How's Figure AI doing these days?
In our humanoid robotics market deck, you will find everything you need to understand the market
SUMMARY
Figure AI is doing well these days, and it now looks like a credible humanoid robotics frontrunner rather than a pure hype company.
The strongest change is that Figure is no longer relying on one type of proof. BMW runtime, BotQ production data, Catalyst customer expansion, Helix demos, field-service hiring, and package-sorting endurance all point toward the same operating direction.
BMW remains the cleanest proof that Figure AI is doing real work. The company has disclosed production-line contribution, runtime, cycle-time targets, placement success, and zero human resets per shift, which are much more useful than another polished robot video.
The package-sorting livestream was more interesting than it first looked. The viral part got attention, but the real signal was dull endurance: robots doing the same repetitive task for days, near human speed, while still showing accuracy limits.
Helix is becoming the center of the story. The moat is less “one amazing robot” and more the learning loop between deployed robots, human data collection, teleoperation infrastructure, video pretraining, and repeated task transfer.
BotQ may be the most underappreciated signal because humanoid robotics will not scale without manufacturing discipline. Figure’s disclosed throughput, actuator, battery, yield, and end-of-line metrics suggest the company is now attacking the boring industrialization problems.
The commercial proof is improving, but it is still uneven. BMW has operating metrics, while Catalyst and Brookfield create promising customer pathways without yet showing public runtime, intervention rates, uptime, or customer economics.
Teleoperation is not proof that Figure is fake, but it is proof that autonomy is still being built through heavy human data work. The important question is whether that human-data engine becomes a scalable training advantage or a hidden deployment cost.
Competition is getting stronger, not weaker. Agility has external logistics proof, Apptronik has a very deep strategic investor base, and Neura’s recent funding shows that capital is backing several full-stack humanoid players at once.
The $39 billion valuation only works if Figure becomes a robot-labor platform. It looks stretched if the business is just expensive robot hardware sold one customer site at a time.
Safety is becoming a real deployment gate. Factories and warehouses can manage risk through controlled workflows, but homes raise the bar on certification, liability, privacy, trust, and everyday human proximity.
The practical conclusion is that Figure AI deserves serious attention now, but it is not a proven winner yet. The next test is whether internal proof turns into externally verified customer economics across more than BMW.
This market map, featured in our humanoid robotics market deck, highlights top companies and startups in the humanoid robotics market
Is Figure AI actually doing real work now?
Figure AI is doing real work now, and BMW remains the best proof that this is no longer just a demo company.
The strongest signal is indeed Figure’s November 2025 BMW disclosure.
The company said its F.02 robot contributed to 30,000 BMW X3 vehicles, loaded more than 90,000 sheet-metal parts, logged over 1,250 runtime hours, and operated 10-hour weekday shifts at BMW’s Spartanburg plant.
The more useful detail is the KPI layer: Figure disclosed an 84-second total cycle-time target, a 37-second load-time target, over 99% placement success per shift, and zero human resets per shift. That is the kind of boring metric we want, because real robotics companies are judged on cycle time and interventions, not on whether the video looks futuristic.
There is also a subtler pattern in how the BMW proof evolved. In October 2025, Adcock posted that Figure had passed five months on the BMW X3 body-shop production line, and press coverage around that claim noted previous scrutiny over the real scale of the deployment.
Then in November, Figure came back with a more detailed production note. That sequence matters. It suggests the company understood that the market no longer wanted “we’re at BMW” but evidence of runtime, repetition, and customer-site endurance.
However, BMW has not publicly disclosed the economics of the deployment, the number of robots, or whether Figure is expanding across lines. So Figure has real factory work, but not yet the kind of customer-verified scale data that would close the debate.
If you want more recent data on this point, please see our latest humanoid robotics market report.
Is Figure AI still just making viral robot videos?
Figure AI still uses viral videos aggressively, but not only.
The May 2026 package-sorting livestream is the clearest example. Figure first framed it as an 8-hour autonomous shift, then the run stretched past 24 hours, then to roughly 81 hours, with more than 100,000 packages reportedly sorted.
The interesting part? Figure let people watch the robots do the same dull task for days. In humanoids, boring repetition is actually more valuable than a spectacular 30-second demo.
The human-vs-robot race made the signal more useful. Business Insider reported in May 2026 that Figure’s robot lost to an intern by only 192 packages over 10 hours: 12,924 packages for the human versus 12,732 for the robot, with average handling times of 2.79 seconds and 2.83 seconds.
That is a much more informative comparison than “near human speed.” The robot was slower, but only slightly slower in a repetitive endurance task, and it does not need bathroom or meal breaks. For logistics, that gap matters.
The negative read is also important. Roboticist Ayanna Howard told Business Insider the technology was still not ready for deployment because of accuracy issues such as barcode orientation mistakes and packages being knocked off the belt.
Figure is showing endurance before it has fully shown industrial-grade reliability.
As this chart shows, and as featured in our humanoid robotics market deck, search interest in where to buy robots has been rising steadily
Is Figure AI’s Helix system becoming a real moat now?
Yes, Helix is now the strongest part of Figure AI’s story, because it connects the robot demos, the hiring pattern, and the manufacturing ramp into one learning loop.
The proof has been building quickly. Figure introduced Helix as a vision-language-action model in February 2025, then showed logistics progress through 2025, household tasks such as laundry and dishwasher loading, and Helix 02 full-body autonomy in January 2026.
By May 2026, the company was showing two humanoids resetting a bedroom with a single learned policy and no central planner or message passing. That matters because humanoid robotics is less about one task and more about whether the same learning system keeps transferring.
The deeper signal is in the jobs, not the videos. Figure’s recent hiring page shows roles across Helix AI, video pretraining, reinforcement learning, data infrastructure, data quality, data operations, teleoperation hardware, Helix data creators, and humanoid robot pilots. One job post describes the Pilot team as people who wear sensors or teleoperation equipment, guide robots through behaviors, collect training data, and upload that data to Figure’s AI training system. That is not a side function. It tells us Figure is building an internal data factory around Helix.
This is where the OpenAI split becomes more meaningful. Adcock’s 2025 decision to end the OpenAI collaboration looked risky at the time. By March 2026, he was reportedly saying the partnership brought little technical value and that robotics AI could not be outsourced.
The recent Helix cadence supports that view more than it did a year ago. Today, Figure’s AI strategy looks less like “we lost OpenAI” and more like “we decided embodied AI has to be vertically integrated.”
If you want more recent data on this point, please see our latest humanoid robotics market report.
Is Figure AI becoming a real manufacturer these days?
Figure AI is now trying to become a manufacturer, and BotQ is probably the most underappreciated recent signal.
In April 2026, Figure said BotQ had produced more than 350 Figure 03 robots and increased output from one robot per day to one robot per hour in under 120 days. A 24x throughput jump is easy to hype, but the more revealing numbers are the manufacturing-quality ones: more than 9,000 actuators across 10+ SKUs, 500+ battery packs, 99.3% battery first-pass yield, and more than 80% robot end-of-line first-pass yield. Those are actual factory metrics.
The hiring data points in the same direction. Figure’s job board recently listed roles for MES/SCADA, manufacturing test, PCBA test, metrology, quality engineering, supplier quality in Asia, remanufacturing, service tooling, field service, and commercial-site deployment. That mix looks like a company preparing for defects, field failures, component variance, customer-site maintenance, and robot refurbishment. In other words, the messy middle of industrialization.
The smart read is that Figure’s current edge may be less about having the best-looking humanoid and more about compressing the iteration loop. If the company can produce, test, deploy, break, service, and retrain robots faster than peers, the moat compounds. If BotQ numbers are mostly internal-stage throughput, the market will need external deployment data before rewarding the company as a scaled manufacturer.
This chart, featured in our humanoid robotics market deck, illustrates yearly venture capital funding for humanoid robotics startups
Is Figure AI getting real customers beyond BMW now?
Figure AI is getting real customers beyond BMW, but today the customer evidence is still concentrated and early.
The freshest customer signal is Catalyst Brands. In May 2026, Figure announced a commercial agreement to deploy humanoids into Catalyst’s distribution and logistics network, starting at the Reno, Nevada distribution logistics center. Catalyst is not a random pilot logo: it sits behind brands such as JCPenney, Aéropostale, and Brooks Brothers. Even more interesting, Figure described the deal as the first commercial bridge between Figure and a Brookfield portfolio company.
That Brookfield angle is more important than it looks. In September 2025, Figure announced a strategic partnership with Brookfield to deploy humanoids across residential, office, and logistics real estate. If Catalyst becomes the first operational bridge into that ecosystem, Figure gets more than one warehouse. It gets a path to repeatable environments where it can collect data, refine tasks, and sell automation into a portfolio rather than one customer at a time.
But we should be strict. Catalyst has announced a partnership, not a BMW-style operating dataset. We do not yet have public numbers on packages moved at Reno, hours worked, intervention rates, uptime, or expansion plans.
So yes, Figure is expanding beyond BMW, but its commercial proof still rests much more on signed pathways than on repeated customer-published KPIs.
Is Figure AI secretly more dependent on teleoperation than it says?
Figure AI is probably not fake autonomous, but, yes, the company is still heavily dependent on human data work behind the scenes.
The public demos are being framed as autonomous. During the May 2026 package-sorting push, Adcock and Figure repeatedly emphasized no teleoperation, no human resets, and onboard Helix autonomy. Given how much skepticism exists around humanoid demos, that public insistence matters. TechRadar and Business Insider both reported that viewers questioned odd robot movements, while Adcock pushed back on the idea that the stream was remote-controlled.
At the same time, Figure’s own job postings show how much human data still sits underneath the autonomy. The company has been hiring humanoid robot pilots, teleoperation hardware technicians, AI data operations leads, data quality managers, Helix data creators, and commercial launch operators. One pilot role explicitly asks candidates to wear teleoperation equipment, guide robots through designated behaviors, upload collected data, and stand for 8+ hours a day.
The right interpretation is that Figure is currently autonomous in some narrow public tasks, while still using humans aggressively to build the dataset that makes future autonomy better.
That is actually a healthier story than pretending robots learn from nowhere. The diligence question is whether the human-data engine becomes cheaper and more scalable over time, or whether it hides a labor-intensive training model that slows deployment economics.
If you want more recent data on this point, please see our latest humanoid robotics market report.
This chart, featured in our humanoid robotics market deck, shows how Agility Robotics is capturing share in humanoid robotics
Is Figure AI still ahead of humanoid competitors now?
Figure AI is ahead on integrated speed, but it is not clearly ahead on third-party commercial proof.
Agility Robotics is the key counterexample. In November 2025, Agility said Digit had moved more than 100,000 totes in live commercial deployment at a GXO facility. That is a narrower use case than Figure’s general-purpose ambition, but it is a clean customer-environment metric. Figure’s package-sorting livestream may be more viral, and BMW is more industrially complex, yet Agility’s GXO tote number is exactly the type of external logistics proof that customers understand quickly.
Apptronik is another pressure point. In February 2026, Apptronik closed a $520 million extension, bringing its Series A to more than $935 million, with investors and partners including Google, Mercedes-Benz, John Deere, Jabil, GXO, and Qatar Investment Authority. That does not make Apptronik better than Figure, but it shows that the market is not waiting for one winner. Large industrial customers are placing multiple bets.
And then there is Neura. In June 2026, Neura announced up to $1.4 billion in Series C funding at roughly a $7 billion valuation, backed by names including Nvidia, Amazon, Qualcomm, Bosch, Schaeffler, and the European Investment Bank. Neura also talked about serial production and real-world training environments. That makes Figure’s $39 billion valuation look less lonely as a category signal, but also more exposed: capital is flowing to several full-stack humanoid players now.
Is Figure AI’s $39B valuation still sane today?
Figure AI’s $39 billion valuation is only sane if we underwrite it as a robot-labor platform, not as a robot hardware company.
The September 2025 Series C was enormous: more than $1 billion at a $39 billion post-money valuation, with strategic investors across chips, infrastructure, telecom, enterprise software, and real assets. Figure said the proceeds would go into Helix, BotQ, GPU infrastructure, human video, and multimodal sensor data. That use of funds matches the platform thesis: collect data, train models, manufacture robots, deploy fleets, and repeat.
The issue is that the revenue proof has not caught up with the valuation. We have BMW runtime, Catalyst access, BotQ throughput, and strong investor demand. We do not have public ARR, gross margins, robot ASPs, fleet utilization economics, service costs, or customer renewal data. For a $39 billion private company, that is a big missing block.
There was also a small but telling market-heat signal in 2025. TechCrunch reported that Figure sent cease-and-desist letters to secondary-market brokers after Adcock had talked up investor demand. That does not undermine the company’s technology, but it tells us private-market excitement around Figure was running ahead of normal transparency.
Today, the valuation is not absurd if Helix plus BotQ becomes a scaled labor network. However, it is very stretched if Figure ends up selling expensive robots one deployment at a time.
If you want more recent data on this point, please see our latest humanoid robotics market report.
This chart, featured in our humanoid robotics market deck, illustrates yearly funding for humanoid robotics startups
Is Brett Adcock becoming a bigger risk for Figure AI now?
Adcock is still a major asset for Figure AI, but Hark makes the key-person question harder to ignore.
In May 2026, Business Insider reported that Adcock’s new AI hardware startup Hark raised $700 million at a $6 billion valuation, with Parkway leading and investors including Nvidia, AMD Ventures, and Salesforce Ventures. Hark is building personalized AI systems and hardware, and the report said Figure robots already use some Hark models. That creates a plausible strategic link: Hark could strengthen Figure’s AI layer, interface, and personalization stack.
But the timing is unusual. While Figure robots were livestreaming package sorting to millions of viewers, Adcock was also preparing a separate $700 million AI hardware company. Figure is already trying to solve robot hardware, embodied AI, manufacturing, data collection, safety, commercial deployment, and customer-site operations. That is a lot for one CEO, even a very effective one.
Investors should probably diligence the operating bench under him more carefully now. If Figure has strong manufacturing, AI, safety, and commercial leaders who can run the machine day to day, Hark becomes leverage. If the company is still too founder-centric, Hark becomes a new execution risk.
Is Figure AI’s safety issue getting bigger lately?
Figure AI’s safety issue is getting bigger because the company is moving from controlled demos toward human environments.
The most visible concern is the whistleblower lawsuit filed by former safety engineer Robert Gruendel in late 2025. Public reports say Gruendel alleged he was fired after warning that Figure robots were powerful enough to fracture a human skull and after raising concerns that safety materials had been weakened for investors. Figure denied wrongdoing and reportedly said he was terminated for poor performance. These are allegations, but they are material allegations because humanoids operate near workers, furniture, children, pets, and industrial equipment.
Figure’s own product design also shows that safety is not theoretical. Figure 03 added washable soft goods, foam around pinch points, reduced size and weight versus Figure 02, battery protections, and UN38.3 battery certification. TIME’s October 2025 profile also captured the home-safety and privacy tension clearly: Adcock described home safety as one of the hardest problems the company faces, and the article raised the issue that domestic robots will collect intimate home data.
The bigger point is that safety may become a deployment gate, not a PR issue. If Figure stays in structured factories and warehouses, safety can be engineered through zones, workflows, speed limits, and training. Homes are much harder. Figure can probably keep proving commercial tasks before homes, but every step toward domestic deployment raises the certification, privacy, liability, and trust bar.
If you want more recent data on this point, please see our latest humanoid robotics market report.
This chart, featured in our humanoid robotics market deck, compares the main business model options for humanoid robot manufacturers
Is Figure AI really aiming for homes again now?
Figure AI is definitely aiming for homes again, but the near-term business still looks commercial.
Figure 03 was explicitly designed with home usability in mind: soft goods, tactile sensors, palm cameras, wireless charging, safer battery architecture, and a more consumer-friendly body. The recent Helix demos also lean household: dishwasher loading, folding laundry, living-room tidying, and bedroom reset. TIME’s October 2025 profile reported that Adcock sees useful home work as possible within single-digit years, while also noting that Figure 03 was not yet ready for full domestic use.
The commercial side is still where the proof will come first. BMW gives Figure a production-line use case. Catalyst gives it a logistics-distribution use case. The job board now includes commercial site leads, deployment engineers, field service technicians, humanoid robot operators, and launch technicians in places such as Reno and Los Angeles. That looks like a company preparing to operate robots in customer environments before asking families to bring them into kitchens.
So Figure’s current strategy looks sensible: train toward the home, monetize and validate in commercial settings. A humanoid that cannot handle repeated warehouse or factory tasks safely will not deserve to be trusted at home. Figure seems to understand that, even if the consumer narrative remains more exciting for media and investors.
How is Figure AI doing these days?
Figure AI is doing well these days, and it’s definitely a credible frontrunner, but not a proven winner yet.
The company’s strongest recent pattern is that several independent-looking signals now point in the same direction. BMW metrics show real factory work. BotQ numbers show manufacturing iteration. Job postings show commercial deployment, field service, quality, teleoperation, and data operations. Catalyst shows customer expansion beyond BMW. Helix demos show a coherent AI learning strategy. The package livestream shows endurance, even if it also exposed accuracy limits. That is enough evidence to say Figure is genuinely advancing.
The unresolved part is commercialization quality. We still need more customer-side metrics: fleet hours, task success rate, reset frequency, cost per hour, robot uptime, safety incidents, customer expansion, and revenue retention. Without those numbers, Figure remains ahead of the obvious hype curve but not fully de-risked.
So, Figure AI is one of the few humanoid startups that deserves serious investor attention today, but the next six to twelve months need to turn internal proof into externally verified customer economics.
If Figure publishes more BMW-style datasets across Catalyst or another customer site, the company’s $39 billion story becomes much easier to defend.
If the next wave remains mostly livestreams and founder posts, the gap between public excitement and operating proof will become harder to ignore.
This chart, featured in our humanoid robotics market deck, shows the revenue mix across customer segments in the humanoid robotics market
| Question | Answer / check | Evidence used |
|---|---|---|
| Is Figure AI doing real work now? | Yes. BMW gives Figure real production-line evidence, though not scaled economics yet. | BMW X3 production contribution, 90,000+ parts, 1,250+ runtime hours, 10-hour shifts, published cycle-time and reset KPIs |
| Is Figure AI still mostly viral videos? | Less than before. The videos now test endurance, but accuracy still limits deployment. | May 2026 livestream, 81-hour run, 100,000+ packages, intern race, expert criticism on barcode and handling errors |
| Is Helix becoming a moat? | Yes. The moat is the learning loop between robots, human data, and deployment. | Helix 02, household demos, two-robot learned policy, video pretraining roles, teleoperation/data-collection hiring |
| Is Figure becoming a manufacturer? | Yes. BotQ shows Figure is attacking yield, testing, and field reliability. | 350+ Figure 03 units, 24x throughput, 99.3% battery yield, 80%+ robot EOL yield, MES/quality/remanufacturing roles |
| Is Figure getting customers beyond BMW? | Yes, but customer proof is still early. | Catalyst Brands Reno agreement, Brookfield portfolio bridge, commercial-site hiring, lack of Catalyst runtime metrics |
| Is Figure hiding teleoperation dependence? | No clear evidence of fake demos, but human data collection is central. | Public no-teleoperation claims, viewer skepticism, humanoid pilot roles, teleoperation hardware roles, Helix data creator roles |
| Is Figure ahead of competitors? | Ahead on integrated speed, not clearly ahead on external proof. | Agility GXO 100,000 totes, Apptronik $935M+ Series A, Neura $1.4B raise, Tesla Optimus scale ambition |
| Is the $39B valuation sane? | Only under a robot-labor-platform thesis. | $1B+ Series C, Helix/BotQ/data use of funds, limited public revenue evidence, secondary-market heat |
| Is Adcock becoming a risk? | Hark creates leverage, but also increases key-person diligence needs. | Hark $700M raise, $6B valuation, shared AI-model link with Figure, Adcock running both companies |
| Is safety getting bigger lately? | Yes. Safety is becoming a deployment gate. | Gruendel lawsuit allegations, Figure denial, Figure 03 soft goods, UN38.3 battery certification, home-data concerns |
| Is Figure aiming for homes again? | Yes, but commercial environments remain the real proof path. | Figure 03 home design, household Helix demos, Brookfield residential angle, BMW/Catalyst/commercial deployment roles |
OUR METHODOLOGY
This analysis tests how Figure AI is doing today based on the recent public signals available across customer deployment, autonomy, manufacturing, data collection, commercial expansion, competitive position, valuation, safety, and home-readiness.
Figure AI is difficult to judge from intuition alone because the public signals point in different directions at the same time: real deployments, viral demos, aggressive fundraising, limited economics, emerging safety questions, and fast-moving competition.
So we broke the main question into the dimensions that actually matter for a humanoid robotics company. For each dimension, we looked at recent signals, prioritized the most concrete ones, compared them against what was still missing, and then formed a grounded read.
We gave the most weight to operating evidence: customer-site metrics, runtime, task volume, cycle time, reset frequency, manufacturing yield, deployment roles, and field-service preparation. We treated videos, partnerships, and valuation headlines as useful signals only when they helped clarify a real operating trend.
That structure is what makes the final answer clearer. Figure AI now has enough evidence to be taken seriously as a real robotics company, but the evidence is not yet strong enough to call it a fully de-risked commercial robotics platform.
We are not affiliated with Figure AI, its investors, its customers, or any company mentioned in this analysis. We do not own shares in Figure AI or hold any related private-market position. This analysis is for informational purposes only and should not be read as investment advice, a recommendation to buy or sell any security, or a solicitation to make any investment decision.
Key sources used for this analysis include: Figure AI on BMW production deployment, Business Insider on the robot-vs-intern package-sorting race, Business Insider on the package-sorting livestream, TechRadar on livestream skepticism, Figure AI on Helix, Business Insider on the household task demo, Figure AI on BotQ and Figure 03 production, Figure AI on the Catalyst Brands agreement, JCPenney / Catalyst Brands on humanoid automation, Figure AI on Brookfield, Figure AI on its Series C and $39B post-money valuation, TechCrunch on the $39B valuation, TechCrunch on secondary-market cease-and-desist letters, Agility Robotics on Digit moving more than 100,000 totes, Apptronik on its $935M+ Series A, NEURA Robotics on its Series C, Business Insider on Brett Adcock and Hark, Figure AI on Figure 03 product and safety design, TIME on Figure 03, home robotics, safety, and privacy, and Figure AI’s Greenhouse job post for humanoid robot pilots and data collection.
This chart, featured in our humanoid robotics market deck, shows how factory humanoid robot technology has evolved over time
Related blog posts
- How strong is fundraising in the humanoid robotics market right now?
- The startups that have raised the most funding in the humanoid robotics market
- The most highly valued startups in the humanoid robotics market
