What are the top physical AI startups by revenue today?

The ranking of top startups in the physical AI market by revenue

Below is a table ranking all the companies in this market by their current revenue scale. You can find our methodology at the end of this page.

If you want a deeper understanding of the market and its current dynamics, get our report covering the physical AI market.

Which physical AI companies are making the most money today?

The largest physical AI companies by revenue are not necessarily the humanoid robotics startups attracting the most investment. Roborock, Symbotic, Mobileye, Ecovacs Robotics, and Anduril Industries operate at a much larger commercial scale than most newer embodied AI companies. However, the answer changes significantly depending on whether we include publicly listed robotics manufacturers, autonomous driving businesses, defense companies, and consumer robots.

Roborock reported RMB18.695 billion in 2025 revenue, equivalent to approximately $2.6 billion. Symbotic generated $2.247 billion in fiscal 2025, while Mobileye reported $1.894 billion. Anduril is another major contender, with Sacra estimating approximately $2.2 billion in 2025 revenue, although that figure is less authoritative than audited financial results.

These businesses represent different forms of physical AI. Symbotic automates warehouse operations, Mobileye develops intelligent driving systems, Anduril builds autonomous defense technology, and Roborock sells intelligent cleaning robots.

Among younger robotics-focused companies, Geek+, Horizon Robotics, UBTECH, Unitree Robotics, and Skild AI are important names. Geek+ reported RMB3.171 billion in 2025 revenue, while Unitree generated RMB1.699 billion. Skild AI announced more than $100 million in annual recurring revenue in September 2026.

There are two distinct groups: established companies generating hundreds of millions or billions in recognized revenue, and newer AI-native robotics companies whose commercial operations are expanding but generally remain smaller.

For this analysis, physical AI means artificial intelligence integrated into systems that perceive, navigate, manipulate, or act in the physical world. We include robotics, autonomous vehicles, autonomous defense platforms, and directly related control software. We distinguish private startups from publicly listed companies and use company-wide revenue only when the business is substantially relevant to physical AI.

Company	Latest reported scale	Metric	Main business
Roborock	~$2.6B	FY2025 revenue	Consumer robotics
Symbotic	$2.247B	FY2025 revenue	Warehouse automation
Anduril Industries	~$2.2B	2025 revenue estimate	Autonomous defense
Mobileye	$1.894B	FY2025 revenue	Intelligent driving
Ecovacs Robotics	~$1.5B	FY2025 service-robotics segment revenue	Consumer robotics
Applied Intuition	~$830M	Estimated 2025 ARR	Autonomous systems software
Horizon Robotics	~$523M	FY2025 revenue	Autonomous driving AI
Geek+	~$441M	FY2025 revenue	Warehouse robotics
PROCEPT BioRobotics	$308.1M	FY2025 revenue	Robotic surgery
Shield AI	$300M	2025 revenue	Autonomous defense
UBTECH Robotics	~$279M	FY2025 revenue	Humanoid and service robotics
Unitree Robotics	~$236M	FY2025 revenue	Humanoid and quadruped robots

The dollar equivalents for Chinese companies are approximate conversions using exchange rates representative of the reported period. The table mixes audited revenue, segment revenue, management-reported revenue, and explicitly identified estimates. It is a comparison of commercial scale, not a strictly comparable accounting ranking. Some businesses are public companies rather than private startups.

Why are physical AI startups attracting so much attention now?

Physical AI startups are attracting attention because several categories of autonomous machines are moving from controlled demonstrations into commercial operations, while new robotics foundation models promise to make machines more adaptable.

Symbotic's annual revenue grew 26% to $2.247 billion in fiscal 2025. Geek+ increased revenue by 31.6% to RMB3.171 billion during 2025. Horizon Robotics grew 57.7% to RMB3.758 billion over the same calendar year.

These are materially different companies, but customers are already purchasing AI-enabled systems for warehouses and vehicles at meaningful scale.

At the same time, Skild AI announced that it crossed $100 million in ARR only ten months after its first commercial deployment. The company also reported more than $50 million in recognized revenue since deployments began.

Still, established robotics businesses and newer general-purpose AI developers remain at very different commercial stages.

What counts as a physical AI startup?

A physical AI startup is a company developing intelligent systems that can interpret their surroundings and take action in the physical world. The most relevant businesses include autonomous mobile robots, humanoids, self-driving vehicles, intelligent industrial machines, and software that controls these systems.

Physical AI is broader than humanoid robotics.

Symbotic uses AI-enabled robotics to move and organize goods inside warehouses. Waymo uses autonomous driving systems to transport passengers. Shield AI develops autonomous aircraft software. Skild AI develops models intended to control different robotic platforms.

All four companies participate in physical AI, although their products and revenue models are substantially different.

We also need to distinguish startup origins from current ownership. Symbotic, Mobileye, Horizon Robotics, Geek+, and UBTECH are publicly traded businesses. They remain useful comparisons because they establish the commercial scale that private physical AI startups are attempting to reach.

NVIDIA, Tesla, and Baidu require additional care. They have substantial physical AI activities, but their consolidated revenue includes enormous businesses unrelated to autonomous machines. Including their total revenue would make a startup ranking almost meaningless.

Our practical boundary is to prioritize companies whose products directly enable intelligent physical action, while identifying public comparables and excluding unrelated corporate revenue from company-level comparisons.

Which private physical AI startups have the highest revenue?

Anduril Industries and Applied Intuition appear among the highest-revenue private physical AI contenders, but their figures rely partly on third-party estimates. Shield AI, Quantum Systems, Saronic Technologies, and Locus Robotics provide additional evidence of substantial commercial activity.

Sacra estimates Anduril generated approximately $2.2 billion in 2025 revenue. That is an unusually large figure for a venture-backed physical AI company. However, Anduril's business includes a broad portfolio of defense systems, not exclusively autonomous software or robots.

Applied Intuition is another major contender. Sacra estimated approximately $830 million in ARR around the end of 2025. Its products support vehicle development, simulation, and autonomous systems, but ARR is not equivalent to recognized annual revenue.

Shield AI reported $300 million in 2025 revenue, according to Fortune's December 2025 reporting. Sacra separately estimated approximately $330 million in 2025 revenue for Quantum Systems and $200 million for Saronic Technologies.

Warehouse robotics is also represented by Locus Robotics, for which Sacra estimated approximately $180 million in ARR in mid-2026.

Among these companies, Shield AI's revenue has a stronger direct company-reporting basis than the independent estimates for several competitors.

Anduril leads this private-company comparison by estimated commercial scale, while Applied Intuition is a major software-oriented contender. Shield AI offers an important example of directly reported autonomous-systems revenue.

A definitive ranking remains unavailable because private-company disclosures are not uniform and some important competitors do not publish revenue at all.

How much revenue does Anduril make?

Anduril Industries is already operating at billion-dollar revenue scale, making it one of the largest private companies connected to physical AI. Its revenue, however, reflects a broad autonomous defense and military technology business rather than robotics alone.

Anduril disclosed that it reached $1 billion in revenue in 2024, according to Fortune's December 2025 reporting on the defense technology sector.

Sacra subsequently estimated approximately $2.2 billion for calendar 2025, suggesting that the company's commercial operations may have expanded substantially.

That estimated increase would represent approximately 120% growth from the earlier $1 billion figure, although we cannot treat the calculation as verified growth because the figures have different disclosure quality.

Anduril sells integrated defense capabilities involving autonomous aircraft, surveillance, sensors, command systems, and other military hardware. This business model allows the company to address contracts much larger than those available to many early-stage robotics software providers.

For perspective, its estimated 2025 revenue is more than seven times Shield AI's company-reported $300 million. The comparison includes different product portfolios and reporting methods, though.

Anduril is clearly a commercial leader in autonomous defense technology. Its estimated $2.2 billion is not a measure of revenue exclusively from AI software or autonomous robotics.

Is Symbotic the biggest warehouse robotics company by revenue?

Symbotic is among the largest publicly traded pure-play AI warehouse automation companies, with $2.247 billion in fiscal 2025 revenue. Its commercial scale is substantially greater than the latest reported annual revenue of Geek+ and most private warehouse robotics specialists.

Symbotic's fiscal 2025 investor results reported 26% revenue growth and adjusted EBITDA of $147 million, alongside a $91 million net loss.

The company generated $618 million in its fourth fiscal quarter alone, compared with $565 million in the corresponding quarter of the previous year.

Geek+, another major warehouse robotics provider, reported approximately $441 million in calendar 2025 revenue after conversion from RMB. That means Symbotic's annual revenue was roughly five times Geek+'s.

But these are not identical businesses. Symbotic typically supplies large integrated warehouse automation systems, while Geek+ specializes in autonomous mobile robots and broader robotic fulfillment solutions.

The contrast becomes clearer when examining customers and deployment models. Symbotic's systems can involve major infrastructure installations, while mobile robotics providers can serve a wider variety of warehouse configurations.

Revenue leadership demonstrates Symbotic's success in securing and delivering large-scale automation projects. It does not establish leadership in robot unit shipments, installed fleets, or recurring software revenue.

How fast is Geek+ growing compared with Symbotic?

Geek+ is growing faster than Symbotic in percentage terms, although Symbotic remains much larger in absolute revenue.

Geek+ reported RMB3.171 billion in 2025 revenue, up from RMB2.409 billion in 2024. The increase of RMB762 million represents 31.6% annual growth.

Symbotic reported 26% growth to $2.247 billion in fiscal 2025.

Geek+'s audited annual report shows that 99.9% of 2025 revenue came from robotics solutions. Its business is therefore closely aligned with physical AI rather than being a diversified technology conglomerate.

The company also reported RMB4.137 billion in new orders during 2025, approximately 31.7% higher than the previous year.

Orders were equivalent to about 1.30 times recognized annual revenue. This indicates substantial commercial demand, although orders may be fulfilled in later periods.

Geek+ reported an adjusted net profit of RMB43.8 million in 2025, compared with an adjusted loss of RMB92.2 million in 2024. Its statutory net result remained slightly negative.

Are autonomous driving companies bigger than humanoid robotics startups?

Autonomous driving companies generally have larger verified revenue bases than the best-known humanoid robotics startups. They have benefited from years of automotive integration and production contracts.

Mobileye generated $1.894 billion in 2025 revenue, increasing 15% from $1.654 billion in 2024.

Horizon Robotics reported RMB3.758 billion, approximately $523 million, in 2025 revenue, representing 57.7% annual growth.

By comparison, UBTECH generated RMB2.001 billion, or approximately $279 million, in company-wide revenue, while Unitree reported RMB1.699 billion, approximately $236 million.

The gap is larger when we isolate humanoid products. UBTECH's full-size humanoid robot products and services produced RMB820.6 million in 2025, significantly below the company's consolidated revenue.

Autonomous driving companies benefit from existing automotive supply chains. Once an intelligent driving product enters vehicle production, it can generate revenue across large manufacturing volumes.

Humanoid robots generally require new customer workflows, deployment methods, maintenance systems, and operating economics.

Autonomous driving has a clear commercial advantage today, even if that does not settle which technology will eventually address more applications.

How much revenue does Unitree Robotics make?

Unitree Robotics generated RMB1.699 billion in 2025 revenue, approximately $236 million, making it one of the most commercially developed humanoid and quadruped robotics companies.

Its IPO-related disclosures provide unusually detailed financial information for a business in this category.

According to the company's prospectus information reported through the Shanghai Stock Exchange, Unitree's revenue increased from RMB159.13 million in 2023 to RMB392.77 million in 2024 and RMB1.699 billion in 2025.

That is a roughly 10.7-fold increase over two years.

The growth also reflects a major shift in product demand. Humanoid robots represented only 1.88% of Unitree's revenue in 2023, while humanoid revenue reached approximately RMB868 million in 2025.

That was just over half of its core operating revenue.

Unitree shipped more than 5,500 humanoid robots during 2025, according to the prospectus information. This makes the company particularly important because it provides both revenue and shipment data rather than relying solely on demonstration videos or fundraising announcements.

Its reported first-half 2026 revenue expectation of RMB1.052 billion to RMB1.128 billion also indicated continued growth, though expected figures should not be confused with finalized annual results.

Is UBTECH making more money than Unitree?

UBTECH generated more total company revenue than Unitree in 2025, but Unitree generated slightly more revenue from humanoid robots specifically.

UBTECH reported RMB2.001 billion in consolidated revenue, compared with Unitree's RMB1.699 billion.

That gives UBTECH approximately 18% more company-wide revenue.

However, UBTECH's full-size humanoid robot products and services generated RMB820.6 million during 2025, whereas Unitree reported approximately RMB868 million in humanoid revenue.

Unitree therefore led this specific comparison by approximately RMB47 million, or about 6%.

Both companies experienced major changes in their revenue composition.

UBTECH's humanoid revenue increased from RMB35.6 million in 2024 to RMB820.6 million in 2025, representing more than 20-fold growth. Unitree's humanoid business also expanded from a relatively small part of its operations into its largest product category.

Metric	UBTECH Robotics	Unitree Robotics
2025 total revenue	RMB2.001B	RMB1.699B
Approximate USD equivalent	$279M	$236M
2025 humanoid revenue	RMB820.6M	~RMB868M
Humanoid share of revenue	41.0%	~51%
Main non-humanoid activities	Service and other intelligent robots	Quadruped robots and other products
Revenue evidence	Annual financial results	IPO prospectus
Total revenue leader	UBTECH	-
Humanoid revenue leader	-	Unitree

Is Figure AI making significant revenue yet?

Figure AI has demonstrated meaningful industrial deployment progress, but there is insufficient public financial evidence to place it among the largest physical AI startups by revenue.

The company has attracted considerable attention through its humanoid robotics development and industrial partnerships.

Its BMW factory deployment provides one concrete measure of operational progress.

In November 2025, Figure reported that its robots had loaded more than 90,000 parts during their deployment at BMW.

That's real work in a production environment, rather than another laboratory demonstration.

However, the number of parts handled does not tell us how much BMW paid Figure, what percentage of the deployment was commercially contracted, or how much recurring revenue the company generated.

Unitree and UBTECH disclosed hundreds of millions of renminbi in humanoid-related revenue during 2025. Figure's publicly described industrial activity does not provide an equivalent financial measure.

How much revenue is Skild AI generating, and are robot foundation models becoming a real business?

Skild AI is one of the fastest-growing commercially disclosed physical AI software startups, having exceeded $100 million in annual recurring revenue within ten months of its first commercial deployment. Its results provide some of the strongest evidence that robot foundation models are becoming a real business.

In September 2026, founders Deepak Pathak and Abhinav Gupta announced that Skild had crossed that threshold and recognized more than $50 million in revenue since beginning deployments.

The company also reported more than 60 paying customers across warehouses, factories, data centers, delivery, security, and other applications.

The most interesting detail is the composition of revenue.

According to Skild, mobility represented approximately 10% of revenue, while the remaining 90% came mainly from manipulation tasks.

Robot manipulation involves interacting with physical objects, often in changing environments where traditional automation can require substantial programming.

Skild also disclosed that autonomous mobile robotics solutions associated with its Zebra/Fetch Robotics acquisition represented approximately 4% of revenue.

The disclosure suggests that the majority of its reported recurring revenue was not simply inherited from the acquired mobile robotics business.

We should still distinguish $100 million in ARR from $100 million in annual recognized revenue. Skild's reported $50 million of cumulative recognized revenue is the stronger measure of actual sales booked over its initial deployment period.

The company's disclosed work with NVIDIA and Foxconn on robotic manipulation for Blackwell production systems illustrates a potentially valuable industrial application.

Other companies, including Physical Intelligence, Generalist AI, and several emerging embodied AI laboratories, are developing technologies with similar ambitions. However, their public technical demonstrations and investment rounds generally do not provide revenue evidence comparable with Skild's disclosure.

Can physical AI software companies make more money than robot manufacturers?

Physical AI software companies can generate substantial revenue without manufacturing every robot themselves, but the largest established hardware-and-systems businesses still have a major advantage in reported sales.

Applied Intuition is an important example of the software opportunity.

Sacra estimated approximately $830 million in ARR for the company around the end of 2025. Its products support autonomous systems development, simulation, and vehicle software.

Skild AI represents a different software approach, developing intelligence that can operate across multiple robotic applications and hardware configurations.

Its $100 million-plus ARR provides evidence of a rapidly developing commercial model.

By comparison, Symbotic reported $2.247 billion in fiscal 2025 revenue, and Geek+ generated approximately $441 million in calendar 2025 revenue from integrated robotics solutions.

The revenue figures are not directly comparable because physical systems companies recognize equipment and project revenue, while software businesses may report recurring subscription or licensing revenue.

There is also an economic distinction.

Manufacturing and installing robots can produce substantial revenue but involve equipment, components, logistics, maintenance, and working capital.

Software can potentially be deployed across more machines without equivalent increases in manufacturing expenditure. However, development, computing, deployment, and customer support costs remain important.

Are warehouse robots or humanoid robots generating more revenue?

Warehouse robotics is currently a larger demonstrated revenue market than humanoid robotics among the companies for which reliable financial information is available.

Symbotic generated $2.247 billion in fiscal 2025 revenue. Geek+ generated RMB3.171 billion, approximately $441 million, during calendar 2025.

Together, these two businesses reported roughly $2.69 billion in annual revenue, acknowledging that their fiscal periods differ.

By comparison, UBTECH and Unitree generated approximately RMB1.689 billion combined from humanoid products during 2025.

That is approximately $235 million, based on their disclosed humanoid revenue figures.

The two warehouse robotics providers therefore generated more than eleven times the combined humanoid-specific revenue of the two Chinese humanoid leaders.

This is not a comprehensive market-size calculation. It compares four important companies rather than every warehouse and humanoid manufacturer. But the scale difference is hard to ignore.

Warehouse automation benefits from clearly defined economic applications: transporting inventory, organizing storage, picking orders, and increasing fulfillment capacity.

Humanoid robots promise broader flexibility, but many deployments remain early and require customers to establish new operational procedures.

Which types of physical AI startups make the most revenue?

Physical AI companies serving established industrial, automotive, and defense customers currently demonstrate the largest revenue opportunities. These customers can support large contracts because autonomous systems address expensive, recurring operational requirements.

The clearest examples are warehouse automation and automotive intelligence.

Symbotic's fiscal 2025 revenue exceeded $2.2 billion, while Mobileye generated nearly $1.9 billion and Horizon Robotics reported approximately $523 million.

Defense technology also supports large commercial scale. Anduril's estimated revenue reached approximately $2.2 billion, while Shield AI reported $300 million in 2025.

Consumer robotics should not be ignored. Roborock's approximately $2.6 billion in annual revenue demonstrates that intelligent machines can already reach mass consumer markets.

Humanoid robotics and general-purpose robot software are developing more quickly from smaller revenue bases.

Unitree generated approximately $236 million in company-wide revenue during 2025, while Skild reported more than $100 million in ARR in September 2026.

Established purchasing processes give industrial and automotive businesses an advantage over markets where customers still need to prove entirely new uses for robots.

Physical AI segment	Representative company	Reported commercial scale
Consumer robotics	Roborock	~$2.6B FY2025 revenue
Warehouse automation	Symbotic	$2.247B FY2025 revenue
Autonomous defense	Anduril	~$2.2B estimated 2025 revenue
Intelligent driving	Mobileye	$1.894B FY2025 revenue
Autonomous systems software	Applied Intuition	~$830M estimated ARR
Warehouse mobile robotics	Geek+	~$441M FY2025 revenue
Robotic surgery	PROCEPT BioRobotics	$308.1M FY2025 revenue
Humanoid and quadruped robotics	Unitree	~$236M FY2025 revenue
Robot foundation models	Skild AI	>$100M ARR
Robotaxis	Waymo	~$355M estimated early-2026 run-rate

Is Waymo making more revenue than other robotaxi startups?

Waymo appears to be one of the largest commercial robotaxi operators by revenue, although its financial scale is based on estimates rather than separately disclosed audited results.

Sacra estimated that Waymo reached approximately $355 million in annualized revenue in February 2026.

The same research estimated approximately $284 million at the end of 2025 and $125 million at the end of 2024.

That implies the annualized revenue measure nearly tripled over roughly fourteen months.

The operating data also indicate substantial expansion.

Sacra reported that Waymo's weekly paid rides increased from approximately 200,000 in early 2025 to 450,000 by December 2025.

By comparison, Pony.ai reported $90 million in recognized revenue in 2025, while WeRide disclosed RMB684.6 million, approximately $95 million.

Those figures include broader autonomous driving businesses and are not pure robotaxi fare revenue.

Waymo's estimated annualized revenue is roughly 3.7 times WeRide's full-year recognized revenue, but the comparison combines different periods, accounting measures, and business models.

Are autonomous delivery robots actually making money?

Autonomous delivery robots are generating commercial activity, but the leading operators disclose much more about deliveries and deployments than annual revenue.

Serve Robotics provides one of the clearer financial examples.

Its second-quarter 2026 financial results reported approximately $3.24 million in quarterly revenue.

Starship Technologies reported reaching 10 million cumulative autonomous deliveries by April 2026.

Zipline reported more than 2.5 million cumulative commercial deliveries by July 2026.

These operating milestones show that autonomous delivery is already being used commercially, beyond isolated pilot programs.

However, cumulative deliveries cannot be treated as annual revenue.

Delivery services may be paid through different arrangements involving retailers, restaurants, logistics partners, enterprise contracts, and consumers. The revenue recognized by the robotics provider can differ substantially from the amount customers pay for the underlying order.

Compared with warehouse automation, autonomous delivery remains a category where operational scale is often easier to observe than financial scale.

Are autonomous defense startups generating serious revenue?

Autonomous defense is already producing hundreds of millions of dollars in revenue for leading startups, and Anduril may have reached a multibillion-dollar annual scale.

Anduril and Shield AI provide the clearest comparison.

Anduril's 2024 revenue reached $1 billion, according to company information cited by Fortune. Sacra estimated approximately $2.2 billion in 2025.

Shield AI reported $300 million in 2025 revenue, according to Fortune's December 2025 coverage.

Quantum Systems represents another important business, with Sacra estimating approximately $330 million in 2025 revenue. Saronic Technologies was estimated at approximately $200 million.

These companies serve different military requirements.

Anduril develops integrated defense systems, Shield AI focuses heavily on autonomy software and aircraft, Quantum Systems produces reconnaissance drones, and Saronic develops autonomous maritime vessels.

Defense demand extends across airborne, maritime, and integrated autonomous systems.

However, defense contracts introduce an important distinction between contract awards, funded backlog, deliveries, and recognized revenue.

A company announcing a multiyear contract has not necessarily earned the entire contract value.

How much revenue are agricultural robotics startups generating?

Agricultural robotics already has meaningful commercial adoption, but its leading revenue figures are much less visible than those of warehouse and automotive robotics companies.

XAG is a useful example.

Reporting from Guangzhou government sources in July 2026 indicated that XAG's 2025 revenue exceeded RMB1.166 billion, equivalent to more than approximately $160 million.

The company develops agricultural drones and autonomous equipment used for spraying, spreading, and other farming operations.

Agricultural robots perform tasks with direct operating costs, including labor, fuel, equipment utilization, and crop treatment.

The underlying market is also different from humanoid robotics.

A farmer purchasing or using an autonomous spraying system generally does not need to believe in the future of general-purpose AI. The purchase decision can be evaluated against existing agricultural machinery and service costs.

Other companies, including Ecorobotix, Carbon Robotics, and autonomous tractor developers, demonstrate the breadth of the category, although directly comparable current revenue figures are not consistently public.

XAG's reported sales show that agricultural physical AI can support substantial commercial businesses.

One company's revenue cannot establish a market-wide ranking, but it shows how far commercial physical AI extends beyond warehouses, humanoids, and robotaxis.

Is China ahead of the United States in physical AI revenue?

China has a particularly strong commercial position in consumer robotics and humanoid robot manufacturing, while the United States hosts several of the largest revenue-generating autonomous systems and warehouse automation businesses.

China-based Roborock generated RMB18.695 billion in 2025 revenue, and Ecovacs reported approximately RMB10.680 billion from its service-robotics segment.

Chinese businesses also have a substantial presence in intelligent driving and warehouse automation.

Horizon Robotics generated RMB3.758 billion, Geek+ generated RMB3.171 billion, and Unitree reported RMB1.699 billion.

In the United States, Symbotic reported $2.247 billion in fiscal 2025 revenue, and Anduril was estimated at approximately $2.2 billion. Mobileye, an Israel-headquartered company listed in the United States, generated $1.894 billion.

The national comparison is complicated because these companies serve different industries and are not a complete census of physical AI businesses.

China has demonstrated significant manufacturing capacity and consumer robotics revenue. The United States has major commercial positions in large-scale warehouse automation, autonomous defense, and software-intensive physical AI.

The humanoid category provides a particularly interesting contrast: Unitree and UBTECH have disclosed substantial humanoid product revenue, while leading American companies such as Figure and Agility Robotics have generally demonstrated much smaller verified revenue or less financial transparency.

Are the biggest physical AI startups profitable yet?

Some physical AI businesses have reached profitability or positive adjusted earnings, but high revenue does not guarantee positive net income.

Geek+ provides an instructive example.

The company reported RMB3.171 billion in 2025 revenue and adjusted net profit of RMB43.8 million, while its statutory net result remained a loss of RMB10.4 million.

That was a considerable improvement from its RMB831.5 million statutory loss in 2024.

Symbotic generated $2.247 billion in fiscal 2025 revenue and $147 million in adjusted EBITDA, but still reported a $91 million net loss.

PROCEPT BioRobotics also illustrates the cost of scaling specialized physical systems. It reported $308.1 million in 2025 revenue and a 64% gross margin, yet continued to incur substantial operating expenses.

Horizon Robotics generated RMB3.758 billion in annual revenue, but recorded an adjusted net loss of RMB2.812 billion in 2025.

Businesses with comparable technological relevance can have very different financial profiles.

Can smaller physical AI startups catch the revenue leaders?

Smaller physical AI startups can close the revenue gap, but doing so requires more than rapid technical progress. They need repeatable commercial deployments and a way to scale without rebuilding their systems for every customer.

The difference in starting scale is substantial.

Symbotic generated $2.247 billion in fiscal 2025 revenue. Skild AI reported more than $100 million in ARR in September 2026.

Those figures cannot be compared as identical accounting measures, but they illustrate the distance between a mature automation business and a rapidly growing AI-native software company.

Skild's revenue expansion shows that a smaller company can reach significant commercial scale quickly when its technology addresses multiple applications.

Unitree demonstrates another route.

Its revenue increased from RMB159.13 million in 2023 to RMB1.699 billion in 2025, an increase of more than tenfold.

Geek+ followed a more established systems trajectory, growing annual revenue by 31.6% in 2025 while improving profitability.

These companies show three different growth models: general-purpose robotics software, manufactured intelligent robots, and integrated industrial automation systems.

The strongest commercial evidence currently favors businesses that can reproduce deployments across customers and facilities.

What are the top physical AI startups by revenue today?

The top physical AI companies by demonstrated revenue are currently led by established robotics, autonomous driving, and defense businesses, while the strongest emerging AI-native challengers include Unitree Robotics, UBTECH, and Skild AI.

Roborock's approximately $2.6 billion in 2025 revenue makes it one of the largest robotics-oriented businesses in the broader market. Symbotic's $2.247 billion in audited fiscal revenue makes it a particularly strong example of large-scale enterprise physical AI.

Among private companies, Anduril stands out with an estimated $2.2 billion in 2025 revenue, while Applied Intuition's estimated $830 million ARR suggests substantial autonomous software activity. These figures should remain explicitly labeled as estimates rather than treated as audited results.

At the next level, Horizon Robotics, Geek+, Shield AI, Unitree, and UBTECH have demonstrated meaningful commercial scale.

The humanoid comparison is particularly revealing.

Unitree and UBTECH each generated more than RMB800 million from humanoid-related products during 2025, while Figure and several other highly visible American humanoid startups have not disclosed comparable revenue figures.

Skild AI presents a different challenge to the established hierarchy. Its $100 million-plus ARR, more than $50 million in recognized revenue, and 60-plus paying customers demonstrate that general-purpose robot intelligence can become a material commercial business without depending entirely on sales of proprietary robot hardware.

Our final judgment is that physical AI is already a multibillion-dollar commercial market, but most of its proven revenue remains concentrated in specialized applications rather than general-purpose humanoid robots.

For now, Symbotic, Anduril, Mobileye, Roborock, and other established operators define the revenue ceiling. Unitree, UBTECH, and Skild AI are among the clearest evidence-backed challengers that could change the ranking as the market develops.