8 Physical AI Startups Racing to Build the Robot Economy
Humanoids and robot brains just raised tens of billions. Here are the 8 Physical AI startups every VC should be tracking before the next round prices in.
Key Takeaways
- The money is real: Robotics startups raised nearly $14 billion in 2025, topping even the 2021 venture peak, according to Crunchbase.
- Two bets, one wave: The winners split into humanoid hardware makers (Figure, Apptronik, 1X) and the "robot brain" foundation-model labs (Physical Intelligence, Skild AI, Generalist).
- Valuations are vertical: Figure hit a $39B post-money valuation and Skild AI tripled to $14B in seven months, proof the market is pricing Physical AI like frontier AI, not industrial automation.
- DeepMind is the talent farm: Three of the eight were founded or co-founded by ex-Google DeepMind robotics researchers.
- Independence is the filter: Every company here is still private and standalone, which is exactly why they belong on a watchlist.
As of June 2026, based on public reporting. Funding figures sourced from company press releases, Crunchbase, TechCrunch, CNBC, Bloomberg, and The Robot Report.
For a decade, "AI" meant text on a screen. Now it has hands.
Physical AI is the new front line, and the capital agrees. Robotics startups pulled in nearly $14 billion in venture funding in 2025, up from $8.2 billion in 2024 and even past the $13.1 billion peak of 2021, according to Crunchbase. The chatbot era trained models to talk. This one is training them to move.
Physical AI is artificial intelligence that perceives, reasons about, and acts in the real world through a physical body, whether that body is a humanoid robot, an industrial arm, or a wheeled platform on a construction site. The thesis is simple and enormous: software that can do physical labor is a market the size of the global workforce.
These are the eight startups racing to own it. Every one was verified as independent, private, and operating as of June 2026.
Why Is Physical AI Suddenly Worth Tens of Billions?
The market finally believes the robots will ship.
According to Goldman Sachs, the global market for humanoid robots could reach $38 billion by 2035, a forecast the bank revised upward roughly sixfold as supply chains and AI models matured faster than expected. That re-rate is exactly why valuations here look more like frontier AI than legacy industrial robotics.
The pattern splits cleanly. Some companies build the body. Some build the brain. The smartest investors are tracking both, because whoever controls the general-purpose model that runs on every robot owns the most valuable layer in the stack.
Here are the eight that matter.
FigureGeneral-purpose humanoid robots for work and home
Figure builds general-purpose humanoid robots, with its latest model Figure 03 designed to work in homes and warehouses, powered by an in-house AI system called Helix.
Figure was founded in 2022 by Brett Adcock, the serial entrepreneur behind Archer Aviation and Vettery. The company has assembled one of the deepest humanoid teams in the industry.
In September 2025, Figure exceeded $1 billion in committed capital in its Series C at a $39 billion post-money valuation, led by Parkway Venture Capital with Brookfield, NVIDIA, Intel Capital, Salesforce, and Qualcomm Ventures participating, per the company's press release.
Why it's spicy: A $39 billion valuation for a pre-mass-production hardware company is the boldest bet on this list. Figure is wagering that vertical integration, building both the robot and the brain, beats the foundation-model labs that only sell software. If they are right, they own the whole stack.
Physical IntelligenceFoundation models that control any robot
Physical Intelligence builds general-purpose foundation models for robots, releasing its pi-zero and pi-0.7 vision-language-action models that can fold laundry, bus tables, and assemble boxes across different robot hardware.
The company was co-founded in 2024 by a group of AI academics and former Google DeepMind researchers, including Sergey Levine, who pitched the idea as "ChatGPT, but for robots."
Physical Intelligence closed a $600 million Series B in November 2025 led by CapitalG at a $5.6 billion valuation, after a $400 million Series A backed by Jeff Bezos, OpenAI, and Thrive Capital. As of March 2026, Bloomberg reported it was in talks to raise roughly $1 billion more at an $11 billion-plus valuation.
Why it's spicy: This is the purest "robot brain" play in the world. If one model can drive any robot from any maker, Physical Intelligence becomes the operating system every hardware company has to license. That is a winner-take-most position.
Skild AIAn omni-bodied AI brain for robots
Skild AI is building what it calls an "omni-bodied" foundation model, a single robot brain trained largely on human videos that can control security robots, mobile manipulators, and packing arms.
Skild was founded in 2023 by Deepak Pathak and Abhinav Gupta, both former Carnegie Mellon robotics professors.
In January 2026, Skild AI raised roughly $1.4 billion in a Series C led by SoftBank at a valuation over $14 billion, with NVIDIA's NVentures, Jeff Bezos, Samsung, and LG joining, according to the company's own announcement. That tripled its valuation in about seven months.
Why it's spicy: Skild's bet on learning from human video sidesteps the single biggest bottleneck in robotics, the shortage of real-world training data. If watching humans really does scale, Skild's data advantage compounds while competitors are still teleoperating robots one demo at a time.
ApptronikApollo humanoid robots for industrial work
Apptronik builds Apollo, a 5-foot-8 humanoid robot designed for case picking, palletization, and machine tending in warehouses and factories, already piloting with Mercedes-Benz and GXO Logistics.
The company, based in Austin, is led by co-founder and CEO Jeff Cardenas and spun out of the University of Texas Human Centered Robotics Lab.
According to CNBC, Apptronik raised $520 million in February 2026 at roughly a $5 billion valuation, an extension that brought its Series A to a remarkable $935 million in total. The company also has a partnership with Google DeepMind to power Apollo with Gemini Robotics.
Why it's spicy: Apptronik is the humanoid play with real industrial pilots and a DeepMind brain partnership, hedging the build-vs-buy question by doing both. Paying customers in logistics give it something most humanoid startups still lack: a revenue story.
1X TechnologiesNEO humanoid robots for the home
1X Technologies builds NEO, a humanoid robot aimed squarely at private homes, with its entire first year of production reportedly selling out within days of opening preorders in October 2025.
1X was founded in 2014 by Norwegian roboticist Bernt Bornich, originally as Halodi Robotics. It is backed by OpenAI.
According to TechCrunch and Sifted, 1X is targeting a roughly $10 billion valuation in a new round of up to $1 billion, after previously raising a $100 million Series B that included EQT Ventures and Samsung.
Why it's spicy: Everyone else is chasing the factory. 1X is chasing your living room, a far harder safety and trust problem, but a vastly bigger market if they crack it. Selling out year-one production before shipping is the kind of demand signal VCs dream about.
GeneralistGeneral-purpose AI models that run robots
Generalist trains general-purpose AI models for robots, releasing GEN-1, which it describes as the first general model to reach mastery of simple physical tasks across factories, warehouses, and homes.
The company was founded in 2025 by Pete Florence and Andy Zeng, both former Google DeepMind senior researchers, alongside Andy Barry, formerly of Boston Dynamics.
According to Bloomberg and The Robot Report, Generalist raised $400 million in May 2026 at a $2 billion valuation, led by Radical Ventures with NVIDIA, Bezos Expeditions, Fei-Fei Li, and Zoom founder Eric Yuan participating.
Why it's spicy: Going from stealth to a $2 billion valuation in roughly a year is the kind of velocity that only happens when the best researchers in the field leave to start something. The Fei-Fei Li check is its own signal, she literally coined the modern computer-vision playbook.
Field AIA risk-aware robot brain for the real world
Field AI builds autonomy software, its EDGE robot brain and Field Foundation Models, that lets robots operate without GPS or pre-mapping in unpredictable environments like construction sites, mines, and energy facilities.
Field AI was founded by CEO Ali Agha, a veteran of NASA's Jet Propulsion Laboratory, with a team drawn from DeepMind, Tesla, and NVIDIA.
According to CNBC and GeekWire, Field AI raised a round in August 2025 that pushed total funding to roughly $405 million at a $2 billion valuation, backed by NVIDIA's NVentures, Bezos Expeditions, Bill Gates's Gates Frontier, Khosla Ventures, and Temasek.
Why it's spicy: Field AI skipped the flashy humanoid and went straight for the brutal, high-value environments where robots actually break. Production deployments across three continents mean this is operating revenue, not a demo reel. The investor list reads like a who's-who that usually only co-invests on a conviction thesis.
Agility RoboticsDigit, the first humanoid in production deployment
Agility Robotics makes Digit, a bipedal humanoid it calls the first humanoid robot in production deployment, with paying customers including Amazon, GXO, and Schaeffler.
Agility was founded in 2015 as a spinoff from Oregon State University by Jonathan Hurst, Damion Shelton, and Mikhail Jones.
According to SiliconANGLE and Tech Startups, Agility raised a $400 million Series C in 2025 at a valuation north of $1.75 billion, with backing that has included Amazon's Industrial Innovation Fund, SoftBank, DCVC, and Playground Global. Roughly 80% of Digit's nearly 6,000 parts are sourced domestically.
Why it's spicy: While rivals run demo videos, Agility quietly put humanoids on real warehouse floors with real customers. Being first to genuine production deployment is a moat made of operational scar tissue, the kind no amount of capital buys overnight.
How Should You Score a Physical AI Startup?
Hype is cheap in robotics. Demo videos are cheaper. So how do you separate the next Figure from the next cautionary tale?
The same way the best investors evaluate any frontier bet: by interrogating the team, the moat, and the traction underneath the headline number. A $14 billion valuation is not a thesis, it is a price.
This is exactly the kind of structured comparison Unicorn Screener is built for. It is a data-driven scoring tool that runs research-backed evaluations across founder quality, market dynamics, traction signals, and competitive positioning, so you can rank a humanoid hardware maker against a foundation-model lab on the same axes. See how today's most-watched names stack up on our /leaderboard before the next round reprices them.
If you want the underlying framework first, our guide on how to evaluate AI startups before writing the check walks through the diligence questions that matter most when the technology is moving faster than the financials.
What These Eight Have in Common
Step back and the pattern is obvious.
The brain may matter more than the body. Three of these companies (Physical Intelligence, Skild AI, Generalist) sell no hardware at all. They are betting the foundation model becomes the value layer, the same way it did in software AI.
DeepMind is the founder pipeline. Physical Intelligence, Generalist, and the talent inside Field AI and Apptronik all trace back to elite robotics research labs. In frontier categories, the team is the moat.
Strategic money is piling in. NVIDIA, Samsung, LG, Salesforce, and Bezos appear across multiple cap tables. When the chipmakers and electronics giants invest directly, they are buying a seat at the table for a market they expect to be enormous.
The same vertical-depth logic that drove the AI agent startups funded by top VCs in 2026 is now playing out in atoms instead of pixels. The winners pick a hard physical problem and go all the way in.
One caveat worth stating plainly: no scoring model can guarantee outcomes, and most of these valuations are priced on a future that has not shipped yet. Hardware is unforgiving, timelines slip, and today's leader can be lapped by next year's lab. Track the signals, not just the headlines.
What This Means for You
-
Decide brain or body. The foundation-model labs and the humanoid makers are different risk profiles. Pick your layer deliberately before you fall for the demo.
-
Weight the founding team heavily. In a field this young, ex-DeepMind and ex-Boston Dynamics pedigree is among the most reliable early signals you have.
-
Hunt for real deployments. Agility and Field AI have paying customers in the wild. Production revenue beats a viral video every time.
-
Score before you chase. These rounds are already priced. The edge is finding the next one early. Try Unicorn Screener to run a structured evaluation, and read how to evaluate startup unicorn potential in 2026 before your next check.
Want to screen startups like a top-tier VC? Score any startup for free with our research-backed evaluation model.