What a Humanoid Robot Actually Costs: Bill of Materials, Price Curve, and Break-Even Math

What does a humanoid robot actually cost?

There is no single sticker price for a humanoid robot, because the industry spans at least three different products sold at three different price points. At the low end, Unitree’s G1 starts at $16,000 for a base research/hobbyist configuration, climbing to roughly $73,900–$74,000 for fully specced EDU units. In the middle, 1X Technologies prices its home robot NEO at $20,000 outright or $499 a month on subscription. At the high end, Agility Robotics lists its warehouse-oriented Digit at around $250,000, usually sold through a Robots-as-a-Service (RaaS) lease rather than a one-time purchase. Tesla’s Optimus doesn’t have a retail price yet, but Elon Musk has repeated a long-term consumer target of $20,000–$30,000 — a figure that describes an ambition, not a bill of materials.

That last distinction — price versus cost — is the whole story of humanoid robot unit economics. What a manufacturer charges and what a robot costs to build are two different numbers, and the gap between them is currently being subsidized by venture capital, not covered by margin. Understanding both sides of that gap is the point of this piece: the bill of materials (robot BOM), how it’s expected to fall over time, and the break-even math that will eventually decide which of today’s humanoid robot programs survive.

What’s actually inside a humanoid robot’s bill of materials?

A humanoid robot’s BOM is dominated by actuation, not by the “brain.” Roughly speaking, the largest cost buckets are:

• Actuators and joints — dozens of servo motors, harmonic/planetary gearboxes, and torque sensors that give the robot 40+ degrees of freedom. Tesla’s Optimus Gen 2 reportedly uses more than 40 servo motors alone.
• Rare-earth permanent magnets — each actuator typically needs neodymium-iron-boron (NdFeB) magnets for torque density; Optimus is estimated to require about 3.5 kg of NdFeB magnets per unit.
• Battery and power electronics — a humanoid-scale pack sized for multiple hours of bipedal operation.
• Sensing and compute — cameras, IMUs, force/torque sensors, and an onboard compute stack for perception and control.
• Structural components — the chassis, hands, and actuator housings, often the least glamorous but non-trivial share of cost.

This is why where the components are sourced changes the number so dramatically. Tesla Optimus Gen 2’s bill of materials has been estimated at roughly $46,000 when sourced through China’s supply chain, versus roughly $131,000 if the same robot were built without Chinese components — nearly a tripling of cost. That gap alone tells you more about where humanoid robot cost actually lives (rare-earth magnets, precision gearboxes, and low-cost servo manufacturing) than any spec sheet does.

In fairness, “the” BOM number is contested across the industry. Some industry-wide estimates for mainstream humanoids put the range at $30,000–$150,000; the Optimus-specific figures above put it at $46,000–$131,000; other estimates cluster around $40,000–$50,000. These aren’t necessarily contradictory — they may reflect different methodologies (single-model teardown vs. industry-wide average) — but no single number should be treated as settled. Read every “a humanoid costs $X” claim as an estimate with a methodology behind it, not a catalog price.

Where does China’s supply chain dominance come in?

China produces an estimated 85%–90%+ of the world’s NdFeB permanent magnets, the component that makes compact, high-torque actuators possible. On April 4, 2025, China enacted export controls on seven rare-earth minerals, requiring a Ministry of Commerce export license — a policy that now sits directly upstream of every humanoid robot BOM. This is not an abstract geopolitical footnote: it is the mechanical reason a non-China-sourced Optimus costs roughly three times as much as a China-sourced one. Any unit-economics model for humanoid robots built after early 2025 has to treat rare-earth and magnet supply as a line-item risk, not just a commodity input. For a deeper look at how Chinese manufacturers are competing on this exact cost axis, see our coverage of Unitree and UBTech’s approach to humanoid hardware.

Robot BOM and list price, compared

Robot	Maker	List price / model	Business model	Segment
G1	Unitree	$16,000 (base) – ~$73,900 (EDU)	Direct purchase	Research/developer
NEO	1X Technologies	$20,000 purchase or $499/month	Purchase or subscription	Home/consumer
Optimus	Tesla	No retail price yet; $20,000–$30,000 long-term target stated by Musk	Not yet sold	Industrial/consumer (future)
Digit	Agility Robotics	~$250,000 list	Robots-as-a-Service (RaaS) lease	Industrial/warehouse
Optimus Gen 2 BOM (China-sourced)	Tesla	~$46,000 estimated cost	—	—
Optimus Gen 2 BOM (non-China)	Tesla	~$131,000 estimated cost	—	—

Two things jump out. First, price and cost aren’t comparable across rows — Unitree and 1X publish retail prices, while the Optimus figures are BOM cost estimates, not what Tesla would charge. Second, the business model matters as much as the number: Agility’s $250,000 Digit is rarely paid upfront; it’s typically leased via RaaS, bundling hardware, software, maintenance, and support into a recurring fee — which changes the break-even math entirely (more below).

How is the price curve expected to move?

IDTechEx, in its “Humanoid Robots 2026–2036: Technologies, Markets, and Opportunities” market report published May 20, 2026, projects that average selling prices for humanoid robots will fall from roughly $114,700 in 2024 to roughly $37,000 by 2030 — a decline of more than 68% in six years. That trajectory tracks the same curve every hardware category follows once volume manufacturing kicks in: batteries, solar panels, and industrial robot arms all saw comparable cost declines as production scaled and component sourcing consolidated.

Whether humanoids actually follow that curve depends on production volume materializing, which is precisely the part that hasn’t happened yet. Tesla’s own reporting is a useful reality check: on the January 28, 2026 earnings call, Musk confirmed Tesla was ending Model S and Model X production entirely and converting the Fremont factory’s lines to build Optimus and Cybercab instead — a serious capital commitment. But by the Q1 2026 call on April 22, 2026, Musk pointed to Optimus’s roughly 10,000 unique parts and entirely new production line as the reason 2026 production-rate forecasting was “literally impossible to predict.” Fremont’s Optimus production wasn’t set to begin until late July/August 2026, about four months after the last Model S/X rolled off the line in early May 2026. Reported 2025 output came in at just a few hundred units against an original plan of 5,000 — and in January 2026, Musk admitted that zero Optimus units were doing “useful work” inside Tesla’s own factories, reversing his January 2025 claim that thousands would be doing useful work by the end of that year.

None of this means the price curve is wrong — it means the curve is a projection built on assumed volume, and volume is currently the industry’s least reliable variable. Treat “$37,000 by 2030” as a plausible destination contingent on manufacturing scale-up actually happening on schedule, not a locked-in forecast.

What does break-even actually look like for a buyer?

This is where humanoid robot unit economics gets genuinely uncertain, because credible sources disagree by an order of magnitude. Some industry estimates put operating cost at roughly $2–$3 per hour with payback in as little as six months under high utilization. Others put operating cost closer to $6.70 per hour with typical payback in the 18–36 month range for two-shift industrial deployments. Both ranges come from industry aggregator content rather than a single audited primary study, so the honest takeaway is that payback period is extremely sensitive to assumptions that vary by deployment — utilization rate (one shift vs. two or three), the labor cost being displaced, maintenance and downtime, and whether the deal is a capital purchase or a RaaS lease.

A simplified way to frame the math for any specific deployment:

1. Capital or lease cost — either the purchase price amortized over an expected service life, or the RaaS monthly fee (Agility’s Digit and Figure’s reported ~$1,000/robot/month RaaS pricing both use this model).
2. Operating cost per hour — power, connectivity, remote supervision/teleoperation, and maintenance.
3. Displaced labor cost per hour — what a human doing the same task costs, including overhead, in the specific market and shift pattern.
4. Utilization — hours per day the robot is actually doing productive work, which is usually the single biggest swing factor between a six-month and a three-year payback.

Because utilization varies so much by task and facility, a warehouse pilot and a home-robot subscription will land on completely different payback timelines even with an identical robot model. Our deployment-focused piece on Agility’s Digit rollouts in warehouse settings walks through what utilization looks like in a live industrial context.

Why is venture capital pricing humanoids so far ahead of unit economics?

The valuation side of the industry is running well ahead of the cost curve. Figure AI closed a Series C exceeding $1 billion at a $39 billion post-money valuation on September 16, 2025, led by Parkway Venture Capital with Brookfield, NVIDIA, Macquarie, and Intel Capital among the participants. That’s roughly a 15x jump from Figure’s February 2024 Series B, which raised $675 million at a $2.6 billion valuation — an 18-month re-rating that has little to do with today’s per-unit margins and everything to do with investors betting on the 2030-era price curve materializing. For context on how humanoid economics fit into the broader investment picture, see our market outlook for Physical AI investment.

That gap between paper valuation and proven unit economics is itself a signal worth tracking, not a footnote. A $39 billion valuation prices in years of BOM decline, production-volume ramp, and utilization improvements that haven’t been demonstrated at scale by anyone yet — Tesla included, given its own admission that no Optimus units were doing useful work as of January 2026.

How do the current price points compare across use cases?

• Research and hobbyist: Unitree’s G1 at $16,000 is the accessible entry point, aimed at labs and developers rather than deployed operations.
• Home/consumer: 1X’s NEO at $20,000 (or $499/month) targets household chores; over 10,000 units had reportedly been reserved at launch, suggesting real demand exists even before ROI math is proven.
• Industrial: Agility’s Digit at ~$250,000, almost always leased via RaaS, is priced for facilities that can show clear labor-hour displacement.
• Future consumer/industrial hybrid: Optimus’s $20,000–$30,000 long-term target is aspirational and unbacked by a retail price today — it should be read as a manufacturing goal, not a current market fact.

If you’re new to how these robots are actually built and controlled, our pillar guide on what a humanoid robot is and how it works is the right starting point, and our head-to-head on how Figure’s roadmap compares with Tesla’s Optimus goes deeper on two of the highest-profile programs referenced above. For the fuller landscape of players racing toward a commercial humanoid, the Humanoids: The Race for the Robot Body hub collects our full coverage of this cluster. And if terms like BOM, RaaS, or NdFeB need unpacking, our glossary defines them in plain language.