What are the top startups in the robotics market?
In our robotics market deck, you will find everything you need to understand the market
SUMMARY
The top startups in the robotics market today are Zipline, Exotec and Locus Robotics on proven operating scale, with Figure AI, Agility Robotics and AgiBot leading the private humanoid race and Skild AI leading the robot-brain group.
The biggest divide in robotics is no longer between “good” and “bad” technology. It is between companies already repeating useful work at scale and companies whose valuations assume that a much broader robotics market is about to arrive.
Humanoids dominate funding, but they do not dominate deployed fleets. Warehouse and delivery companies still have far more robots doing paid work every day, which makes the current valuation gap unusually large.
Zipline stands out because it combines autonomy with millions of completed commercial missions. The new Uber relationship matters because distribution may now scale faster than Zipline could build consumer demand on its own.
Exotec and Locus show why boring repetition is such a strong robotics advantage. Their customers are no longer testing whether the systems work; they are expanding the same systems across more sites and more workflows.
Figure has the strongest U.S. humanoid upside story, but Agility has cleaner disclosed commercialization and AgiBot has a manufacturing lead that neither Western company currently matches. There is no single humanoid winner yet.
China already appears ahead in humanoid production volume, though shipment numbers are harder to interpret economically because some demand comes from training centers, research institutions and government-backed programs rather than clearly disclosed labor replacement.
The robot-foundation-model race is becoming a separate market. Skild AI has the strongest platform position, FieldAI has the clearest disclosed business traction, and Physical Intelligence remains the most research-heavy bet.
Specialized robotics remains badly underappreciated next to humanoid hype. Gecko Robotics, Dexterity, ANYbotics and Bedrock all attack expensive jobs where customers can already measure the value of removing risk, labor or downtime.
The ranking deliberately favors hard-to-fake evidence: repeated deployments, working fleets, follow-on customer adoption, contract value and operating volume. Funding and valuation still matter, but they are treated as expectations rather than proof.
The practical takeaway is simple: the strongest robotics companies today are not necessarily the ones with the biggest valuations. The leaders are the companies that combine real operating evidence with a credible path to much broader deployment.
This market map, featured in our robotics market deck, highlights top companies and startups in the robotics market
Why is robotics startup funding exploding right now?
Robotics funding is exploding right now because investors think recent AI gains can finally turn robots from narrow machines into adaptable workers.
By late June, Crunchbase had already counted $18.8 billion invested in robotics startups during 2026. That was above the $15 billion raised during all of 2025 and the $14.1 billion peak reached in 2021. The pace is unusual even by the standards of the current AI boom.
Humanoids are absorbing a huge part of that money. Dealroom counted $8.7 billion of humanoid-robotics investment through late July, nearly twice the full-year 2025 record. Figure had previously raised more than $1 billion at a $39 billion valuation. NEURA Robotics then raised $1.4 billion at roughly $7 billion. Skild AI raised another $1.4 billion at more than $14 billion. Apptronik's Series A grew to more than $935 million. More recently, Generalist added nearly $200 million to a round that has now reached $600 million.
Investors are clustering around three bets. The first is humanoid labor. The second is software that can control many different robots. The third is autonomous industrial machinery that already has an obvious job to do. The companies in those three groups are at very different stages, which is why simply ranking robotics startups by valuation produces a pretty bad answer.
Should Figure AI's $39B valuation automatically make it the top robotics startup?
Figure AI's $39 billion valuation does not automatically make it the top robotics startup; we give more weight to repeated customer work, deployment scale and how much larger the company could realistically become from here.
That definition changes which companies qualify too. We include private, independent companies whose main product either performs physical work autonomously or provides the intelligence directly controlling robots. Zipline qualifies even though its robots fly. Skild AI, FieldAI, Physical Intelligence and Generalist qualify even though they mainly build robot intelligence rather than a single proprietary machine.
We exclude companies that have already become public or sit inside larger corporations. Unitree would have ranked very highly a few weeks ago, but its recent Shanghai listing means it has now left our startup universe. Boston Dynamics belongs to Hyundai. Optimus belongs to Tesla. Agility Robotics stays in for now because its planned SPAC transaction has been announced but has yet to turn it into a listed operating company.
The ranking also rewards difficult evidence. A polished robot demonstration tells us that something can work once. Months of repetitive customer work tell us much more about reliability, maintenance, integration and cost. That difference separates several of the highest-valued robotics companies from the companies at the top of our list.
If you want more recent data on this point, please see our latest robotics market report.
As this chart shows, and as featured in our robotics market deck, search interest in robot costs has increased significantly
Are humanoid robots actually leading the robotics market today?
Humanoid startups lead robotics fundraising today, while purpose-built warehouse and delivery robots still have much larger working fleets.
Exotec, Locus Robotics and Zipline have spent years accumulating commercial operations. Their robots now work across hundreds of sites or millions of delivery missions. Humanoid deployments are finally becoming real too, especially at BMW, GXO and Toyota, but the fleets remain far smaller.
That gap explains one of the strangest features of the market. Investors are putting the highest valuations on companies with the least mature operating scale because humanoids could eventually attack a much larger labor market. Specialized robots have the reverse profile: narrower jobs, much more evidence.
Both groups belong near the top. The mistake would be treating future addressable market as if it were already deployed revenue.
| Robotics model | Companies to watch | What has been proven | What we still don't know |
|---|---|---|---|
| Warehouse robotics | Exotec, Locus Robotics | Large fleets working every day across hundreds of sites | How much general-purpose robots eventually take over |
| Autonomous delivery | Zipline | Millions of fully autonomous commercial deliveries | How broadly drone delivery works outside early markets |
| Humanoids | Figure, Agility, AgiBot, Apptronik | Real factory and logistics work is now happening | Cost, reliability and fleet economics at large scale |
| Robot foundation models | Skild AI, FieldAI, Physical Intelligence, Generalist | Models can transfer across tasks and robot types | Whether one software layer can become a true robotics platform |
Is Zipline the strongest robotics startup today?
Zipline is the strongest all-around robotics startup we found today because it combines autonomous technology with a level of commercial repetition that most robotics startups are still years away from reaching.
Zipline recently passed 2.5 million commercial deliveries, with one million of those completed during the previous year alone. The company also said the number of businesses available through its U.S. marketplace grew thirteenfold during the first half of 2026. Every flight is autonomous.
The freshest development makes Zipline more interesting. Uber recently invested in the company and is adding Zipline drones to Uber Eats, starting in markets where Zipline already operates before expanding into dozens of U.S. cities. The two companies are targeting one million drone deliveries per day by the end of 2029. That target is obviously far ahead of current volume, but Uber gives Zipline access to demand without forcing the robotics company to build the entire consumer marketplace itself.
Zipline also raised an $800 million Series H at a $7.6 billion valuation earlier this year. That valuation looks surprisingly restrained next to the $39 billion attached to Figure or the $14 billion-plus attached to Skild AI, considering how much more commercial work Zipline has already completed.
The main objection is category choice. Some readers will naturally think of Zipline as drone delivery rather than robotics. We still include it because the aircraft, autonomy stack, fleet management and physical delivery system are exactly the kind of problems robotics companies solve. On operating evidence, Zipline is our number one.
If you want more recent data on this point, please see our latest robotics market report.
This chart, featured in our robotics market deck, shows annual venture capital investment in robotics startups
Is Exotec the strongest ground-based robotics startup?
Exotec is the strongest ground-based robotics startup in our ranking because its warehouse robots are already operating at a scale that the new humanoid companies have yet to approach.
Exotec has manufactured and deployed more than 10,000 Skypod robots across more than 200 customer sites. The company says those systems have completed more than 938 million robot cycles, and Exotec crossed $1 billion in cumulative sales remarkably quickly for a hardware-heavy automation company.
The more interesting part is repeat deployment. Decathlon has standardized Exotec systems across seven logistics sites in five European countries. Typical sites in that program use roughly 150 to 200 robots, store 100,000 to 125,000 locations and handle around 150,000 to 200,000 items a day. Exotec says throughput across the program has tripled.
Other customers are building very large facilities around the same architecture. Komar's new 760,000-square-foot fulfillment center in Georgia is designed to raise throughput by as much as 50% while increasing storage density by 30%. Renault has used Exotec to cut one spare-parts order-processing workflow from roughly two hours to fifteen minutes.
That repeatability is what puts Exotec so high. Customers are no longer deciding whether warehouse robots work. They are deciding how many sites should use the same system.
Is Locus Robotics already bigger than most humanoid robot startups?
Locus Robotics is already operating at a scale that makes almost every humanoid startup look tiny by comparison.
More than 17,000 Locus robots now operate across 360-plus sites. The fleet has passed 7 billion customer picks and 190 million autonomous navigation hours. Those numbers come from warehouses rather than carefully selected test environments, so they include years of congestion, human interaction, hardware failures, layout changes and peak-season pressure.
The speed of the last billion picks is particularly useful. Locus reached 7 billion only about four and a half months after reaching 6 billion. That works out to roughly 7 million picks per day across the network during that period. We care more about that pace than the round-number milestone because it shows how heavily the installed fleet is actually being used.
Locus is also moving into harder automation. Its new Array system adds more autonomous picking and consolidation while LocusONE coordinates different robot types across the warehouse. The company is gradually expanding from robots that remove walking for human pickers toward a much more automated fulfillment system.
This gives Locus a different route to general-purpose robotics. Instead of waiting for one machine to learn every warehouse job, it can add new specialized robots into an orchestration layer that already runs 17,000 machines.
This chart, featured in our robotics market deck, breaks down Figure’s playbook in robotics
Who's actually winning the humanoid robot race right now?
Figure AI leads the U.S. humanoid race on capital and AI ambition, Agility Robotics leads on disclosed commercial use, and AgiBot leads on manufacturing volume.
Figure has the strongest single U.S. factory reference. BMW says Figure 02 worked for roughly 1,250 hours over ten months in Spartanburg, moved more than 90,000 sheet-metal components and supported production of more than 30,000 BMW X3 vehicles. BMW has since continued working with Figure on a new Figure 03 use case involving sequencing work in the same plant. That continuation is more convincing than another lab video because BMW had enough experience with the previous generation to keep going.
Agility has broader commercial evidence. Digit has passed 100,000 tote movements at GXO, and the deployment has now been running full time for more than a year. Toyota Motor Manufacturing Canada moved from a pilot to a commercial robotics-as-a-service agreement. Mercado Libre has also signed a commercial agreement, while Agility says it has more than $300 million in contract orders.
AgiBot wins a completely different contest. The Chinese company says its 15,000th embodied robot came off the production line in June after the company passed 10,000 only a few months earlier. Omdia data for 2025 put AgiBot above 5,100 humanoid shipments and at roughly 39% of the market.
So there is no clean humanoid champion yet. Figure has the biggest U.S. upside bet, Agility has the cleanest Western commercial record, and AgiBot has already built far more units.
If you want more recent data on this point, please see our latest robotics market report.
Is Agility Robotics more commercially proven than Figure AI?
Agility Robotics has stronger disclosed commercial proof today than Figure AI, even though Figure is valued at roughly fifteen times Agility's proposed SPAC valuation.
Agility's announced transaction values the company at about $2.5 billion before the deal, compared with Figure's $39 billion private valuation. Yet Agility has commercial agreements with GXO, Toyota Motor Manufacturing Canada and Mercado Libre, active work across multiple customer environments and more than $300 million of disclosed contract orders.
The GXO deployment is especially useful because Digit has already repeated the same kind of job more than 100,000 times. Agility has had to improve navigation, turning, tote handling and interaction with existing autonomous mobile robots inside a busy facility. Those boring improvements are exactly what commercial robotics needs.
Figure's BMW work remains technically impressive, and BMW's decision to test Figure 03 keeps Figure firmly near the front. We rank Figure slightly higher overall because its financing, AI stack and general-purpose ambition could create a much larger platform if the technology scales.
We have lower confidence in that ordering than we do in Zipline, Exotec and Locus. Agility could easily move above Figure if contracted orders become a large deployed fleet before Figure broadens beyond a small number of factory workflows.
This chart, featured in our robotics market deck, shows annual funding in robotics startups
Is China already ahead in humanoid robot production?
China is already ahead in humanoid robot manufacturing volume, although we have less visibility into how much of that volume represents long-term productive customer use.
AgiBot gives the clearest private-company example. Reaching 15,000 produced robots puts the company in a completely different manufacturing phase from U.S. humanoid startups that are still discussing hundreds or thousands of future units. The company has also shown multi-day factory work rather than only short demonstrations.
Unitree gives us a useful benchmark now that it is public. The company recently completed its Shanghai market debut after becoming one of the best-known Chinese humanoid manufacturers. Reporting around the listing put Unitree's previous-year revenue at roughly $252 million with about $41 million of net profit. That is a rare financial reference point in a humanoid sector where most Western competitors barely disclose revenue.
We still treat Chinese shipment numbers carefully. Financial Times reporting has documented government-backed training centers buying humanoid robots partly to produce training data. Research institutions, local governments and data facilities can therefore contribute meaningful demand alongside factories that are replacing paid human work.
That nuance changes our ranking of AgiBot. Manufacturing 15,000 robots is a major achievement and makes AgiBot one of the world's most important private robotics companies. We keep Figure and Agility slightly ahead because their customer deployments are easier to interpret economically.
Can Apptronik still catch Figure AI and Agility Robotics?
Apptronik can still become a top-three U.S. humanoid company, but Apollo now needs large production deployments rather than more impressive partnerships.
The company has nearly $1 billion of funding after expanding its Series A to more than $935 million. Its latest financing valued Apptronik above $5.5 billion. Google, Mercedes-Benz, John Deere and Qatar Investment Authority are among the investors.
Apptronik's partner set is unusually useful. Mercedes gives Apollo a real manufacturing environment. Google DeepMind works with Apptronik on robot intelligence. GXO brings warehouse experience. Jabil can manufacture Apollo and can also use the robots inside its own factories.
That combination removes several obvious excuses for slow scaling. Apptronik has capital, AI expertise, manufacturing capacity and customers willing to test the product.
The remaining gap is straightforward: we need to see many Apollo robots doing paid work repeatedly. Figure has BMW's operating record, while Agility has commercial deployments and contracted orders. Apptronik is well positioned, but it is still chasing those two on evidence.
This chart, featured in our robotics market deck, compares the main business model options for warehouse AMR robotics providers
Who's winning the robot-brain race right now?
Skild AI is currently our leader in robot foundation models, while FieldAI has the strongest disclosed commercial evidence and Generalist has become the fastest-moving new challenger.
Skild has raised more than $2 billion, including a $1.4 billion Series C that valued the company above $14 billion. More important than the financing, Skild has started buying and partnering its way into existing robot fleets. It acquired Zebra Technologies' robotics business, formerly Fetch Robotics, and is working with ABB Robotics and Universal Robots.
Skild also released S1 very recently. The company says the model can watch one video demonstration and then perform an unseen manipulation task without task-specific fine-tuning. In Skild's tests, S1 handled tasks lasting up to ten minutes and produced a sevenfold improvement on unseen tasks compared with language prompting. The results come from Skild itself, so we would like to see independent production data, but the ability being tested is exactly the one that could make robot foundation models commercially useful: teaching a new task in minutes instead of collecting hours of teleoperation data.
FieldAI has a much smaller $2 billion valuation but better disclosed business evidence. Business Insider reported that the company has passed $100 million in combined revenue and customer contracts across more than 30 customers. Its models already operate on different machines in construction, mining and other difficult environments.
Physical Intelligence remains the research-heavy contender. Recent reporting now puts its valuation around $11 billion after another large financing. The team has raised enormous amounts of capital to build general robot intelligence, while co-founder Lachy Groom has openly said he does not give investors a commercialization timeline.
Generalist is catching up quickly. TechCrunch reported this week that another roughly $200 million extension took its latest round to $600 million and lifted the valuation to around $3 billion. Its Gen 1.5 model can reportedly learn tasks from video demonstrations only three to twelve seconds long. For now, the company is working with only a handful of customers, so the commercial gap with FieldAI is still large.
| Company | Freshest important evidence | Commercial proof | Our read |
|---|---|---|---|
| Skild AI | S1 learns unseen tasks from video prompts; Fetch Robotics acquisition | Commercial partners plus access to ABB and Universal Robots ecosystems | Strongest platform position |
| FieldAI | More than $100M in revenue and contracts | 30+ customers across several regions and robot types | Best business evidence |
| Physical Intelligence | Recent valuation around $11B | Commercial timeline still deliberately unclear | Highest-end research bet |
| Generalist | Latest round reached $600M at about $3B | Only a handful of customers so far | Fastest-rising challenger |
If you want more recent data on this point, please see our latest robotics market report.
Are Gecko Robotics, Dexterity and ANYbotics stronger than the humanoid hype suggests?
Gecko Robotics, Dexterity and ANYbotics already solve expensive industrial jobs for paying customers, so we rank them above several robotics companies with much bigger funding rounds.
Gecko has perhaps the cleanest contract evidence. The U.S. Navy gave the company a five-year contract with a $71 million ceiling for robotic inspection and asset-health work, starting with 18 Pacific Fleet ships. NAES previously signed a multi-year agreement worth more than $100 million, with the option to exceed $250 million, covering power-generation infrastructure.
Those numbers are large relative to Gecko's funding history and show exactly why specialized robotics can build a good business before humanoids become general. Gecko's robots climb industrial assets, collect data humans struggle to gather safely and feed that information into software used to decide what needs repair.
Dexterity is attacking a harder manipulation problem. FedEx recently expanded Dexterity's autonomous trailer-loading system into its Hagerstown hub after several years of development and testing. Random boxes arrive in unpredictable sequences, and the robot has to decide where each package fits while building a stable wall inside the trailer. FedEx is now moving the program beyond its earlier pilot environment and into a larger production setting.
ANYbotics has a smaller fleet but unusually hard operating conditions. SAP reported that more than 200 ANYmal quadrupeds are already in productive use worldwide. They inspect refineries, chemical facilities, power infrastructure and other sites where stairs, heat, hazardous gases and difficult terrain make normal mobile robots less useful.
These companies will probably never receive Figure-sized valuations if they stay focused on narrow industrial jobs. They may still generate better robotics businesses than many general-purpose challengers because customers already know exactly why they are paying.
This chart, featured in our robotics market deck, breaks down revenue across customer segments in the robotics market
Is NEURA Robotics really Europe's best humanoid startup?
NEURA Robotics is Europe's strongest independent humanoid contender by funding and industrial backing, but its roughly $7 billion valuation still assumes a production ramp that has barely started.
NEURA raised $1.4 billion earlier this year from a group that included Nvidia, Amazon, Bosch, Schaeffler, Qualcomm and other investors. The company wants to build both robots and the Neuraverse software layer used to train and coordinate them.
Its industrial relationships are credible. Schaeffler has discussed deploying a mid-four-digit number of humanoid robots over the long term, and Bosch is helping with industrial scaling. NEURA says it plans to produce around 6,000 humanoids this year and reach tens of thousands next year.
Those targets would immediately change NEURA's position if they turn into working customer fleets. We have much less evidence today than we have for Agility's commercial deployments, Figure's BMW work or AgiBot's production numbers.
Europe is also getting more crowded. UK-based Humanoid recently raised $152 million at a $1.35 billion valuation and has its own large Schaeffler relationship. NEURA remains ahead for now, but European humanoid robotics has moved beyond a one-company race.
Which newer robotics startup could jump into the top tier next?
Bedrock Robotics has the strongest case to jump into our top tier because its construction robots have just crossed from supervised testing into fully autonomous work on live customer sites.
Bedrock retrofits existing excavators rather than manufacturing the heavy machinery itself. Earlier testing moved more than 70,000 cubic yards of material across a 130-acre manufacturing site. The company raised $270 million at a $1.75 billion valuation earlier this year, bringing total funding above $350 million.
The latest step is more important. Bedrock recently announced fully autonomous excavator deployments on live commercial projects, including a Nevada water-treatment facility, a Texas earthworks site involving millions of cubic yards and a 1.2-million-cubic-yard civil project. Removing the operator from the cab changes the economics much more than another supervised test.
Mind Robotics is the other newcomer worth watching closely. The Rivian spinoff has already raised more than $1 billion and reached a valuation of roughly $3.4 billion. Rivian gives the company an obvious first factory and a huge amount of manufacturing knowledge, but Mind is still so young that we have little operating data to rank.
General Intuition is even more speculative. Investors are discussing a valuation around $6 billion only weeks after an earlier financing valued the company at $2.3 billion. Its approach uses huge quantities of gameplay data to teach models how agents move through space and time. The robotics application could become important, but the evidence today sits much closer to an AI research bet than to Bedrock's working excavators.
Bedrock is the newcomer we would move upward fastest if its autonomous deployments keep expanding through customer projects.
This chart, featured in our robotics market deck, shows how home cleaning robot technology has evolved over time
So which robotics startups are actually the best today?
The top robotics startups today are Zipline, Exotec and Locus Robotics on proven operating scale, followed by Figure AI, Agility Robotics and AgiBot as the strongest private humanoid contenders; Skild AI leads our robot-brain group.
Zipline gets our number-one spot because millions of autonomous commercial deliveries give it the rare combination of difficult robotics, repeated consumer use and a credible path to much larger distribution through Uber. Exotec and Locus follow because warehouses already depend on their machines at a scale that general-purpose robotics has yet to reach.
Figure is our highest-ranked company whose position depends heavily on what happens next. BMW has given Figure real production credibility, but a $39 billion valuation assumes an enormous jump from the current deployment base. Agility has better commercial evidence at a fraction of that valuation. AgiBot could move above both if its huge manufacturing numbers translate into transparent, sustained customer usage.
The software race is more uncertain. Skild has the strongest combination of capital, new model capability and access to existing robot fleets. FieldAI has a more convincing business today. Physical Intelligence and Generalist could eventually become more important than either, although we would be guessing if we ranked them that way already.
Our final ranking is deliberately different from the funding leaderboard. We reward robots doing useful work repeatedly, then ask how much broader the company can become. That puts some older warehouse, delivery and inspection businesses ahead of the humanoid names attracting the biggest checks these days.
| Rank | Startup | Why we rank it here | Biggest open question |
|---|---|---|---|
| 1 | Zipline | As seen above, 2.5M+ autonomous commercial deliveries plus a new Uber distribution path | Can drone delivery scale across mainstream U.S. cities? |
| 2 | Exotec | 10,000+ robots, 200+ sites and major repeat warehouse programs | How much warehouse work eventually shifts to more general robots? |
| 3 | Locus Robotics | 17,000+ robots, 7B+ picks and an installed orchestration platform | Can Locus automate much more of the warehouse than transport alone? |
| 4 | Figure AI | Best combination of U.S. humanoid capital, AI ambition and serious BMW factory work | Can Figure build a large paying fleet quickly enough to support $39B? |
| 5 | Agility Robotics | Strongest disclosed Western humanoid commercialization and $300M+ of orders | Can Digit production scale fast enough? |
| 6 | AgiBot | Huge manufacturing lead among private humanoid companies | How much production becomes sustained commercial use? |
| 7 | Skild AI | Strongest robot-foundation-model platform position, plus S1 and existing robot channels | Can a common robot brain become a large software business? |
| 8 | FieldAI | More than $100M of revenue and contracts across 30+ customers | Can FieldAI turn industrial deployments into a broad platform? |
| 9 | Gecko Robotics | Large Navy and power-sector contracts tied to robots already doing useful work | How large can the inspection category become? |
| 10 | Dexterity | FedEx is expanding one of the hardest real-world manipulation deployments | Can complex manipulation become cheap and reliable enough for large fleets? |
| 11 | Apptronik | Nearly $1B raised with Google, Mercedes-Benz and Jabil around Apollo | When do pilots turn into hundreds or thousands of working robots? |
| 12 | ANYbotics | 200+ productive industrial robots in difficult environments | Can industrial inspection support much larger fleets? |
| 13 | Bedrock Robotics | Fully autonomous excavators have now reached live commercial sites | How quickly can Bedrock repeat deployments across contractors and machine types? |
| 14 | NEURA Robotics | Massive financing and unusually strong European industrial backing | Can NEURA turn production targets into working customer fleets? |
| 15 | Physical Intelligence | One of the deepest general robot-intelligence teams with enormous funding | When does the research become a commercial product? |
| 16 | Generalist | $600M latest round, $3B valuation and fast progress on video-taught robot skills | Can a handful of customers become broad production adoption? |
If you want more recent data on this point, please see our latest robotics market report.
OUR METHODOLOGY
This analysis asks a simple question with a messy answer: what are the top startups in the robotics market? Instead of relying on reputation, headline valuations or which companies currently have the most hype, we broke the market into the dimensions that actually separate robotics companies today.
We looked at the freshest meaningful evidence we could find across funding, valuation, installed fleets, autonomous operating volume, customer deployments, contract value, manufacturing scale, technical progress and the potential to expand beyond a company’s current use case. We then assessed those points together rather than letting one metric decide the ranking.
We treated demonstrated strength and future potential separately. A company can have exceptional technology and a huge theoretical market while still having only a small deployed fleet; another can have thousands of machines working every day in a narrower category. That distinction is especially important when comparing humanoids with warehouse, delivery and inspection robots.
Where the evidence pointed in different directions, we gave more weight to the harder-to-fake parts: repeated customer work, large working fleets, follow-on deployments, commercial agreements, contract value and independently visible operating results. Funding still counts, but mainly as evidence of how much room a company has to keep pushing.
We use “startup” here to mean a private, independent company whose core product either performs physical work autonomously or provides the intelligence directly controlling robots. That is why Zipline and robot-foundation-model companies qualify, while public companies and robotics businesses owned by larger corporations do not.
For humanoids and robot foundation models, operating histories are naturally shorter, so we also looked at manufacturing progress, customer pilots that moved into commercial use, model capability, access to existing robot fleets and whether partners were expanding rather than simply announcing a first test.
The final ranking was formed after reviewing the evidence point by point. It is deliberately not a mechanical score: the point is to distinguish companies that are already strong today from companies whose position depends much more heavily on what they may become.
The analysis reflects information available through August 2026. Key sources include Dealroom on humanoid-robotics funding, Figure AI on its Series C and valuation, BMW Group on Figure’s factory work, Skild AI on its Series C, Skild AI on S1, NEURA Robotics on its financing, Apptronik on its expanded Series A, Zipline on deliveries and funding, Uber on the Zipline partnership, Exotec on robot deployment and cumulative sales, Locus Robotics on fleet scale and picks, Agility Robotics on Digit’s GXO deployment, AgiBot on its 15,000th robot, FedEx on Dexterity’s expanded deployment, and SAP on ANYbotics’ productive fleet.
In our robotics market deck, we identify pain points entrepreneurs should prioritize
