMind (RT-2, Gemini Robotics), NVIDIA Research (GR00T). Frontier-lab robotics-AI research programs.

Outlook

• The pace of disclosed hardware-partner licensing relationships beyond Rhoda's own robotic systems.
• Continued FutureVision technical-report and capability disclosures as the company moves further from stealth-exit posture.
• The disclosure of founder team and senior leadership identification.
• The Series B or adjacent fundraising timeline given the comparatively recent Series A.
• Whether the video-predictive-control technical bet produces measurable advantages over vision-language-model-and-action-head peer architectures in commercial deployments.

Sources

• Business Wire: Rhoda AI Exits Stealth with $450 Million Series A. Stealth-exit and Series A reference.
• The Robot Report: Rhoda AI exits stealth with $450M to train robots from video. Industry coverage.
• Tech Funding News: Khosla-backed Rhoda raises $450M at $1.7B valuation. Investor and valuation detail.
• Interesting Engineering: Robot AI trained on millions of videos. FutureVision technical-approach reference.
• Rhoda AI official site. Company reference.