Signals Inbox·July 28, 2026·Humanoid Robotics

Is Figure beating Tesla in factory robots?

Figure is beating Tesla in factory robots today: its humanoids have logged real production hours at BMW, while Optimus is still being judged mainly on the scale Tesla plans to build.

Q1Why are Figure and Tesla being compared in factory robots now?

Figure has made the comparison with Tesla unavoidable because its humanoid robots are currently doing the kind of factory work that Optimus has promised for years.

Figure’s first BMW program moved beyond a staged demonstration and supported the production of more than 30,000 X3 vehicles. BMW has now brought the newer Figure 03 into the same Spartanburg plant for a harder logistics job: picking unsorted components, placing them in the correct sequence and moving a heavy trolley along the line. BMW describes the earlier project as successful and says the new robot will work in its day-to-day production environment.

Tesla is heading in a different direction. Its latest available quarterly report focuses on preparing two enormous Optimus factories, including a first line in Fremont and a later line in Texas. The report says both facilities remain under construction and describes Optimus as progressing toward mass production. It gives no comparable account of Optimus contributing to current vehicle production.

That is why the comparison feels relevant now. Figure is building its case through factory deployments. Tesla is building its case through expected scale.

Q2What would it mean for Figure to beat Tesla in factory robots?

For factory robots, Figure beats Tesla only if its humanoids are doing more useful and repeatable production work today.

We are not trying to predict which company will eventually sell the most humanoids across factories, homes and other markets. Tesla could still dominate that much larger race.

The comparison here is narrower. We need to know which company has robots completing real factory tasks, how long those robots can work without help, whether the jobs are commercially useful and whether the company can build enough reliable units to serve customers.

Demo difficulty still counts, especially when it shows a capability that factories need. But a polished video carries less weight than months of work on an active production line. Announced capacity also carries less weight than robots that have already left the factory and completed useful work.

Using that definition, Figure starts ahead. Tesla has the stronger route to massive volume, but volume remains a future advantage until working Optimus units begin arriving in meaningful numbers.

Q3Is Figure doing more real factory work than Tesla Optimus today?

Figure currently has far stronger evidence of real factory work than Tesla Optimus.

Figure says its second-generation robots accumulated more than 1,250 operating hours at BMW, loaded over 90,000 sheet-metal parts and contributed to more than 30,000 X3 vehicles. The robots ran ten-hour weekday shifts during the active deployment. BMW independently confirmed that they worked under real production conditions and performed precise, repeatable tasks.

Tesla has shown Optimus performing individual factory-style activities, but it has not published a similar production record. We do not have a current total for Optimus operating hours, production cycles, interventions, parts handled or vehicles supported.

Tesla’s latest formal disclosure instead lists its Optimus facilities in California and Texas as being under construction. Tesla is clearly spending seriously on the program. It just does not show Optimus playing a meaningful role in vehicle production yet.

Current factory evidence for Figure and Tesla Optimus

Q4Was Figure’s BMW deployment a real production job?

Figure’s BMW deployment was a genuine production job, although it covered one narrow task rather than an entire assembly process.

Figure 02 collected three sheet-metal components and placed them into a welding fixture. Conventional six-axis industrial arms then welded the parts and moved them into the main production line. The humanoid handled the flexible loading step; the specialized machines kept the high-speed welding work.

Figure set clear targets for the deployment. A complete cycle had to finish within 84 seconds, including a 37-second loading window. All three pieces needed to be positioned correctly in more than 99% of cycles during a shift, within a five-millimetre tolerance. The eventual goal was to complete each shift without a human pausing or resetting the robot.

Figure did not publish the final result for every target, so we cannot calculate the exact success or intervention rate. Its claim of “minimal hardware failures” also comes from Figure rather than an independent audit.

Still, the deployment produced the kind of unglamorous engineering lesson that usually comes from real usage. The forearm became Figure 02’s most common hardware failure point. Figure responded by removing a communications board and moving cables out of the wrist design for Figure 03, reducing the number of parts that could fail.

BMW’s decision to run a second project carries more weight than Figure’s own description. Manufacturers rarely expand factory tests that taught them nothing useful.

Q5Is Tesla Optimus actually useful inside Tesla’s factories now?

Tesla Optimus still looks like a development and pre-production program inside Tesla’s factories. Tesla has not published evidence of production work comparable with Figure at BMW.

Tesla undoubtedly tests Optimus in its own facilities. Its Fremont site describes itself as a hub for Optimus production, and Tesla has recruited engineers to build robot batteries, manufacturing equipment and high-volume assembly processes. This is a serious product program, not a side experiment.

The missing piece is measurable factory output.

Tesla’s recent shareholder reports discuss a “production-primed” Optimus design, the preparation of supply chains and the construction of robot production lines. They do not say that a fleet of Optimus robots is running a recurring Tesla production job, nor do they provide shift hours, task success rates or parts moved.

Tesla may be keeping good internal results private. But private progress cannot support a public claim of leadership. Until Tesla shows what Optimus does, how often it does it and how much human help it needs, Figure owns the stronger factory record.

Q6Is Figure ahead of Tesla in autonomous manipulation?

Figure currently shows more convincing autonomous manipulation than Tesla, especially when walking and handling objects must happen together.

Figure’s Helix 02 system completed a continuous four-minute kitchen task using only onboard sensors, according to the company. The robot walked around the room, unloaded and reloaded a dishwasher, stacked items and started the machine without a reset or human intervention. Figure counted 61 separate locomotion and manipulation actions.

The hardware was built around the same problem. Figure 03 has cameras in its palms, allowing the robot to keep seeing an object when its head cameras are blocked by its hands or a container. Tactile sensors in the fingertips provide another source of feedback when objects move or slip. Figure has demonstrated the system separating small metal components from clutter and handling objects that require delicate force control.

Figure has now carried that approach back into manufacturing. In BMW’s latest sequencing project, Figure 03 must identify parts that have shifted inside a container, use both hands to move them and adjust its feet and balance while pulling a wheeled trolley. It is a more varied job than the earlier sheet-metal-loading cycle, although BMW has not released long-term output or reliability figures yet.

Tesla’s broader AI experience remains a serious advantage. The company trains vision-based systems at enormous scale and develops its own inference hardware. What it has not released is much current detail showing how the latest Optimus model handles contact, object slippage, balance and recovery during extended autonomous work.

Figure leads on what outsiders can inspect. That does not prove Helix would beat Optimus across a neutral set of factory tests; nobody has run one.

Q7Is Figure manufacturing more humanoid robots than Tesla now?

Figure is currently manufacturing more completed humanoid robots than Tesla has publicly disclosed.

Figure reported that BotQ had delivered more than 350 Figure 03 robots by late April. It also demonstrated an assembly cycle of one completed robot per hour, up from one per day less than four months earlier. Figure carefully described this as a demonstrated cycle time, so we should not assume BotQ runs at that speed every hour of every day.

The robots have several destinations. Figure uses them for AI training, data collection, internal development and commercial projects. The figure of 350 does not mean that 350 robots are working at paying customer sites.

Even so, producing hundreds of current-generation units is a meaningful step beyond hand-building a few prototypes. Figure has installed dedicated module lines, connected more than 150 manufacturing workstations and introduced over 50 inspection points across its assembly process.

Figure’s initial BotQ line was designed for up to 12,000 robots annually. Tesla has announced vastly larger capacity, but its latest report still categorizes the California and Texas Optimus facilities as being under construction. Tesla has not disclosed a current total for completed, production-intended Optimus units.

Right now, Figure has the robot count. Tesla has the larger blueprint.

Q8Can Tesla’s factory scale wipe out Figure’s lead?

Tesla can wipe out Figure’s current lead if it successfully turns Optimus into a mass-produced product. The gap between planned capacity and working robots is still enormous, though.

Tesla says its first Optimus line in Fremont is being designed for one million robots per year. A later Texas facility is being prepared for long-term capacity of ten million annually. Those targets are roughly 80 and 800 times larger than the stated 12,000-unit capacity of Figure’s first BotQ line.

Tesla has already proved that it can manufacture complicated electromechanical products at high volume. In its latest completed quarter, the company produced 451,758 vehicles. It also controls major parts of its battery, motor, power-electronics, software and AI stack.

Robots bring a different set of bottlenecks. A car can tolerate limited variation in its assembly process because every unit follows essentially the same design. A humanoid needs dozens of compact actuators, dexterous hands, cameras, sensors, computing hardware and joints that must remain reliable after repeated impacts and movements.

Tesla’s own reporting warns that installed capacity can differ sharply from the actual production rate because of equipment uptime, component availability, factory upgrades and other constraints. Its million-robot line is a design target, not a current run rate.

Tesla’s manufacturing advantage is real. It simply has not reached Optimus production yet.

Q9Does Figure have stronger factory customers than Tesla?

Figure currently has stronger outside customer validation than Tesla because BMW has tested the robots, published the outcome and returned for another project.

BMW’s follow-up is particularly useful. The company has moved from a predictable body-shop task to a sequencing problem that appears frequently in automotive logistics. Components can arrive rotated, shifted or partly hidden, and Figure 03 must place each one into the correct trolley slot. BMW says the use case offers room for further development and scaling.

Figure has also signed a commercial agreement with Catalyst Brands, which operates JCPenney, Aéropostale and Brooks Brothers. The first deployment is planned for its distribution centre in Reno, where Figure robots will target physically demanding supply-chain work. Figure has not disclosed the contract value, number of robots or deployment schedule, so the agreement remains early evidence rather than proof of a large business.

Tesla has one convenient first customer: Tesla. Internal deployment allows its engineers to change the robot, factory and task together without negotiating with another company.

That convenience comes with a drawback. Internal teams can tolerate unfinished hardware for longer than an outside customer. BMW’s involvement shows that Figure can enter another manufacturer’s production environment, connect with its processes and earn a second assignment.

No comparable external Optimus customer has been announced.

Q10Can Figure or Tesla prove the economics of factory humanoids?

Neither Figure nor Tesla has proved that factory humanoids currently deliver an attractive financial return.

BMW and Figure have not revealed the cost of the robots, integration work, maintenance, engineering support or downtime. Without those numbers, we cannot compare Figure’s hourly cost with a worker, a fixed industrial arm or a purpose-built loading machine.

Figure is making visible progress on unit cost. Figure 03 replaced many slow, machined components with stamped, moulded and die-cast parts. The company says its new 2.3-kWh battery costs 78% less than the Figure 02 battery while providing about five hours of operation and faster charging. Figure has not disclosed the full robot cost or customer price.

Tesla has not published a current Optimus cost either. Its expected advantage comes from very high production volumes and shared expertise from vehicles, batteries and power electronics. That argument becomes convincing only after Tesla builds the new robot at scale.

The type of work also determines whether a humanoid makes sense. At BMW, conventional arms still handled the welding because they are fast and well suited to that fixed process. Figure handled the awkward movement between containers, fixtures and human-designed workspaces. BMW describes humanoids as a complement to existing automation, particularly for repetitive, physically difficult or safety-sensitive jobs.

That is probably the credible factory role for now: filling changing gaps between workers and specialized machines. Replacing every industrial robot would be slower, more expensive and unnecessary.

Q11What could make Tesla overtake Figure in factory robots soon?

Tesla would move ahead quickly if Optimus begins logging substantial factory hours while its new production lines start delivering reliable units.

The standard should be concrete. Tesla would need to identify an active production task, disclose how many Optimus robots perform it, report output across several months and show that human interventions are becoming rare. Producing hundreds or thousands of the latest design would strengthen the case further.

Figure has weaknesses Tesla could exploit. Its 350-plus robots are spread across development, data collection and commercial programs rather than a large paying fleet. BotQ has demonstrated a one-hour assembly cycle, but Figure has not released a sustained monthly production rate.

The newest BMW project remains early too. We have seen Figure 03 perform the sequencing workflow, and BMW has confirmed that deployment work is underway. We do not yet know how many hours it will operate, whether the robot can maintain production speed or how often a person must intervene.

Tesla’s latest available quarterly report still places Optimus before mass production, with both planned manufacturing sites under construction. Figure has some breathing room, though probably less than its current evidence advantage suggests.

Q12Is Figure beating Tesla in factory robots?

Yes. Figure is currently beating Tesla in factory robots, although Tesla remains capable of overturning the result through manufacturing scale.

Figure leads where factory robotics can be measured today: production hours, components handled, customer confirmation, repeat deployments and completed current-generation robots. Its BMW record is still the only detailed production dataset available from either company.

Tesla leads in resources, existing manufacturing capacity and the ambition of its planned robot factories. Those advantages could eventually become decisive. For now, Tesla asks us to judge Optimus largely by what its factories are expected to produce, while Figure can point to what its robots have already done.

The economic question remains open for both companies. Figure has proved that a humanoid can contribute to an automotive production line. It has not shown that the robot delivers a strong return after purchase, integration and support costs. Tesla has not reached that test publicly.

So the title gets a clear answer: Figure is ahead in useful factory deployment today. Tesla is the more dangerous long-term competitor, but it is currently chasing rather than leading.

Figure versus Tesla in factory humanoids