Humanoid robotics startups have raised $11 billion so far in 2026, more than double the $4.8 billion raised across all of 2025, according to Dealroom data from July. Figure AI leads the pack with roughly $1.9 billion in total funding and a $39 billion valuation as of October 1, 2026. The money is real. The robots on factory floors are not, at least not yet.
That gap between capital and capability is the story. Investors are pricing in a future where bipedal machines sort parts, lift boxes, and stand in for scarce labor. Plant managers, meanwhile, are still running pilots that last months and end with the robot back in its crate.
The reason this matters right now is timing. Industrial robot shipments have grown steadily for a decade, but humanoids are a different bet. They cost more, break more often, and need supervision. The 2026 funding wave says the capital is committed. Whether the hardware earns it back is an open question.
Where the Money Went
Robotics and physical AI companies raised $33.4 billion in total during 2026, per Manufacturing Dive's reporting on first-half data. Of that, robotics startups alone pulled in more than $18.8 billion in the first six months, compared with a record $15 billion for all of 2025. The acceleration is not subtle.
Figure AI sits at the top of most trackers, with Neura Robotics close behind. Deep Robotics raised $140 million in February 2026. The humanoid index maintained by industry watchers now tracks more than 100 companies, and the count keeps rising.
The pitch is consistent across these firms. Humanoids can use tools built for humans, walk through spaces built for humans, and slot into workflows without rebuilding a line. That flexibility is worth a premium if it holds up.
The Factory Floor Test
The premium only matters if the machines work. Most humanoid deployments today are narrow pilots: a single task, a fixed station, a human operator nearby. Figure has run trials at BMW's Spartanburg plant. Agility Robotics has tested its Digit robot at Amazon fulfillment centers. Both programs are real, and both are limited.
The limits show up in uptime. Battery life on current humanoids runs two to four hours under load. Joints and hands wear faster than fixed-arm robots. When a humanoid falls, it can damage itself and the equipment around it. None of this is fatal, but it is expensive, and plant managers price risk honestly.
It is not clear that the current generation of hardware will hold up under daily industrial use. That is the honest read, and it is why several manufacturers have slowed their purchase timelines even as they keep talking to vendors.
The Hardware Bottleneck
Hands are the hard part. A humanoid that can walk and balance still struggles with fine manipulation: picking up a small part, inserting it, and repeating that ten thousand times a day. Human hands do this without thought. Robot hands do it with cables, motors, and software that is still being tuned.
Actuators are the other constraint. The motors that move a humanoid's joints need to be light, strong, and cheap. Right now they are usually two of those three. Companies like Figure and Tesla have invested in building their own actuator supply chains, which signals how serious the bottleneck is.
Software is catching up faster than hardware. Vision models and reinforcement learning have improved grasping and locomotion in lab settings. The jump from lab to factory is where projects stall, because factory floors are messy, loud, and unforgiving.
Regulation and the Airspace Question
Drones are a separate but related story, and they face their own ceiling. The FAA's beyond-visual-line-of-sight rules remain the gatekeeper for scaled delivery. Amazon Prime Air, Wing, and Zipline all operate under waivers or limited approvals. None of them runs a national network the way a logistics company would need.
The BVLOS rulemaking has been slow. Congress has pushed the agency, and the agency has pushed back with safety concerns. The result is a patchwork: rural medical delivery works, suburban package delivery mostly does not.
The Drone Parallel
Drone delivery offers a cautionary tale for the humanoid crowd. The technology arrived years before the rules did. Zipline now runs medical deliveries at scale in Rwanda and Ghana, and has expanded to parts of the United States. Wing delivers in a handful of suburbs. Neither is the mass-market service that early investors promised.
The lesson is that capital alone does not create a market. Regulation, infrastructure, and unit economics all have to line up. Humanoids face a version of the same problem, with the added complication that they interact with humans in shared space.
What Buyers Are Actually Doing
Buyers are hedging. Instead of buying humanoids outright, large manufacturers are signing pilot agreements with milestone clauses. Payment follows proof, not promise. That structure protects the buyer and starves the startup of recurring revenue.
The startups need that recurring revenue to justify their valuations. A $39 billion valuation on $1.9 billion raised implies an expectation of billions in annual revenue within a few years. No humanoid company is close to that today.
The tension is structural, not personal. Investors and buyers are looking at the same machine and pricing it differently. One side is buying a story. The other is buying uptime.
The Cheaper Path Forward
Not every robot company is chasing the humanoid form. Mobile bases with a single arm cost a fraction of a bipedal machine and cover most warehouse tasks. Fixed industrial arms are cheaper still and have decades of reliability data behind them.
The humanoid bet is that general-purpose hardware will eventually be cheap and reliable enough to win on flexibility. That may happen. It has not happened yet, and the funding numbers do not change that.
For more on industrial automation coverage, see Manufacturing Dive's robotics reporting. For related coverage on this site, see our robotics and drones category.
The next twelve months will show whether pilot programs convert to purchase orders. If they do, the funding surge looks prescient. If they do not, the sector faces a correction that will make the 2026 numbers look like a peak rather than a baseline.