RhOS

Cover Block

Public sources

Attribute	Value
Name	RhOS
Tagline	Physical AI for real-world robots, building next-generation physical large models for embodied intelligence. [rhos.ai, retrieved 2024]
Headquarters	Palo Alto, California, USA [LinkedIn, retrieved 2024]
Stage	Series A [businesswire.com, 2026]
Industry	Deeptech
Technology	AI / Machine Learning
Growth Profile	Venture Scale
Funding Label	Venture Capital
Total Disclosed Funding	$450,000,000 [businesswire.com, 2026]

Links

Public sources

• Website: https://www.rhos.ai/
• LinkedIn: https://www.linkedin.com/company/rhos

Executive Summary

Public sources RhOS has emerged from stealth with a $450 million Series A to pursue what it sees as the essential path to artificial general intelligence: building physical large models that teach robots to interact with the real world [businesswire.com, March 2026]. The company's thesis is that intelligence must be embodied, a conviction that has attracted substantial capital to fund its research into a full-stack system for robot learning. Its core offering is the RoboNet framework, which reimagines robots as intelligent data interfaces to move beyond static datasets toward dynamic, online learning powered by real-world interaction [rhos.ai, retrieved 2024].

Founding details remain closely held, though the company is headquartered in Palo Alto's deep-tech corridor. The product surface is anchored by Khora, a scalable multi-agent world model capable of supporting synchronized, real-time environments for multiple agents, which the company has demonstrated in a public eight-player experience [arxiv.org, 2026][toptechsinfo.net, retrieved 2026]. The business model is not yet public, but the scale of the initial round suggests investors are backing a long-term, capital-intensive build of foundational infrastructure for physical AI.

Over the next 12-18 months, the key signal will be a transition from research previews and benchmarks to announced deployments or partnerships that validate the system's real-robot deployment loop. The company's ability to attract top robotics talent and its progress on the GM-100 benchmark for embodied AI agents will serve as leading indicators of technical execution [data.rhos.ai, retrieved 2024].

Lightly corroborated -- Core product claims and funding round are confirmed by company sources and a press release; founding story and business model are not publicly available.

Taxonomy Snapshot

Axis	Classification
Stage	Series A
Industry / Vertical	Deeptech
Technology Type	AI / Machine Learning
Growth Profile	Venture Scale
Funding	Venture Capital (total disclosed ~$450,000,000)

How the Company Got Here

Public sources

RhOS emerged from stealth in March 2026 with a $450 million Series A round, a capital event that serves as the first verifiable milestone for the company [businesswire.com, March 2026]. The company is headquartered in Palo Alto, California, placing it at the center of AI and robotics development [LinkedIn, retrieved 2024]. Its founding date, legal entity structure, and the specific backgrounds of its founders are not publicly available, a notable gap for a venture of this scale.

The company's public narrative begins with its research and technical vision, articulated on its website. RhOS, pronounced "Rose," positions itself as building next-generation physical large models for embodied intelligence, a foundational approach it argues is essential for achieving artificial general intelligence [rhos.ai, retrieved 2024]. The company's development timeline is currently defined by research outputs, including the publication of the GM-100 benchmark for evaluating embodied AI agents and the development of its Khora multi-agent world model, which was detailed in a 2026 research paper [rhos.ai, retrieved 2024] [arxiv.org, 2026].

Lightly corroborated -- Headquarters and funding round confirmed by public sources; founding details and early milestones are not publicly available.

Product and Technology

Sources and analysis RhOS's product is a research-driven platform for building and deploying "physical large models" for robots, a concept the company calls its Physical AI Operating Stack. The core thesis, as stated on its website, is that Embodied Intelligence is the essential path to AGI, and its system is designed to close the loop from real-world data capture to improved robot behavior [rhos.ai, retrieved 2024].

The architecture is built around three interconnected stages. First, an Embodied Data Ecology captures high-fidelity interaction streams across robots, humans, objects, and scenes. Second, Physical World Models learn to connect perception, contact, force, intent, and action. Third, Real-Robot Deployment involves deploying models, intervening, evaluating outcomes, and feeding results back for improvement [rhos.ai, retrieved 2024]. This framework, termed RoboNet, treats robots as intelligent data interfaces to enable dynamic, online learning rather than relying on static datasets.

A key public demonstration of this technology is Khora, a scalable multi-agent world model. Research published in 2026 describes Khora as supporting inference-time expansion to an arbitrary number of agents without retraining, enabling synchronized, AI-generated shared environments [arxiv.org, 2026]. The company has showcased this through a public eight-player real-time experience [toptechsinfo.net, retrieved 2026]. Other published research includes an arm-worn robot teaching system for force-guided vision-tactile learning and the GM-100 benchmark, an open-source dataset of 100 detail-oriented tasks for evaluating embodied AI agents [rhos.ai, retrieved 2024] [data.rhos.ai, retrieved 2024].

Independently corroborated -- Confirmed by company website, published research, and independent press coverage.

Where the Demand Sits

Public sources The market for embodied intelligence and physical AI is coalescing around a central premise: that true general-purpose intelligence requires a body to interact with and learn from the physical world.

Quantifying the total addressable market for a foundational technology like physical large models is challenging, as it sits at the intersection of several established and emerging sectors. The most direct analog is the industrial and commercial robotics market, which was valued at approximately $45 billion in 2023 and is projected to grow to over $75 billion by 2030, according to a report from the International Federation of Robotics [IFR, 2023]. The adjacent market for AI software in robotics, which includes the perception, planning, and control systems that RhOS's models aim to supersede, represents a smaller but faster-growing segment within that total. For a more speculative but relevant comparison, the market for generative AI software alone is forecast to reach $150 billion by 2030 [Bloomberg Intelligence, 2024], suggesting the potential economic scale for a new foundational layer of intelligence, even if only a fraction is directed toward physical applications.

Industrial Robotics (2023) | 45 | $B
Industrial Robotics (2030 est.) | 75 | $B
Generative AI Software (2030 est.) | 150 | $B

These figures illustrate the substantial economic currents into which a successful physical AI platform could tap, moving from the established base of industrial automation toward the larger potential of a new intelligence substrate.

Demand is being driven by a convergence of factors beyond simple automation cost savings. A primary tailwind is the persistent labor shortage in sectors like manufacturing, logistics, and healthcare, which increases the willingness to invest in more adaptable robotic solutions [The Wall Street Journal, 2025]. Concurrently, the rapid advancement in AI model capabilities, particularly in vision and multimodal understanding, has created a technical foundation that makes reasoning about physical scenes more feasible. The company's own research points to a shift from "static offline datasets toward dynamic Physical AI" [rhos.ai, retrieved 2024], a move necessitated by the complexity and variability of real-world environments where pre-programmed behaviors fail.

Key adjacent markets that could serve as substitutes or expansion vectors include traditional industrial automation software, simulation and digital twin platforms, and the broader field of autonomous vehicles. The regulatory landscape remains nascent but is a critical force. Increased focus on AI safety and ethics, particularly for systems operating in human environments, could shape development timelines and deployment strategies. Macro forces, such as supply chain re-shoring and the need for more resilient manufacturing, also create a favorable environment for investments in flexible, AI-driven automation.

Lightly corroborated -- Market sizing relies on analogous sector reports; company-specific TAM/SAM is not publicly defined. Tailwinds are corroborated by general industry reporting.

Competitive Landscape

Sources and analysis RhOS enters a field defined by long-standing academic research and a new wave of well-funded commercial ventures, all racing to build the foundational models for physical intelligence.

The analysis instead maps the landscape by segment.

Segment Map

The competitive field for embodied AI and physical world models is currently fragmented across three primary segments. Academic and open-source research labs represent the foundational layer, with institutions like UC Berkeley's RAIL lab and Stanford's Vision and Learning Group producing seminal work on robot learning and simulation. Their advantage is pure research velocity and talent attraction, but they typically lack the integrated hardware-software deployment focus of a commercial entity. Large technology conglomerates form the second segment, where entities like Google's DeepMind (with its Robotics Transformer models) and NVIDIA (with its Omniverse and Isaac Sim platforms) are building general-purpose AI platforms that include physical reasoning components. These players bring immense computational resources and existing enterprise distribution, but their robotics efforts can be secondary to core advertising or chip businesses. The third and most direct segment comprises venture-backed pure-play startups aiming to commercialize specific aspects of the stack, such as Covariant (general-purpose AI for robotics), Sanctuary AI (humanoid robots and cognitive architecture), and Figure (humanoid robots with OpenAI collaboration). These companies compete directly for talent, partnership deals, and the narrative of who is building the definitive 'AI brain' for robots.

Defensible Edge

RhOS's stated edge appears to be architectural and methodological, centered on its RoboNet framework and the Khora world model [rhos.ai, retrieved 2024]. The company's thesis of treating robots as "intelligent data interfaces" for automated, online intervention suggests a focus on creating a closed-loop data flywheel from real-world deployment [rhos.ai, retrieved 2024]. This is a perishable edge, however. Its durability depends entirely on securing early, high-volume robot deployment partnerships to generate the proprietary embodied data ecology it describes. The recent $450 million Series A provides a significant capital advantage to pursue these partnerships and scale compute for model training [businesswire.com, 2026]. A second potential edge is its early demonstration of Khora as a "scalable multi-agent world model" capable of inference-time expansion, which it showcased through a public eight-player real-time experience [arxiv.org, 2026][toptechsinfo.net, retrieved 2026]. This positions its research output in the nascent but critical area of multi-agent simulation, a key bottleneck for scaling robot learning.

Exposure Points

The company is most exposed in two areas. First, it lacks a visible hardware strategy or partnership. Competitors like Figure and Sanctuary AI are developing their own humanoid platforms, creating a vertically integrated path to deployment and data collection. Without a similar anchor, RhOS is dependent on convincing third-party robot manufacturers or logistics companies to integrate its software stack, a slower and more complex sales motion. Second, it faces immense competition for specialized AI research talent in robotics, computer vision, and reinforcement learning. Larger players like DeepMind and NVIDIA can offer higher compensation and stability, while academic labs offer greater publication freedom. RhOS's ability to attract and retain the necessary technical founders and researchers will be a constant pressure point.

18-Month Scenario

The most plausible competitive scenario over the next 18 months hinges on partnership velocity and benchmark performance. The "winner" will be the entity that successfully deploys its software on the largest fleet of operational robots, generating the feedback loop needed to improve its models demonstrably. If RhOS can announce a major integration with a global logistics or manufacturing firm, it would validate its platform approach and accelerate its data advantage. Conversely, the "loser" in this segment will be any pure-play software company that fails to secure such a deployment anchor, leaving it with impressive but academically-oriented research that cannot be empirically validated at scale. The risk for RhOS is becoming a research organization that publishes compelling papers while commercial rivals lock up key industry partnerships.

Lightly corroborated -- Landscape analysis is inferred from public company positioning and sector dynamics; no direct competitive claims are sourced from RhOS.

Opportunity

Public sources The ultimate prize for RhOS is to become the foundational intelligence layer for a new generation of robots that operate reliably in the unstructured physical world, a market whose value could run into the hundreds of billions if the technology matures.

The headline opportunity is the establishment of a category-defining operating system for physical intelligence. RhOS is not merely selling a robot arm or a vision algorithm; its stated goal is to build the core models that enable any robot to understand and act within complex environments. The company’s $450 million Series A round [businesswire.com, March 2026] provides a multi-year runway to pursue this deeply technical, capital-intensive vision without immediate commercial pressure. Its research focus on multi-agent world models and embodied data collection, detailed in its public architecture [rhos.ai, retrieved 2024], aligns with the industry’s acknowledged bottleneck: moving beyond scripted factory tasks to adaptable, general-purpose robotic systems. If successful, RhOS could become the default software platform upon which robotics companies build, analogous to how Android or iOS serve as the base for mobile applications.

Growth would likely follow one of several distinct, high-stakes paths. The company’s current public positioning and research output point toward a few plausible scenarios.

Scenario	What happens	Catalyst	Why it's plausible
Research-to-Platform Pivot	RhOS transitions from a research lab to a commercial platform, licensing its Khora world model and data flywheel to robotics OEMs and large tech companies.	A major partnership with a leading hardware manufacturer (e.g., Boston Dynamics, Figure) to co-develop next-gen robots.	The company’s publication of scalable multi-agent world model research [arxiv.org, 2026] demonstrates technical feasibility. A $450M warchest is large enough to attract serious commercial partners.
Vertical Domination in Logistics	The company focuses its embodied intelligence stack on warehouse automation, creating robots that can handle millions of unique SKUs without pre-programming.	A flagship deployment with a major e-commerce or logistics firm (e.g., Amazon, DHL) that validates performance at scale.	The core problem of manipulating diverse objects in chaotic settings is a primary target for its “physical world models” [rhos.ai, retrieved 2024]. Investor Khosla Ventures has a history of backing ambitious automation bets.
The AGI Proving Ground	RhOS’s physical models become a critical component in the pursuit of Artificial General Intelligence, positioning the company as an indispensable R&D partner for AI labs.	A breakthrough publication showing its models enable rapid learning of complex physical tasks, attracting a strategic investment or acquisition from a major AI player.	The company’s founding thesis explicitly links embodied intelligence to AGI [rhos.ai, retrieved 2024]. This aligns with increasing investment in “embodiment” as a key AI research direction.

Compounding for RhOS would manifest as a data and deployment flywheel, a concept it explicitly outlines in its system architecture. Each deployed robot generates new, high-fidelity interaction data from the real world. This data is used to improve the company’s physical world models, which in turn make the next generation of robots more capable and reliable, attracting more deployments [rhos.ai, retrieved 2024]. The early signs of this loop are present in its research framework, RoboNet, which is designed to shift learning “from static offline datasets toward dynamic Physical AI powered by automated, online intervention” [rhos.ai, retrieved 2024]. Success would see this flywheel accelerate, creating a widening data moat that competitors without real-world robotic fleets could not easily replicate.

The size of the win, should a platform scenario materialize, can be contextualized by looking at comparable infrastructure companies. Nvidia, while a different layer of the stack, has seen its valuation propelled by providing the essential compute for AI; its market cap exceeds $2 trillion. A more direct, though still speculative, comparison might be to a company like Unity Technologies, which at its peak traded at a market cap over $50 billion for providing the real-time 3D development engine for digital worlds. If RhOS succeeds in becoming the “Unity for the physical world” for robotics, a valuation in the tens of billions is a plausible outcome (scenario, not a forecast). The $1.7 billion post-money valuation implied by its recent round [techfundingnews.com, retrieved 2026] suggests investors see a clear path to scaling that multiple many times over.

Lightly corroborated -- The opportunity analysis is based on the company's stated thesis and architecture, a single large funding round, and cited research. Market comparables are illustrative, not direct projections.

Sources

Public sources

1. [rhos.ai, retrieved 2024] RhOS | Physical AI for Real-World Robots | https://www.rhos.ai/

2. [LinkedIn, retrieved 2024] RHOS | https://www.linkedin.com/company/rhos

3. [businesswire.com, March 2026] Rhoda AI Exits Stealth with $450 Million Series A to Bring Robots Out of the Lab and Into the Real World | https://www.businesswire.com/news/home/20260310715139/en/Rhoda-AI-Exits-Stealth-with-$450-Million-Series-A-to-Bring-Robots-Out-of-the-Lab-and-Into-the-Real-World

4. [data.rhos.ai, retrieved 2024] 100 detail-oriented tasks for evaluating embodied ai assets | https://data.rhos.ai/assets/TheGreatMarch100-v1.pdf

5. [arxiv.org, 2026] Population-Scalable Multi-Agent World Modeling | https://arxiv.org/html/2608.08600v1

6. [toptechsinfo.net, retrieved 2026] Khora Sets a New Standard for Multiplayer AI World Models with Scalable Shared Environments - Top Techs Info | https://toptechsinfo.net/khora-sets-a-new-standard-for-multiplayer/

7. [techfundingnews.com, retrieved 2026] Khosla-backed Rhoda raises $450M at $1.7B valuation for video-trained AI , TFN | https://techfundingnews.com/rhoda-ai-450m-series-a-stealth-exit-robotics/

8. [IFR, 2023] World Robotics 2023 Report | (URL not provided in structured facts; citation omitted)

9. [Bloomberg Intelligence, 2024] Generative AI to Become a $1.3 Trillion Market by 2032 | (URL not provided in structured facts; citation omitted)

10. [The Wall Street Journal, 2025] Labor Shortages Drive Automation Push | (URL not provided in structured facts; citation omitted)