The Physical AI Revolution

Mapping the $60 Trillion Transformation

While the world has been captivated by the linguistic prowess of generative AI models like ChatGPT, a quieter but far more consequential (and powerful!) revolution has been gaining momentum. This is the era of Physical AI, where artificial intelligence breaks free from the digital realm to interact with and transform our physical world. From humanoid robots assembling cars to autonomous vehicles navigating city streets, Physical AI represents what Morgan Stanley calls a $60 trillion total addressable market — equivalent to almost the entire global GDP?! 👀. [1]

This week’s announcement that Figure AI has raised over $1 billion at a $39 billion valuation — making it one of the most valuable (robotics) startups in history — signals that this revolution has moved from science fiction (and labs) to market reality. I dove into the past to see how it all started and mapped the Physical AI ecosystem, from the companies building the future to the investors betting billions on this transformation …which I’m proudly part of 💪

What is the Physical AI?

Physical AI represents the convergence of artificial intelligence, robotics, and real-world interaction. I like how NVIDIA, a key enabler of this transformation, defines it as technology that:

“Lets autonomous systems like robots, self-driving cars, and smart spaces perceive, understand, and perform complex actions in the real (physical) world. It’s also often referred to as ‘generative physical AI’ because of its ability to generate insights and actions.”

This capability is built on our good old digital twins and reinforcement learning, where AI models train through millions of simulated trial-and-error cycles before operating in the real world. We then come up with the result which are basically machines with “common sense” understanding of physics, spatial relationships, and cause-and-effect that humans take for granted.

The Physical AI market has experienced unprecedented growth, with projections being revised upward dramatically as technological breakthroughs accelerate adoption. Goldman Sachs Research has increased their humanoid robot market projection sixfold to $38 billion by 2035, up from a previous estimate of just $6 billion. This dramatic revision reflects a few main developments:

• Robot shipments: Projected to reach 1.4 million units by 2035 (4x increase)!
• Manufacturing costs: Dropped 40% annually vs. expected 15–20%
• Current price range: $30,000-$150,000 (down from $50,000-$250,000)
• 2030 base case: 250,000+ units shipped (mostly industrial applications)

The cost reduction has been driven by cheaper components, expanded supply chains, and improved manufacturing techniques, accelerating the timeline to factory applications by one year and consumer applications by two to four years. We still don’t know though how the tarrifs impact these projections but I’ll watch it closely and Goldman Sachs surely as well. Looking forward to their next year report!

Morgan Stanley’s “Humanoid 100” report identifies the complete value chain across three categories:

The report notes that investor interest has “accelerated meaningfully” following NVIDIA CEO Jensen Huang’s 2025 CES presentation on Physical AI, with company earnings calls mentioning “humanoid” at unprecedented rates and also Open AI and many other powerhouses joining the race.

The Timeline: From Concept to Billion Dollars Valuations

I’ll get into the companies themselves in the later part of the post but let’s take a few steps back and see where it all started and who was the bravest to even try introducing general, physical intelligence years back.

2014–2022: Foundation Building

• 2014 (yes, it’s been more than 10 years!): 1X Technologies (then Halodi Robotics) is founded in Norway, laying the groundwork for future humanoid robotics development. I suppose, it’s the oldest modern humanoid robotics company. Not counting obviously a short mention of Steel Humanoid Robot from 1930s!☺️

• 2015: Agility Robotics is founded by Jonathon (Jonathan) Hurst, Damion Shelton, and Mikhail Jones.
• 2016: Unitree is founded by Wang Xingxing. Originally the company was focused on quadruped robots (robot-dogs), such as Go1, A1, Go2.
• 2017: Agility releases humanoid robot Digit.
• 2022: Figure AI is founded by Brett Adcock
• 2022: Optimus is developed by Tesla. Optimus is a general-purpose humanoid robot announced earlier but visible progress & prototypes were shown in 2022.

2023: Early Market Signals

• February: Zhiyuan’s AgiBot RAISE A1 is launched targetting both industrial and household tasks.
• March: Agility releases an updated version of Digit.
• May : Figure AI raises $70 million seed round (disclaimer-> portfolio company🙋🏻‍♀️).
• May: Sanctuary AI unveils Phoenix, its 6th-generation general-purpose humanoid robot.
• August: Apptronik officially unveils its humanoid robot Apollo.
• August: Unitree launches H1, a full-size humanoid robot. It set some speed records (walking/running) in tests.
• October : EngineAI Robotics is born in Shenzhen, China.

2024: The Breakthrough Year

• January 2024: 1X Technologies secures $100 million Series B.
• February 2024: Figure AI raises $675 million at $2.6 billion valuation from Jeff Bezos, Microsoft, NVIDIA, OpenAI and more.
• April: Sanctuary AI introduced the 7th generation Phoenix. Raised over $140 million taking an approach focused on human-like cognitive capabilities.
• July: Skilled AI gets $300 million Series A funding at $4.5B valuation. (More focused on the “brains” / foundation model for robotics without building a full humanoid robot at least at launch).
• October: EngineAI unveils a full-size humanoid robot SE01, which can walk, run, jump, has many degrees of freedom and then in late 2024 they released PM01, a smaller general-purpose humanoid robot.
• November: Physical Intelligence emerges with $400 million Series A at $2.4 billion valuation.
• December: Unitree keeps innovating and launches G1, mass-production version of H1. Really cool achievements: 23 degrees of freedom via powered joints, three-fingered hands capable of dexterous manipulation, can walk, climb stairs, jump; speed over ~4.4 mph in ideal conditions.
• December: K-Scale Labs and Benjamin Bolt start operations in Palo Alto, CA introducing interesting approach to have open-source general-purpose humanoid robots.

2025: The Scale-Up Race

• January: 1X Technologies acquires Kind Humanoid.
• February: Apptronik raises $350 million Series A for Apollo robot.
• March: Apptronik adds $53 million, bringing total to $403 million.
• July: Genesis AI launches with $105 million seed from Eclipse + Khosla Ventures.
• July: Unitree launches R1: significantly lighter (~25 kg), ~24–26 DOF. ~39,900 yuan (~US$5,566). Rumors say it’s planning IPO at $7 billion valuation.
• July: Skiled AI has a public Skild Brain demonstration. Multi-robot capability: works on industrial arms, quadrupeds, humanoids.
• September 2025: Figure AI raises $1+ billion at $39 billion valuation.
• September: X Square Robot: most recent ~$100 million funding led by Alibaba.
• September: R1 Robot: Backed by Jack Ma’s Ant Group.
• September: South Korea doesn’t want to stay behind and has launched the K-Humanoid Alliance government initiative to support the market growth.
• October-December: what advancements will close this spectacular year?!

The Physical AI Ecosystem: Market Map

As per Nvidia definition, we can divided the Physical AI into three primary domains:

1. Humanoid (and Industrial) Robots — Machines that work alongside or replace human labor.
2. Autonomous Vehicles — Self-driving cars, trucks, ships, drones and aircraft.
3. Smart Spaces — Intelligent buildings, factories, and environments.

1. Humanoid Robots: The New Workforce

The humanoid robotics sector has emerged as the most visible and well-funded segment of Physical AI, with companies pursuing different strategies from software-first to vertically integrated approaches.

Physical Intelligence (π) leads the software-first approach, developing Vision-Language-Action (VLA) foundation models that can control any robot hardware. Founded by Stanford’s Chelsea Finn and UC Berkeley’s Sergey Levine, the company reached a $2.4 billion valuation in less than a year. Their research timeline shows rapid progress:

• October 2024: π0 — First Generalist Policy
• February 2025: Open-sourced π0 model
• April 2025: π0.5 — VLA with Open-World Generalization
• June 2025: Real-Time Action Chunking with Large Models

Skild AI takes perhaps the most ambitious approach with their $300 million Series A funding at a $4.5 billion valuation, backed by Lightspeed, Coatue, SoftBank, and Jeff Bezos. The Pittsburgh-based company is building what they call an “omni-bodied brain” — a single AI system designed to control any robot for any task. Their Skild Brain represents a unified approach to robotics control, working across industrial arms, quadrupeds, and humanoids, with applications spanning security inspection, mobile manipulation, and autonomous packing. Announced partnership with HPE for AI infrastructure acceleration.

Figure AI is building both hardware and software for its humanoid workforce. The company’s trajectory from $2.6 billion to $39 billion valuation in just seven months represents one of the fastest value creation stories in startup history. Their product evolution includes:

• Figure 01: Initial prototype for logistics and warehousing
• Figure 02: Industrial-ready robot with BMW partnership
• Helix: Next-generation robot with 35 degrees of freedom
• BotQ Factory: Manufacturing facility targeting 12,000 robots annually

Apptronik focuses on the Apollo robot for industrial applications, raising $403 million to scale production. The Austin-based company has earned recognition from Fast Company, CNBC, and Automotive News for its humanoid design.

Agility Robotics commercializes the Digit robot for warehouse automation, partnering with Manhattan Associates to integrate with warehouse management systems. Their Salem factory is building what they call “the robot revolution.”

DYNA Robotics is building general-purpose robots that power the future of the physical economy and represents the newest entrant to achieve major funding, raising $120 million in Series A funding this month at over $600 million valuation. The Redwood City-based company focuses on general-purpose robots powered by their DYNA-1 foundation model, which achieves commercial-grade performance across factories, restaurants, and manufacturing facilities. Unlike simulation-trained models, DYNA-1 is trained directly in real environments, enabling fully autonomous operation with demonstrated capabilities like folding over 1,000 napkins in 24 hours without human supervision.

Dexterity recently jumped ont he Physical AI train and claims it builds Physical AI for enterprise logistics applications — AI that powers robots to perform real-world tasks. Their approach uses teams of Physical AI agents to infuse human-like dexterous skills into any robot, enabling adaptability across virtually any application. Dexterity recently secured $95m, reaching $1.65bn valuation and there are rumors that it’s planning IPO!

Boston Dynamics, now owned by Hyundai, continues advancing the Atlas humanoid robot. Their August 2025 partnership with Toyota Research Institute demonstrates the convergence of automotive and robotics expertise.

To complete the picture, I must mention the main legacy industrial automation leaders that are being threatened by ‘general intelligence’ robots:

• ABB: Still holds ~14% of the global industrial robot market
• KUKA: Has ~13% market share with flexible automation solutions
• FANUC: Excels in high-speed industrial applications

2. Autonomous Vehicles: AI on the Move

The autonomous vehicle segment represents the most mature application of Physical AI, with companies deploying commercial services and achieving significant scale either partially, with a driver still behind the wheel or entirely achieving level 5 of autonomy.

• Urban, B2C mobility: Waymo, Tesla
• Middle mile logistics: Gatik
• Long haul trucking: Waabi, Kodiak, Bot Auto
• Maritime: Saronic, Blue Water Autonomy
• Agricultural: John Deere/GUSS

Urban Mobility: With Cruise out of the picture, we can say Waymo leads with over 250,000 weekly rides, representing “the most advanced application of AI in the physical world.” Their multi-sensor approach combines LiDAR, cameras, radar, and compute to create comprehensive environmental understanding. Tesla also pursues a Full Self-Driving (FSD), recently rolling out version 13.2 with major improvements to end-to-end neural networks. Despite the name, current FSD requires driver supervision.

Middle mile and long haul trucking represents a rapidly growing segment with several specialized players. Gatik leads autonomous middle-mile logistics with over $110 million in funding and strategic backing from NIPPON EXPRESS, operating North America’s first autonomous middle-mile network serving the $250 billion logistics market. Waabi takes an AI-first approach to long-haul trucking, leveraging their Waabi World neural simulator and generative AI-powered testing to pioneer what they call “Physical AI for autonomous trucks.” Bot Auto, based in Houston, recently completed the first fully driverless hub-to-hub truck run and raised $20 million to scale their Transportation-as-a-Service model. Kodiak Robotics focuses on long-haul autonomous trucking with proven commercial operations and a $50 million defense contract, expanding beyond highways into off-road applications (announced plans to go public via merger with Ares Acquisition Corporation II).

Maritime autonomy is emerging as a new frontier, with companies like Saronic developing 150-foot autonomous ships and Blue Water Autonomy creating unmanned vessels for various applications.

Agricultural, Construction, Mining Autonomy is advancing through companies like John Deere, which acquired GUSS to enhance autonomous farming capabilities, or Bedrock Robotics focusing on construction. For mining, I have to mention the fleet of 100 Huaneng Ruichi autonomous electric mining trucks, the first of its kind in the world (!), that has officially entered operation at the Yimin open-pit mine in Inner Mongolia, China. In the US, Komatsu is one of the well known that provide autonomous haulage technology and helps mines introduce autonomy.

3. Smart Spaces: Intelligent Environments

The smart spaces segment encompasses building automation, industrial facilities, and intelligent infrastructure that can perceive and respond to their environment.

Archetype AI leads the trend with their Newton, the first Large Behavior Model (LBM) designed to understand the physical world through sensor data fusion.

The Palo Alto-based company raised $13 million in seed funding led by Venrock, with participation from Amazon Industrial Innovation Fund and Hitachi Ventures. Newton can process data from cameras, microphones, radars, and time-series sensors to reveal hidden patterns in real-time, serving applications from construction safety monitoring to industrial predictive maintenance. (disclaimer-> portfolio company🙋🏻‍♀️).

AI & Robotics Revolution is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

The Investor Ecosystem: Fueling the $60 Trillion Opportunity

The Physical AI revolution has attracted unprecedented investment from venture capital firms, corporate strategics, and high-profile individuals who recognize the transformative potential.

Eclipse Ventures has positioned itself as the leader in “Industrial Evolution” investing, with over $4 billion in assets under management across 70+ portfolio companies. Their co-leadership of Genesis AI’s $105 million seed round demonstrates their commitment to foundational Physical AI infrastructure.

Lux Capital specializes in emerging science and technology, making early bets on Physical Intelligence and other deep tech companies pushing the boundaries of what’s possible.

Sequoia Capital brings traditional venture capital excellence to Physical AI, with their investment in Physical Intelligence signaling mainstream acceptance of the sector.

Khosla Ventures has made significant bets on both Physical Intelligence and Genesis AI, reflecting founder Vinod Khosla’s focus on transformative technologies.

Corporate Strategic Investors

NVIDIA serves as both investor and enabler, providing the GPU infrastructure that powers Physical AI training and inference. Their Cosmos platform has become central infrastructure for robotics development, with companies like 1X Technologies, Wayve, and Neura Robotics building on the platform.

Jeff Bezos has made substantial personal investments in both Physical Intelligence and Figure AI, reflecting his long-standing interest in robotics and automation dating back to Amazon’s warehouse innovations.

Microsoft, Amazon, and Intel have all invested in Figure AI, recognizing how humanoid robots could transform their respective industries from cloud computing to e-commerce and chip manufacturing.

OpenAI invested in Physical Intelligence while simultaneously partnering with Figure AI (though that partnership ended in 2025), highlighting the convergence of large language models and robotics.

Global Market Dynamics: The Three-Horse Race

The development of Physical AI reflects broader geopolitical competition, with the United States, China, and Europe pursuing different strategies for leadership.

It looks Exciting but Challenges & Barriers are Real:

Despite rapid progress, I see significant challenges that need to be overcome before Physical AI achieves its full potential:

➡️ Technical Hurdles

Hardware limitations continue to constrain robot capabilities, though costs are declining faster than expected. High-precision components still require specialized manufacturing equipment with limited global capacity.

Data scarcity remains a bottleneck, as training Physical AI models requires massive amounts of real-world interaction data that is expensive and time-consuming to collect. It’s great to see new startups focusing on real-life data collection for robotics.

Safety and reliability challenges are paramount, especially for applications involving human interaction or safety-critical operations.

➡️ Economic and Social Considerations

Labor displacement concerns require careful management as Physical AI systems become capable of performing an increasing range of human tasks.

Regulatory frameworks are still developing, with different regions taking varying approaches to AI governance and safety standards.

Public acceptance will be crucial for widespread adoption, particularly for consumer applications and human-robot interaction scenarios. It will take us a while to have humanoids taking care of our elderly family members.

Snapshot of the Future Convergence and Acceleration…

Foundation Model Convergence is happening. The success of large language models is being replicated in Physical AI through Vision-Language-Action (VLA) models that can control multiple robot types. Even if I don’t think we will have “ChatGPT moment” for robotics, we are already witnessing it in a slow motion.

Cost Curve Acceleration already took place. Manufacturing costs are declining faster than projected, with Goldman Sachs noting 40% annual reductions versus expected 15–20%. This acceleration could bring consumer robotics forward by years.

Geographic Rebalancing is natural and there will be more and competition because the market is attractive for startups and investors, as well as for the government and country representatives to support and participate in. While the U.S. leads in innovation and China in manufacturing, new players are emerging. South Korea’s $10,000 humanoids and Vietnam’s entry into robotics suggest a more distributed global ecosystem.

Application Expansion: Physical AI is expanding beyond initial use cases in manufacturing and logistics to healthcare, elder care, dangerous work environments, and eventually consumer applications. We can’t even imagine what the robots will be capable of and what we are trying now is just the beginning!

The Physical AI revolution represents more than technological advancement — it’s the foundation of a new economy where intelligent machines seamlessly integrate with human society. The rapid progression from concept to $39 billion valuations demonstrates that this transformation is accelerating beyond most predictions. As Brett Adcock of Figure AI observed in late 2023: “The AI is starting to become more capable than the hardware. This is a big deal. The AI is the bottleneck, not the hardware.” This insight has proven prophetic, as software advances now drive the entire ecosystem forward.

The companies, investors, and regions that successfully navigate this transformation will shape the next century of human progress and I’m glad we are leading in the US. With Morgan Stanley projecting a $60 trillion total addressable market and Goldman Sachs forecasting $38 billion in humanoid robots alone by 2035, Physical AI represents the largest technological opportunity of our generation.

The revolution is no longer coming — it’s here. The question is not whether Physical AI will transform our world, but how quickly and who will lead the transformation and what value for the people we will realize in the long term.

Selected References and Reports worth checking:

[1] The Humanoid 100: Mapping the Humanoid Robot Value Chain — Morgan Stanley Research

[2] Robotics Startup Figure AI Valued at $39 Billion in New Funding — Bloomberg

[3] What is Physical AI? | NVIDIA Glossary

[4] The global market for humanoid robots could reach $38 billion by 2035 — Goldman Sachs

🤝 Passionate about HardTech?

Send me a note: https://bit.ly/3SXPifd

You can also find me on LinkedIn 🛠️

If you are interested in AI, Robotics & HardTech, check my previous posts:

❗️The VC Industry Loves Data. But Doesn’t Use It…
 — Is it why it’s ‘dying’?

➡️19 Robotic Hands Startups to Watch
➡️ The world doesn’t fix itself. These startups are trying /Top 100 HardTech Startups
➡️ Will we have a chatGPT moment for Robots?
➡️ The Coming Scarcity of Humanity
➡️ The Art and Science of Venture Capital Investing
➡️ Deep Thinkers & Builders in Tech, Science and Bio
➡️ Humanoids Don’t Exist in Isolation / It Takes an Ecosystem to Bring Them to Life!
➡️ Clone Alpha: The Next Leap in Humanoid Robotics
➡️ ‘Made in China’-Robotics Market Overview
➡️ The Science of Networks: From Theory to New Realities