Robotics Startup Funding Hits $9.9B in Q3 as Physical AI Surges

A record quarter in a record year

In the third quarter of 2026, venture investors put $9.9 billion into robotics startups worldwide. That is more than double the $4.3 billion Crunchbase counted for the same quarter of 2025.13 It is also the latest in a series of quarters at or close to $10 billion, so this was not a one-off spike.13 Over the first nine months of the year, robotics funding rose about 185% compared with the same period last year, and Crunchbase leaves defense companies out of that total.12

The money is concentrated in two countries. Chinese robotics startups have raised $13.2 billion so far this year and U.S. startups $12.1 billion. Germany comes third with $1.7 billion.13 The rest of the startup market shows the same pattern. Crunchbase found that robotics was one of several physical-AI sectors, along with aerospace, data centers, semiconductors and energy, that each raised $10 billion or more in Q3. A record 27 companies closed rounds of $1 billion or more that quarter.14

My reading of the data is that capital has shifted, not just grown. Investors used to see robotics as an expensive hardware gamble. Now they treat it as a place where foundation models can create large platform businesses. Whether the robots in the field can justify that view is a separate question, and this quarter's figures do not answer it.

Who got the money

The biggest rounds of the year show what investors are buying. Guangzhou-based Xpeng Robotics, which makes both AI models and physical robots, raised $900 million. Shenzhen's AI² Robotics raised a $736 million Series C to build robots for manufacturing, healthcare and services.13 In the U.S., Skild AI raised $1.4 billion for Skild Brain, its robotics foundation model. Travis Kalanick's Atoms closed a billion-dollar round that also covers his ghost-kitchen business and robotics units for mining and transport.13 In Germany, Neura Robotics raised $1.4 billion to build cognitive robots and the AI and data infrastructure behind them.13

The spread of these deals matters. The large rounds are not limited to one robot format. Crunchbase says about 75 robotics companies on its unicorn board have raised money in the past three years, and 45 of them became unicorns this year. They range from data and model companies to businesses aimed at heavy industry, manufacturing, logistics and the home.13 October continued the pattern. Bedrock Robotics raised $600 million at a $4.1 billion valuation for autonomous excavators, two months after its machines started working on active construction sites.7

Humanoids: lots of capital, numbers that don't agree

Humanoid robots get much of the attention in this cycle, but the trackers count them very differently. Dealroom data, as reported in July, had humanoid startups at $8.6 billion raised in the first half of 2026. That was already 1.8 times the total for all of 2025.17 Tracxn reports $6.61 billion in equity funding across 42 humanoid rounds through September, compared with $2.05 billion a year earlier.8 The two counts differ by billions of dollars, so any single humanoid total should be treated with caution. Both sources agree on the direction, though: humanoid funding has roughly tripled or more.

Public markets are now involved too. Unitree Robotics went public, raising $905 million, and debuted at a $9 billion valuation. It was one of several Chinese robotics listings that jumped on their first day of trading.14 Smaller deals keep coming as well. Minerva Humanoids came out of stealth with about $10 million in pre-seed funding led by General Catalyst. It is building a semi-autonomous humanoid called Roger for oil-and-gas work and public safety, and specialists control it remotely through a VR headset.1 Beijing's CASBOT announced a Series A++ round worth several hundred million yuan. It says it has more than RMB 400 million in orders tied to footwear manufacturing, although that figure is order intake reported by the company, not recognized revenue.9

Some of the people deploying robots now are skeptical. Jon Miller Schwartz, CEO of the warehouse robotics startup Ultra, argues that humanoids get most of the attention while other robot types do most of the real work, and he estimates humanoids are about five years from scaling.32 I think that skepticism is largely correct. The humanoid boom is still mostly a bet on future deployment.

The foundation-model layer is where valuations are highest

The most striking repricing is happening in robot "brains," not robot bodies. Physical Intelligence is reportedly in talks at a valuation near $20 billion, up from $2.4 billion less than two years ago.37 Other coverage put its talks closer to an $11 billion valuation. The reports may describe different points in the negotiation, and the gap is a sign of how fast these numbers are moving.26 Skild AI's $1.4 billion raise reportedly valued it at about $14 billion.23 Skild also says it passed $100 million in annual recurring revenue ten months after its first commercial deployment, which is unusually concrete evidence in a sector short on it.27

Below those leaders, the field is filling up. Dream Labs, founded by former leads of Nvidia's GR00T robotics research, raised a $90 million seed round. It also launched what it calls the first publicly available API for a robot foundation model, which turns a camera feed and a plain-English instruction into motor commands.2223 London's Machenta raised $40 million, the largest seed round ever for a U.K. robotics startup. It is building a model that learns tasks from a single demonstration and is pre-trained entirely on simulated data.2529 The company has 11 employees and no near-term plans to sell its technology.4

These two companies show the main split in embodied-AI research: what data the models learn from. Dream-1 is trained on video of people doing everyday tasks.22 Machenta says it uses no real-world pixels at all.4 Most rivals collect real demonstration data instead.26 Encord co-CEO Ulrik Stig Hansen summed up the problem. Unlike language models, robotics has no internet's worth of training data. It has to be collected from vision, touch, force and torque sensors, and storing and checking it is expensive.13 He described the field as still in its pre-training phase, with post-training coming soon.13 That helps explain why data-infrastructure companies and simulation-first startups attract so much money. Whoever solves the data shortage could set the terms for everyone else.

Warehouses: smaller checks, more proof

Warehouse automation tells a quieter story. One tracker counts 50 warehouse and logistics robotics deals in 2026 with $2.4 billion disclosed, a small fraction of the money going to humanoids and models.31 Even so, this is where robots are clearly working. Ultra raised $62 million across a seed and a $50 million Series A led by Framework Ventures. It leases packing robots to warehouses for a monthly fee and gets its AI from Physical Intelligence.3234 The company says its machines have packed more than half a million orders.32

That split between body and brain looks like the industry's emerging structure. One company builds and services the hardware, and the model provider gets real-world data to train on.32 Destro AI raised $8 million for a software layer that coordinates warehouse robots from different makers. One of its investors described the gap between a working demo and a working deployment as enormous.35 Vecna Robotics raised $31 million and says demand for its CaseFlow product has more than doubled year over year. Maven Robotics raised $100 million to build general-purpose logistics robots.36

How to read the boom

The totals need two caveats. The first is that the figures depend heavily on who is counting. Crunchbase had robotics at $18.8 billion by mid-June, against $15 billion for all of 2025.16 A PitchBook-based count had more than $23 billion by early June, against $26 billion for 2025.20 Every tracker shows a record year, but they disagree on the starting point.

The second caveat matters more: the boom is driven by megadeals. One analysis of financings through late September found that rounds of $100 million or more made up about 84% of known robotics capital. Funding in rounds under $50 million actually fell, from $1.02 billion to $882 million, compared with the same period of 2025.38 So the headline growth is real, but it comes from a small number of very large bets, not a market where most startups find it easier to raise.

I take the Q3 result as strong evidence that investors are convinced, and weak evidence that robots are being deployed at scale. The companies most likely to hold their value are the ones with real usage behind their valuations: Skild's recurring revenue, Bedrock's construction contracts, and the warehouse fleets packing orders every day.27732 For the rest, especially the humanoid makers with pilot projects but little revenue, the next year will show whether their capital can keep up with the slow pace of physical deployment.37