Wright's Law Predicts How Cheap Robots Will Get — Here's the Trajectory

What Is Wright's Law?

In 1936, Theodore Wright published a paper on the economics of aircraft manufacturing. He had noticed something precise and repeatable: every time cumulative production of an aircraft doubled, the cost of producing each unit fell by a fixed percentage. The relationship was not approximate. It was a law — a mathematical regularity that held across years and across manufacturers.

Wright's Law states: for every cumulative doubling of units produced, costs fall by a constant percentage.

The percentage varies by technology — solar panels decline faster than aircraft, batteries faster than wind turbines — but the shape of the curve is universal. It is not driven by calendar time. It is driven by production volume. Make more of something, learn more about making it, improve the process, reduce waste, achieve economies of scale. Repeat. The cost falls on a predictable schedule tied to cumulative output, not to the passage of years.

Wright's Law has since been validated across semiconductors, hard drives, LED lighting, wind turbines, and — most consequentially for the current moment — solar photovoltaics and lithium-ion batteries. In every case, it was dismissed as inapplicable by expert consensus. In every case, the curve held.

It is now applying to robots. The data, sourced from ARK Investment Management LLC's Big Ideas 2026 analysis, is unambiguous. The only question is how far the curve runs and how fast.

Source for all cost trajectory data in this article: ARK Investment Management LLC, Big Ideas 2026.

The Solar Precedent: $100/Watt to $0.20/Watt

In 1976, solar photovoltaic panels cost approximately $100 per watt of generating capacity. The consensus among energy analysts was straightforward: solar was too expensive to compete with coal or natural gas and would remain so for the foreseeable future. The physics were proven. The manufacturing economics were not.

By 2024, solar panels cost approximately $0.20 per watt — a 99.8% reduction over 48 years.

That is not a gradual improvement. It is one of the most dramatic cost collapses in the history of manufactured goods. Solar is now the cheapest source of electricity ever built, in most geographies, at utility scale. The expert consensus of 1980 was not wrong about the physics. It was wrong about Wright's Law.

The solar curve followed the Wright's Law prediction with remarkable fidelity. Every time cumulative global solar panel production doubled — which happened repeatedly as the industry scaled through Germany's feed-in tariffs, then China's manufacturing buildout — costs fell by approximately 20-25%. The curve compounded. Forty-eight years of doublings produced a 500x cost reduction.

The lesson is not that experts are unintelligent. It is that linear intuition is systematically wrong about exponential processes. When you look at a 99.8% cost reduction in retrospect, it appears inevitable. When you look forward from 1980, it appears impossible. Wright's Law does not care about that distinction.

The Battery Precedent: $10,000/kWh to $80/kWh

Lithium-ion battery packs for electric vehicles cost approximately $10,000 per kilowatt-hour in the early 1990s. The consensus among automotive analysts through most of the 2000s was that electric vehicles could not compete with internal combustion engines on cost or range, and that battery costs would not fall fast enough to change that within any commercially relevant timeframe.

By 2024, battery pack costs had reached approximately $80 per kilowatt-hour — a 99.2% reduction.

The electric vehicle transition from impossible to inevitable happened within a 15-year window as batteries followed Wright's Law down their cost curve. LFP (lithium iron phosphate) cells — now the dominant chemistry in Chinese-manufactured EVs and robots — have followed an even steeper decline curve than earlier nickel-based chemistries.

This matters directly for robot economics. Humanoid robots share critical component categories with electric vehicles: battery packs, electric motors, power electronics, and sensor systems. The supply chains that drove battery cost collapse are the same supply chains that are now being applied to robot actuators and power systems. The manufacturing knowledge accumulated across billions of battery cells is transferable.

For a full analysis of the cost structure underpinning the autonomous vehicle market: Robotaxi Cost Per Mile 2026 →

Applying Wright's Law to Robots

The robot cost data confirms that Wright's Law is already in effect — and the curve has been running for decades.

Industrial robots have declined approximately 50% in real terms since 2000, according to IFR World Robotics data. That decline has tracked cumulative installation volume closely. Global industrial robot installations have grown from roughly 80,000 units per year in 2000 to over 500,000 per year today. Each doubling of the installed base has been accompanied by a predictable cost reduction.

Robot vacuums are the most accessible data point for consumer intuition. The first Roomba launched in 2002 at $199. Current premium models with full base station docking, self-emptying, mop washing, and obstacle avoidance sell for approximately $150-300 in real terms — while delivering capabilities that would have been considered laboratory robotics in 2002. Adjusted for capability, the cost decline is closer to 90%.

Humanoid robot bill-of-materials costs have followed the steepest curve of any robot category:

YearBOM CostGeographyNotes
2023~$130,000United StatesPrototype-era production
2025~$50,000ChinaEarly-stage manufacturing scale
2026~$16,000ChinaUnitree G1 commercial launch price

Source: ARK Investment Management LLC, Big Ideas 2026; Unitree Robotics 2026.

That is a 62% reduction in three years — and it is not primarily driven by design improvements. It is driven by manufacturing scale. Chinese robotics manufacturers — Unitree, AgileX, Fourier — have applied the same factory-floor learning that drove LFP battery cost collapse to electric motors, planetary gearboxes, and actuator assemblies. The component cost structure is collapsing on the same Wright's Law trajectory.

The Unitree G1 at $16,000 is currently the Wright's Law frontier for commercially available humanoid platforms. The Unitree H1, its predecessor, launched at $90,000. That is a 5x cost reduction between consecutive generations from the same manufacturer, within 24 months. For context on where these platforms sit relative to each other: Unitree G1 vs H1 comparison →

For a broader analysis of what robot cost parity means for human labour economics: Robot vs Human Cost Comparison 2026 →

Where the Curve Lands: The $5,000-10,000 Humanoid

Applying Wright's Law to humanoid robots requires an assumption about the slope of the learning curve — how much costs fall per cumulative doubling of production. The battery analogy is the most structurally appropriate benchmark, given the component overlap.

Lithium-ion batteries have followed an approximately 18% cost reduction per doubling of cumulative production. If humanoid robot production follows a comparable curve — which the 2023-2026 data suggests is conservative, given the faster-than-battery decline already observed — the trajectory is as follows:

Five doublings of production from the current base — which, given the rate of investment and manufacturing buildout underway in China — is achievable within 8 to 12 years. The result is a capable humanoid robot available at $5,000-6,000.

For the broader economic analysis of what this means for the robotics market: The Robotics Economic Opportunity 2026 →

The Tesla Optimus represents the highest-profile attempt to compress this timeline through vertical integration — applying automotive-scale manufacturing to humanoid production from the outset. If Tesla's volume targets are realised, it would accelerate the doublings and steepen the cost curve further. For a comparison of where current frontier humanoid platforms sit: Unitree G1 vs Agility Robotics Digit →

Browse all humanoid robot platforms currently tracked in the directory: Humanoid Robot Category →

What $5,000-10,000 Humanoids Actually Mean

The economic implications of Wright's Law reaching its endpoint for humanoid robots are not marginal. They are structural.

At a $5,000 purchase price and an estimated $2-3 per hour operating cost (electricity, maintenance amortisation, connectivity), a humanoid robot's total cost of operation over a three-year working life is approximately $7,500-10,000.

The US federal minimum wage is $7.25 per hour. At 2,000 working hours per year over three years, a minimum-wage human worker costs a business approximately $43,500 in wages alone, before payroll taxes, benefits, absenteeism, turnover, and training costs. The fully-loaded human labour cost for a low-skill role in the United States typically runs $55,000-70,000 over three years.

At $5,000-10,000 total robot cost of operation versus $55,000-70,000 fully-loaded human cost, the economic crossover is not a narrow arbitrage. It is a 5-7x cost differential. That differential exists at any task a $5,000 humanoid can perform adequately. The threshold for adequacy — not perfection, not superiority, but adequate performance — is what determines the displacement radius.

The Universal Robots UR5e is the current standard for industrial collaborative robotics at the task level — structured manipulation in factory environments. The $5,000-10,000 humanoid does not replace the UR5e in its domain. It opens a different domain: unstructured environments, mixed-task roles, settings where a single-purpose industrial arm cannot operate. That is the majority of global labour.

At the consumer end of the spectrum, the Roborock S8 MaxV Ultra illustrates what Wright's Law produces at the cleaning robot level — a fully autonomous home cleaning system at a consumer price point, performing tasks that required human labour a decade ago. The humanoid is the next step on the same cost curve.

The Electricity Parallel — and Why It Matters for Robot Economics

Wright's Law does not only apply to robot hardware. It applies to the energy that powers robots — and here the ARK 2026 analysis raises a structural point that is under-discussed in robotics economics.

US electricity prices followed a Wright's Law declining curve from the 1890s through 1974. Then nuclear power regulation in the United States produced cost overruns that interrupted the curve. ARK's analysis projects that US electricity prices would be approximately 40% lower today if the nuclear regulatory environment of the 1970s had not disrupted the cost trajectory.

This matters for robot operating economics in the following way: a robot's ongoing cost is dominated by labour (near zero), maintenance, and electricity. At current US electricity prices, a 50kg humanoid robot running an 8-hour shift consumes approximately $0.50-1.00 in electricity. At 40% lower electricity prices — the trajectory Wright's Law predicted before the nuclear interruption — that falls to $0.30-0.60.

The compounding effect is significant over a fleet at scale. A logistics operator running 10,000 humanoid robots on three shifts faces electricity as a material operating line item. The difference between current and Wright's Law electricity pricing is the difference between a robot operating cost of $2.50/hour and $1.80/hour — a 28% reduction in ongoing cost that compounds directly into ROI.

The broader point is that Wright's Law is a system-level phenomenon. Cheap robots running on cheap electricity, built from cheap batteries and actuators, represent an intersection of multiple Wright's Law curves arriving simultaneously. The confluence is not coincidental — it is the consequence of manufacturing knowledge accumulating across interconnected supply chains.

For the full strategic context of where this positions China versus the United States in the robotics race: China vs US Robot Race 2026 →

Geppetto's Position

Every analyst who said solar would never be cheap enough was using linear thinking on an exponential curve. The same analytical error is being made today about humanoid robots.

The Unitree G1 costs $16,000 in 2026. The first Roomba cost $199 in 2002 — and a capable robot vacuum now costs $150-300 while performing tasks the 2002 model could not approach. Wright's Law doesn't care about expert consensus. It cares about cumulative production volume. That volume is now accelerating.

The question is not whether humanoid robot costs will fall to $5,000-10,000. Wright's Law makes that an engineering and supply chain execution problem, not a physics problem. The question is timing — whether the production doublings happen over 8 years or 15 years — and the answer depends on how rapidly Chinese manufacturers scale, how quickly Western manufacturers respond, and whether regulatory environments accelerate or constrain deployment.

The current evidence — $130,000 to $16,000 in three years, Unitree's G1 shipping commercially today, Tesla's manufacturing ambitions, and the full battery/actuator supply chain now oriented toward humanoid production — suggests the 8-year timeline is more plausible than the 15-year one.

Geppetto tracks every humanoid platform with verified specifications as they enter the market. As Wright's Law runs its course, the directory will document each milestone.

Frequently Asked Questions

What is Wright's Law? Wright's Law states that for every cumulative doubling of units produced, costs fall by a fixed percentage. Named after Theodore Wright, who documented the relationship in aircraft manufacturing in 1936, the law has held across solar photovoltaics, batteries, semiconductors, and LED lighting. It is distinct from Moore's Law, which is time-based; Wright's Law is production-volume-based.

Has Wright's Law been validated for robots? Yes. Industrial robot costs have declined approximately 50% in real terms since 2000 as global installation volumes scaled. Robot vacuum prices have fallen roughly 90% on a capability-adjusted basis since 2002. Humanoid robot bill-of-materials costs fell from ~$130,000 (US, 2023) to ~$50,000 (China, 2025) to the Unitree G1's $16,000 commercial launch price in 2026 — a 62% reduction in three years that tracks Wright's Law closely.

How cheap will humanoid robots get? Applying a battery-analogous learning curve (approximately 18% cost reduction per cumulative production doubling) to the current $16,000 humanoid baseline suggests capable humanoid robots reach $5,000-10,000 within 8-12 years of mass production beginning. Mass production is beginning now.

Why does China matter so much for robot cost decline? China's manufacturing ecosystem — built on battery, EV, and electronics production at scale — is the primary engine driving humanoid robot cost decline. The supply chains for actuators, motors, power electronics, and sensors are largely Chinese, and Chinese manufacturers are applying factory-floor learning accumulated across billions of battery cells and tens of millions of EVs to robot production. This is why the G1's $16,000 price point is a Chinese product, not a US or European one.

What does a $5,000 humanoid robot mean for employment? At $5,000 purchase price and $2-3/hour operating cost, a humanoid robot's three-year total cost of operation is $7,500-10,000, compared to $55,000-70,000 fully-loaded for a minimum-wage human worker in the United States. The economic crossover is a 5-7x cost differential at any task the robot can perform adequately. This is not a marginal displacement risk — it is a structural repricing of low-skill labour economics. Geppetto's Robot Jobs Index documents the profession-level displacement risk in detail.

Is Wright's Law guaranteed to hold for robots? No historical law is guaranteed to hold indefinitely, and Wright's Law can be interrupted — as the US electricity case demonstrates, where 1970s nuclear regulation disrupted a decades-long cost decline curve. For robots, plausible interruption scenarios include: regulatory barriers to humanoid deployment that reduce demand and slow production scaling; geopolitical disruptions to the Chinese supply chain; or fundamental actuator physics limits that constrain further miniaturisation. None of these currently appear imminent, but the history of technology forecasting counsels against certainty in either direction.

Further Reading