How much does it cost per hour to operate a humanoid robot?

Based on an original analysis using manufacturer-disclosed pricing, conservative assumptions of a 5-year useful life, 16 hours per day of operation, and 350 operational days per year (28,000 total hours), plus estimated electricity, maintenance, and software costs of $0.77 per hour, all-in costs range from $1.34 per hour for the Unitree G1 to $2.56 per hour for the Figure 03. The 1X NEO costs approximately $1.48 per hour, and the Tesla Optimus approximately $1.66 per hour.

Are humanoid robots cheaper than human workers?

On a pure cost-per-hour basis, yes. At $1.34-$2.56 per hour all-in, humanoid robots are 87-95% cheaper than the average U.S. warehouse worker ($19/hour) and 91-95% cheaper than the average U.S. manufacturing worker ($28/hour). However, current humanoid robots can only perform an estimated 15-30% of tasks autonomously, so the effective cost per useful hour of work is significantly higher. Economic crossover with U.S. labor happens at approximately 8% task capability, meaning even primitive robots are already cheaper for the narrow tasks they can do.

Three Companies Opened Humanoid Robot Factories in 120 Days. At $1.48 Per Hour, the Math Against Human Workers Was Already Done.

One dollar and forty-eight cents. That is the all-in cost of one hour of labor from a 1X Technologies NEO humanoid robot, calculated from its $20,000 purchase price, conservative assumptions about useful life and uptime, and estimated operating expenses that include electricity, maintenance, and software updates. A warehouse worker in the United States costs $19 per hour, and a manufacturing worker costs $28. Nobody has published a side-by-side cost-per-hour comparison across the four companies now mass-producing humanoid robots, which is strange, because the math takes thirty seconds and the conclusion lands like a freight train.

Between January and late April 2026, three separate companies opened humanoid robot factories within a 90-mile radius of each other in California. 1X Technologies inaugurated a 58,000-square-foot facility in Hayward. Figure AI's BotQ plant hit a rate of one completed robot per hour. Tesla announced it would convert its Fremont Model S/X production line to Optimus assembly starting in July. Add Unitree Robotics in Hangzhou, which filed for a $610 million IPO on the back of 75,000 units per year of production capacity, and the global humanoid robot manufacturing base went from effectively zero to four competing factories in roughly 120 days.

Here is the number nobody ran, derived from first principles and publicly available pricing data.

The Cost-Per-Hour Calculation

Start with conservative assumptions: a five-year useful life (unproven but standard for industrial equipment amortization), 16 hours of daily operation representing two shifts, and 350 operational days per year after accounting for maintenance downtime, which yields 28,000 productive hours over the robot's lifetime. Divide the purchase price by 28,000 hours, add operating costs, and the all-in hourly rate emerges.

Operating costs break into three components that are roughly constant across manufacturers. Electricity runs approximately $0.02 per hour based on 150 watts of average draw at $0.15 per kilowatt-hour, a figure consistent with the NEO's published power consumption profile. Maintenance adds roughly $0.54 per hour, derived from a $3,000 annual budget that covers actuator replacement, sensor recalibration, and structural wear, while software and connectivity licensing adds $0.21 per hour at $1,200 annually. Total operating overhead across all manufacturers: $0.77 per hour, since operating costs are more a function of robot mass and complexity than purchase price.

Robot	Price	Capital $/hr	Operating $/hr	All-in $/hr	vs US Warehouse ($19/hr)
Unitree G1	$16,000	$0.57	$0.77	$1.34	93% cheaper
1X NEO	$20,000	$0.71	$0.77	$1.48	92% cheaper
Tesla Optimus	$25,000	$0.89	$0.77	$1.66	91% cheaper
Figure 03	~$50,000*	$1.79	$0.77	$2.56	87% cheaper

*Figure AI has not disclosed consumer pricing; $50,000 estimated from production cost analysis plus expected margin. Tesla's $25,000 reflects Musk's Davos 2026 target. Unitree's $16,000 is the configured G1 price from eWeek reporting.

Even the most expensive robot on this list, Figure's 03 at an estimated $50,000, costs 87 percent less per hour than a U.S. warehouse worker. Compare against Chinese factory labor at roughly $5.50 per hour, and the Unitree G1 at $1.34 still undercuts by 76 percent, which explains why China shipped 84.7 percent of global humanoid units in 2025 even while paying its own workers a fraction of American wages.

Why Cost Per Hour Is the Wrong Number (and the Right One)

A warehouse worker earning $19 per hour can do every task in a warehouse: pick, pack, stack, reorganize shelving at 2 a.m. because a shipment arrived mislabeled, and fix the shrink-wrap machine with a zip tie and a prayer. Current humanoid robots handle maybe 15 to 30 percent of those tasks autonomously, and the generous end of that range requires a structured environment with predictable object placements and minimal improvisation.

This means the effective cost per useful hour of labor is the all-in hourly rate divided by the capability fraction, a calculation that changes the picture dramatically but still favors the robots at surprisingly low capability thresholds.

Robot (all-in $/hr)	At 20% capability	At 50% capability	Breakeven vs $19/hr
Unitree G1 ($1.34)	$6.70	$2.68	7.1% capability
1X NEO ($1.48)	$7.40	$2.96	7.8%
Tesla Optimus ($1.66)	$8.30	$3.32	8.7%
Figure 03 ($2.56)	$12.80	$5.12	13.5%

Crossover with U.S. warehouse wages happens at roughly 8 percent capability for three of four manufacturers, a threshold so low that a robot performing one task in twelve is already cheaper than paying a person for the same hours, before accounting for health insurance, overtime, turnover costs, or the fact that the robot does not call in sick on Super Bowl Monday.

The Factory Race Is Really a Data Race

1X sold out its entire first-year production run of 10,000 NEO units in five days after the October 2025 launch, according to the company's April 30 press release. Figure AI has delivered over 350 operational units and produces nine thousand actuators across ten different SKUs, with battery line first-pass yield at 99.3 percent. Tesla plans to begin Fremont Optimus production in late July 2026, targeting capacity that dwarfs everyone else at one million units per year, eventually scaling to ten million at Giga Texas. Unitree generated $250 million in 2025 revenue with 335 percent year-over-year growth.

None of these companies are primarily in the business of selling robots. They are in the business of generating training data. Every unit deployed into a warehouse, a factory, or a consumer's home generates thousands of hours of embodied interaction data that feeds the next generation of control models. Figure's Helix S0 achieves full-body control conditioned by stereo camera perception with zero-shot sim-to-real transfer, a capability that improves directly as the fleet grows. 1X's NEO runs NVIDIA Jetson Thor for onboard inference, meaning each unit processes and learns from its environment locally before uploading training signals. Tesla's fleet learning approach, proven with Autopilot across millions of vehicles, becomes exponentially more powerful when applied to humanoid manipulation tasks across a million deployed bodies.

UBS analyst Phyllis Wang forecasts 30,000 humanoid units shipped in 2026, while EY projects 10 million deployed units by 2035. Between those two numbers lies the steepest manufacturing ramp since smartphones, and the cost-per-hour math suggests it will happen faster than most forecasters expect, because the economic incentive already exists at capability levels that today's robots have already reached for narrow tasks.

What This Analysis Does Not Know

No manufacturer discloses actual operating costs, failure rates, or mean time between repairs for deployed humanoid robots in commercial settings. The five-year useful life assumption is borrowed from industrial robotics and may be wildly optimistic for machines with dozens of actuators performing variable tasks in unstructured environments; it could also be pessimistic if modular repair and over-the-air software updates extend operational life beyond traditional industrial equipment. Maintenance costs of $3,000 per year are estimated from industrial servo replacement schedules and may significantly understate the reality for bipedal robots experiencing joint wear under load. Unitree's $16,000 price may reflect Chinese government subsidies not visible in the sticker price. Figure AI's $50,000 estimate is derived from production cost analysis rather than any disclosed pricing, and the actual consumer price could be higher or substantially lower depending on margin strategy. Capability percentages (15-30 percent) are rough industry consensus rather than standardized benchmarks, because no agreed-upon benchmark for humanoid task coverage exists.

Most fundamentally, this analysis treats humanoid robots as a labor cost substitute, which ignores the restructuring costs of redesigning workflows, facilities, and supply chains around robotic workers. A warehouse optimized for humans has aisles, shelving heights, break rooms, HVAC, and lighting designed for people. A warehouse optimized for robots looks completely different, and the transition cost from one to the other is not captured in a per-hour comparison.

The Bottom Line

Four companies are now mass-producing humanoid robots at prices that translate to $1.34 to $2.56 per hour of operation. U.S. warehouse labor costs $19. The per-hour gap is so vast that humanoid robots reach economic breakeven against American wages at just 8 percent task capability, a threshold that current models already exceed for structured, repetitive work. If you run a warehouse, a fulfillment center, or a manufacturing line, start evaluating pilot programs now: 1X's NEO waitlist is open, and Figure is deploying units to commercial partners. If you work in logistics or manufacturing, the honest assessment is that roles centered on repetitive physical tasks face displacement pressure within three to five years, not ten, and retraining toward robot fleet management, maintenance, and workflow design is the highest-value career investment available today. If you invest in industrial automation, the metric to watch is not unit shipments but fleet hours, because every hour of deployed operation generates the training data that closes the capability gap between 20 percent and 80 percent, and the cost math at 80 percent capability is not a disruption forecast. It is an extinction event for manual labor pricing.

Sources

GlobeNewsWire (April 30, 2026). 1X Technologies opens NEO factory in Hayward, CA. 58,000 sqft, 200+ employees, 10,000 units/year capacity scaling to 100,000+. VP Manufacturing Vikram Kothari from SpaceX. GlobeNewsWire
Vogon.today (May 2, 2026). Figure AI BotQ factory producing 1 robot/hour, 350+ units delivered, 9,000+ actuators, 99.3% battery first-pass yield. Helix S0 full-body control. Vogon.today
Electrek (April 22, 2026). Tesla Optimus Fremont production starting late July 2026, replacing Model S/X line. Electrek
eWeek (2026). Unitree Robotics $610M IPO filing, $250M 2025 revenue (335% growth), 75,000 units/year capacity, G1 starting at $16,000. eWeek
CleanTechnica (May 1, 2026). Tesla Optimus pricing targets $20,000-$30,000 at scale. Gen 3 hands with 50 actuators. CleanTechnica
U.S. Bureau of Labor Statistics (2025). Average hourly earnings: warehouse and storage workers $19.12/hr; manufacturing production workers $27.89/hr. BLS