Physical AI

World Foundation Models Physics-Based Models

Snapshot

Physical AI is the emerging category of AI systems that sense, reason about, and act in the physical world through robots, machines, and embodied devices. The category is coalescing around the convergence of foundation models, robotics, simulation, edge computing, and real-world telemetry.
ℹ️
“Physical AI is a $40B+ market by 2029.” — IDC, based on its 2025 robotics and operations research. [3kmmxx]
This profile captures the category as of October 2026 , when “physical AI,” “embodied AI,” and “robot foundation models” are increasingly being used to describe overlapping markets. The category is worth tracking because major technology companies, robotics manufacturers, and heavily funded startups are converging on a common thesis: general-purpose AI becomes substantially more valuable when it can perceive and manipulate physical environments. [3kmmxx]

What is this Market Category?

Physical AI comprises software, models, infrastructure, and machines that integrate artificial intelligence into physical systems so they can perceive environments, interpret tasks, make decisions, and execute actions in the real world. [3kmmxx] Products are sold primarily to manufacturers, logistics operators, warehouses, healthcare providers, retailers, laboratories, and—in some cases—consumers. The category includes humanoid robots, collaborative robots, autonomous mobile robots, delivery and cleaning robots, robot-control software, simulation environments, edge-compute stacks, perception systems, and general-purpose “robot brains.” [3kmmxx]
The category solves problems that conventional automation handles poorly: unstructured environments, variable objects and tasks, natural-language instructions, long-tail edge cases, and workflows that change faster than fixed industrial programming can accommodate. Its central commercial promise is to make physical labor more flexible, general-purpose, and economically deployable. IDC describes the category as spanning platforms and software, infrastructure, services, and devices , rather than only robot hardware. [3kmmxx]
Physical AI excludes ordinary software agents that have no physical sensing or actuation, conventional industrial robots executing tightly scripted movements, and generic computer-vision systems that only observe without controlling a physical process. It also excludes most autonomous vehicles and drones if they are treated as separate vertical markets, although their underlying perception, planning, simulation, and control technologies are materially adjacent.
The boundary is disputed: some analysts use “physical AI” broadly for any AI embedded in machines, while others reserve it for systems with embodied, adaptive, and increasingly general-purpose behavior; IDC’s broader framing includes humanoids, cobots, autonomous mobile robots, and delivery and cleaning robots. [3kmmxx]

Why Now?

  • Foundation models are moving from language and vision into action. Physical AI systems increasingly combine visual perception, reasoning, and control rather than relying exclusively on hand-coded task programs; IDC defines embodied-AI robots as systems capable of autonomous perception, learning, decision-making, and action in physical environments. [n2ec18]
  • The economics of robotics software are expanding beyond single-purpose automation. Transformer- and diffusion-based models are being applied to robotic perception and action, supporting a software market that MarketIntelo estimates reached $4.5 billion in 2025 and is growing at 38.0% CAGR. [23t2po]
  • Edge computing and improved visual systems are lowering deployment friction. IDC says breakthrough advances in AI models, visual systems, and edge computing could triple the number of robot applications, with manufacturing, logistics, healthcare, and services as major deployment areas. [n2ec18]
  • Large technology companies are entering the embodiment race. Meta established a division within Reality Labs to develop AI-powered humanoid robots and physical-world AI, joining Nvidia-backed Figure AI and Tesla in the race to commercialize humanoid systems. [2qznwy]
  • Labor shortages and repetitive or hazardous work create a clear initial wedge. Apptronik is targeting warehouse and manufacturing tasks with Apollo, while Figure and other humanoid developers are positioning robots for logistics, manufacturing, and household work. [j79yha] [f3x34y]
  • Capital markets have established credible funding precedents. Figure’s 2025 financing valued the company at $39 billion, while Physical Intelligence reportedly raised $600 million at a $5.6 billion valuation and Apptronik raised $350 million to scale production. [f3x34y] [7cmsdl] [j79yha]

What's Happening?

CAGR and TAM

  • IDC, “Physical AI & Robotics Research,” 2026: IDC frames Physical AI as a $40 billion-plus market by 2029 , drawing on its Worldwide Robotics Forecast and Future of Operations Survey; its methodology treats the category as a convergent market spanning software, infrastructure, services, and devices rather than a narrow robot-hardware segment. [3kmmxx]
  • Mordor Intelligence, “Physical AI Market Size, Share & 2031 Growth Trends Report,” 2026: Mordor estimates the market at $5.06 billion in 2025 , rising to $34.89 billion by 2031 , for a 37.46% CAGR from 2026 to 2031 ; its definition covers AI integrated into physical devices, machines, and environments that perceive, analyze, decide, and act. [dvcx4q]
  • MarketsandMarkets, “Global Physical AI Market Forecast to 2032,” 2026: MarketsandMarkets estimates a much narrower market of $0.89 billion in 2025 , reaching $15.28 billion by 2032 at a 47.2% CAGR from 2026 to 2032 . [gqx4r4] The large difference from IDC and Mordor is primarily a scope dispute: the lower estimate likely captures a narrower software or directly attributable Physical-AI layer, while IDC and broader market studies include devices, infrastructure, services, and established robotics deployments. [3kmmxx] [dvcx4q] [gqx4r4]
  • Fortune Business Insights, “Physical AI Market Size & Forecast Report, 2034,” 2026: Fortune Business Insights estimates $33.42 billion in 2025 , growing to $294.07 billion by 2034 at a 27.5% CAGR ; it reports robotics as the largest segment in 2025, with a 46.9% share. [2fg0mt]

Category creation events

  • IDC formally positioned Physical AI as a distinct market category. IDC’s 2026 research page describes Physical AI as AI embedded directly into robots and machines so they can sense surroundings, decide, and act, and organizes the market across platforms, infrastructure, services, and devices. [3kmmxx]
  • Meta’s entry validated the category beyond traditional robotics. Reuters reported that Meta was establishing an AI-humanoid-robotics division within Reality Labs, with plans involving robot hardware as well as AI, sensors, and software that could power robots made by other companies. [2qznwy]
  • Figure’s $39 billion financing crystallized humanoid robotics as a frontier-AI market. Reuters reported that Figure’s Series C was led by Parkway Venture Capital and included Nvidia, Intel Capital, LG Technology Ventures, Salesforce, T-Mobile Ventures, and Qualcomm Ventures. [f3x34y]
  • Physical Intelligence helped define the “general-purpose robot brain” thesis. Bloomberg reported that the startup raised $600 million at a $5.6 billion valuation for software intended to help robots learn a wide range of tasks. [7cmsdl]

Capital concentration

  • Humanoid platforms are attracting the largest disclosed rounds. Figure’s 2025 Series C exceeded $1 billion and valued the company at $39 billion, with participation from strategic technology and industrial investors. [f3x34y]
  • Manufacturing and logistics are attracting scale-up capital. Apptronik raised $350 million led by B Capital and Capital Factory, with participation from Google, to scale Apollo humanoid robots for warehouses and manufacturing plants. [j79yha]
  • Robot-foundation-model software is becoming a separate capital pool. Physical Intelligence reportedly raised $600 million at a $5.6 billion valuation, while Skild AI was reported to have raised $500 million at a $4.7 billion valuation and was subsequently in discussions for a potential investment exceeding $1 billion at a $14 billion valuation. [7cmsdl] [8sappm]
  • Strategic investors are concentrating around the infrastructure stack. Nvidia appears repeatedly across Figure, Physical Intelligence-related financing, and Skild AI reporting, while Google invested in Apptronik and Meta is developing its own physical-AI capabilities. [f3x34y] [j79yha] [7cmsdl] [8sappm]

Market Incumbents

  • Nvidia — supplies accelerated computing, simulation, robotics software, and developer infrastructure for physical-AI workloads and is a strategic investor in major robotics startups. [f3x34y] [8sappm]
  • Alphabet / Google — combines robotics research, AI models, cloud infrastructure, and strategic investment in companies such as Apptronik. [j79yha]
  • Meta — is building physical-world AI capabilities and a humanoid-robotics effort through Reality Labs. [2qznwy]
  • Tesla — develops humanoid robotics through Optimus and is identified as a major participant in the race to commercialize AI-powered robots. [2qznwy] [f3x34y]
  • Amazon — represents the logistics and warehouse-automation incumbent buyer and operator for autonomous mobile robots, fulfillment systems, and increasingly adaptive robotic workflows.
  • Microsoft — provides cloud, AI, simulation, and enterprise software infrastructure that can support physical-AI deployments, even where the physical machine is supplied by another vendor.
  • Siemens — sits at the industrial-automation layer, connecting manufacturing control, digital twins, industrial software, and machine operations.
  • ABB — ABB — supplies industrial robots, collaborative robots, automation systems, and installed-base relationships that make it a major incumbent route to enterprise physical-AI adoption.

Nvidia

Stage : public (NASDAQ: NVDA) Funding : Public company; market capitalization and revenue should be taken from the latest 2026 quarterly filing rather than private funding disclosures. Nvidia is a strategic participant in Figure and Skild AI financing. [f3x34y] [8sappm] Footprint : Nvidia’s relevant footprint is its installed base of accelerated-computing hardware and software used to train and deploy AI models, plus its robotics ecosystem and strategic investments; the supplied research identifies Nvidia as a recurring participant in the physical-AI financing stack. [f3x34y] [8sappm] Why they're in this category : Nvidia supplies the compute, simulation, and software substrate on which robot perception, planning, training, and inference are built, while also taking strategic positions in robot-foundation-model and humanoid companies. [f3x34y] [8sappm] Coverage : Reuters, “Robotics startup Figure valued at $39 billion in latest funding round” . [f3x34y] 3x34y] TechCrunch, “SoftBank and Nvidia reportedly in talks to fund SkildAI at $14B” . [8sappm]

Alphabet / Google

Stage : public (NASDAQ: GOOGL) Funding : Public company; Apptronik ’s $350 million financing included participation from Alphabet’s Google. [j79yha] Footprint : Google brings global AI-model, cloud, research, and enterprise distribution capabilities; its specific physical-AI footprint includes strategic participation in Apptronik and the broader robotics research ecosystem. [j79yha] Why they're in this category : Google is combining frontier AI and cloud capabilities with strategic exposure to humanoid robotics rather than approaching the market only as a conventional robot manufacturer. [j79yha] Coverage : Reuters, “Apptronik raises $350 million to scale production of humanoid robots” . [j79yha] 79yha]

Tesla

Stage : public (NASDAQ: TSLA) Funding : Public company; its physical-AI position is funded through Tesla’s corporate balance sheet and internal product development. Footprint : Tesla combines a large-scale manufacturing operation, autonomy software, vehicle sensor data, and the Optimus humanoid-robot program; Reuters identifies Tesla as one of the major companies competing in AI-powered humanoid robotics. [2qznwy] [f3x34y] Why they're in this category : Tesla’s angle is to reuse autonomy, perception, simulation, manufacturing, and fleet-learning capabilities from vehicles in a humanoid platform intended for general physical tasks. [2qznwy] [f3x34y] Coverage : Reuters, “Meta plans investments into AI-driven humanoid robots” . [2qznwy]

Market Challengers

  • Figure AI — humanoid-robotics scale-up targeting logistics, manufacturing, and household tasks, valued at $39 billion in its 2025 financing. [f3x34y]
  • Apptronik — develops Apollo humanoid robots for warehouses and manufacturing and raised $350 million to scale production. [j79yha]
  • 1X — humanoid and wheeled-robot developer pursuing physical-AI systems for work and home environments; it was reported to be discussing financing of up to $1 billion at a $10 billion valuation. [8sappm]
  • Agility Robotics — develops Digit humanoid robots for logistics and warehouse operations and competes for enterprise deployments in structured commercial environments.
  • Physical Intelligence — builds general-purpose robot-learning software intended to enable a wide range of tasks across different robot embodiments and reportedly raised $600 million at a $5.6 billion valuation. [7cmsdl]
  • Skild AI — develops a foundational robotics model and raised $500 million at a $4.7 billion valuation, with later reporting of a possible $1 billion-plus financing at a $14 billion valuation. [8sappm]
  • Covariant — applies foundation-model techniques to warehouse manipulation and robotic picking, positioning software as the scalable layer across installed robot fleets.

Figure AI

Stage : late-stage private scale-up Funding : Figure’s 2025 Series C exceeded $1 billion and valued the company at $39 billion; the round was led by Parkway Venture Capital and included Nvidia, Intel Capital, LG Technology Ventures, Salesforce, T-Mobile Ventures, and Qualcomm Ventures. [f3x34y] Footprint : Figure is focused on humanoid robots for logistics, manufacturing, and household work, and its valuation makes it one of the most heavily capitalized independent humanoid developers. [f3x34y] Why they're in this category : Figure is a challenger because it has the capital and public visibility to compete directly with Tesla and major technology companies while using a humanoid form factor to target multiple labor markets. [f3x34y] Coverage : Reuters, “Robotics startup Figure valued at $39 billion in latest funding round” .[5]

Apptronik

Stage : late-stage private scale-up Funding : Apptronik raised $350 million in February 2025 in a round led by B Capital and Capital Factory, with participation from Google. [j79yha] Footprint : Apptronik is developing Apollo for warehouses and manufacturing plants, initially targeting package movement and other supply-chain-oriented tasks. [j79yha] Why they're in this category : Apptronik’s wedge is an enterprise humanoid designed for environments where existing warehouse and manufacturing workflows already create demand for repetitive physical labor. [j79yha] Coverage : Reuters, “Apptronik raises $350 million to scale production of humanoid robots” .[9]

Physical Intelligence

Stage : late-stage private scale-up Funding : Physical Intelligence reportedly raised $600 million at a $5.6 billion valuation , with CapitalG identified in reporting as the lead investor. [7cmsdl] Footprint : The company’s footprint is primarily software and research rather than a disclosed installed hardware fleet; its stated market position is a general-purpose AI system that helps robots learn a wide range of tasks. [7cmsdl] Why they're in this category : Physical Intelligence is a challenger because it is trying to become the model layer across many robot types, avoiding dependence on a single proprietary humanoid embodiment. [7cmsdl] Coverage : Bloomberg, “Robotics Startup Physical Intelligence Valued at $5.6 Billion in New Funding” . [7cmsdl]

Market Innovators

  • Tenyks — early-stage computer-vision and edge-AI company relevant to machine perception and physical-world telemetry.
  • Robust AI — develops intelligent warehouse robots and autonomy software, with a thesis around adaptive rather than purely scripted logistics automation.
  • Standard Bots — early-stage industrial-robotics company focused on making robot programming and deployment accessible to smaller manufacturers.
  • Mujin — develops intelligent robot-control systems for industrial and logistics applications, especially where perception and motion planning must adapt to variable scenes.
  • Veo Robotics — develops vision-based industrial safety and human-robot collaboration systems.
  • K-Scale Labs — founder-led open robotics effort exploring lower-cost humanoid hardware, data collection, and community-driven robot learning.
The supplied search results do not provide sufficiently reliable round dates, totals, lead investors, or employee counts for all of these early-stage companies. Those fields should be verified against company announcements, regulatory filings, or reputable funding databases before investment-grade use.

Robust AI

Stage : early-stage/private; precise financing stage requires primary-source verification. Funding : The supplied search results do not establish a reliable total raised, latest round, lead investor, or date. Footprint : Robust AI ’s relevant footprint is intelligent warehouse robotics and autonomy software; customer count, revenue, employees, and deployment geography require verification from company or financial sources. Why they're in this category : Robust AI belongs in the innovator tier because its strategic thesis is to make warehouse robots adaptive to changing environments rather than dependent on rigid automation scripts. Coverage : Primary company materials and specialist robotics coverage should be consulted before treating its funding or deployments as established facts.

Standard Bots

Stage : early-stage/private; precise financing stage requires primary-source verification. Funding : The supplied search results do not establish a reliable total raised, latest round, lead investor, or date. Footprint : Standard Bots is positioned around accessible industrial robotics and simplified deployment; verified customer, revenue, and geographic data were not returned in the available search results. Why they're in this category : Its novel-bet position is that easier programming and lower deployment complexity can broaden physical-AI adoption beyond the largest manufacturers. Coverage : Company announcements and specialist manufacturing-automation reporting should be used to verify traction and financing.

K-Scale Labs

Stage : early-stage/private; precise financing stage requires primary-source verification. Funding : The supplied search results do not establish a reliable total raised, latest round, lead investor, or date. Footprint : K Scale Labs is associated with open and lower-cost humanoid-robotics experimentation; verified employee count, deployments, and commercial revenue were not returned in the available results. Why they're in this category : K-Scale Labs represents the innovator thesis that open hardware, shared datasets, and lower-cost experimentation can reduce the capital required to participate in humanoid robotics. Coverage : Founder and project documentation should be reviewed before assigning a definitive funding stage or commercial scale.

Industry Coverage and Market Data

Market Reports

  • Physical AI & Robotics Research, 2026 — IDC — defines Physical AI as AI embedded into robots and machines, estimates a $40 billion-plus market by 2029, and frames the category across platforms, software, infrastructure, services, and devices. [3kmmxx]
  • Physical AI Market Size, Share & 2031 Growth Trends Report, 2026 — Mordor Intelligence — estimates $5.06 billion in 2025 and $34.89 billion by 2031 at a 37.46% CAGR; scope includes AI integrated into physical devices, machines, and environments. [dvcx4q]
  • IDC Innovators: Embodied AI Robotics, 2025 — IDC — defines embodied-AI robots as systems with autonomous perception, learning, decision-making, and action, and forecasts the segment exceeding 30% of total robotics-market share by 2030. [n2ec18]
  • Global Physical AI Market Forecast to 2032, 2026 — MarketsandMarkets — uses a narrower estimate of $0.89 billion in 2025 and $15.28 billion by 2032 at a 47.2% CAGR. [gqx4r4]
  • Physical AI Market Size & Forecast Report, 2034 — Fortune Business Insights — estimates $33.42 billion in 2025 and $294.07 billion by 2034 at a 27.5% CAGR, with robotics representing 46.9% of the 2025 market. [2fg0mt]
  • Physical AI and Embodied Intelligence Software Market, 2026 — MarketIntelo — estimates a $4.5 billion 2025 software market and a 38.0% CAGR, attributing growth to transformer- and diffusion-based robotic perception and action. [23t2po]

Industry Articles

  • Meta plans investments into AI-driven humanoid robots, memo shows — Reuters — reports Meta’s entry into humanoid robotics and its interest in AI, sensors, software, and potential robot hardware. [2qznwy]
  • Robotics startup Figure valued at $39 billion in latest funding round — Reuters — covers Figure’s valuation, financing participants, and the wider race among robotics startups and technology companies. [f3x34y]
  • Apptronik raises $350 million to scale production of humanoid robots — Reuters — reports Apptronik’s financing and Apollo’s initial warehouse and manufacturing use cases. [j79yha]
  • Physical AI & Robotics Research — IDC — provides one of the clearest category definitions, including the “sense, decide, and act” framing and the four-layer market structure. [3kmmxx]
  • IDC Innovators: Embodied AI Robotics, 2025 — IDC — explains the shift from conventional robotics toward adaptable, autonomous physical systems. [n2ec18]

Financial News Sources

  • Robotics startup Figure valued at $39 billion in latest funding round — Reuters — reports Figure’s Series C, lead investor, strategic participants, and $39 billion valuation. [f3x34y]
  • Apptronik raises $350 million to scale production of humanoid robots — Reuters — reports the $350 million financing led by B Capital and Capital Factory, with Google participation. [j79yha]
  • Robotics Startup Physical Intelligence Valued at $5.6 Billion in New Funding — Bloomberg — reports Physical Intelligence’s $600 million financing and $5.6 billion valuation. [7cmsdl]
  • SoftBank and Nvidia reportedly in talks to fund SkildAI at $14B — TechCrunch — reports Skild AI’s prior $500 million round at a $4.7 billion valuation and potential later financing above $1 billion at a $14 billion valuation, alongside reported Physical Intelligence and Figure financings. [8sappm]
  • Meta plans investments into AI-driven humanoid robots, memo shows — Reuters — provides financial-market context for Meta’s strategic entry into physical AI and the competitive positioning of Nvidia-backed Figure and Tesla. [2qznwy]

Frontier and Open Questions

  • Will Physical AI become a software-platform market or remain dominated by vertically integrated robot manufacturers? Nvidia, Physical Intelligence, and Skild AI are pursuing reusable infrastructure and model layers, while Tesla, Figure, and Apptronik are pursuing tightly integrated hardware-and-software systems. [2qznwy] [f3x34y] [j79yha] [7cmsdl] [8sappm]
  • Can foundation models generalize across embodiments, tasks, and environments without requiring uneconomic amounts of real-world data? Physical Intelligence and Skild AI are the clearest challengers to single-robot architectures, but the supplied sources do not yet establish durable cross-platform generalization. [7cmsdl] [8sappm]
  • Will humanoids be the dominant form factor, or will specialized mobile and industrial robots capture most early revenue? IDC’s category includes humanoids, cobots, autonomous mobile robots, and delivery and cleaning robots, while Fortune Business Insights reports that robotics broadly—not humanoids alone—held the largest 2025 share. [3kmmxx] [2fg0mt]
  • Does the category include autonomous vehicles, drones, and industrial control systems? A broad definition includes any AI embedded in a machine that senses and acts, while a narrower definition limits Physical AI to general-purpose embodied agents; IDC’s broad four-layer framing favors inclusion, but market reports vary materially in scope. [3kmmxx] [n2ec18] [gqx4r4]
  • Can enterprise customers achieve reliable unit economics before humanoid hardware reaches mass production? Apptronik is targeting warehouses and manufacturing, Figure is targeting logistics, manufacturing, and household tasks, and the outcome will depend on uptime, safety, integration cost, and labor substitution rather than model capability alone. [f3x34y] [j79yha]
  • Will incumbents capture the value through compute, cloud, industrial distribution, and installed bases? Nvidia, Google, Meta, Tesla, Amazon, Siemens, and ABB can distribute physical-AI capabilities through existing platforms, while challengers must prove that their model or robot advantage is strong enough to overcome incumbent channels. [2qznwy] [f3x34y] [j79yha] [8sappm]

Adjacent Concepts and Categories

  • Embodied AI — the broader technical concept covering AI systems that learn and act through physical bodies and environments.
  • Robot Foundation Models — reusable models intended to control multiple robot types and perform many tasks.
  • Humanoid Robotics — the most visible physical-AI form factor and a major destination for current capital.
  • Industrial Automation — the incumbent market where physical AI must prove reliability, safety, and return on investment.
  • Autonomous Mobile Robots — adjacent warehouse, factory, hospital, delivery, and cleaning systems already operating in commercial environments.
  • Simulation-to-Real Transfer — the technique of training or validating robotic behavior in simulated environments before deployment in the physical world.
  • Real-World Telemetry — sensor, interaction, and operational data generated by physical systems and used to improve models.
  • Edge AI — the compute and inference layer that allows robots and machines to respond locally under latency, connectivity, privacy, and safety constraints.

Sources