Agility's $125,000 robot and the gears that set the price

A price list written by the bill of materials

Fact On Sep 30, 2026, Agility Robotics filed an amended S-4 for its merger with Churchill Capital Corp XI and put numbers on what a humanoid costs. The company said it had cut the bill of materials for Digit v4 to about $125,000 per robot. For Digit 5 it models a BOM of about $150,000 at commercial launch, with targets of about $75,000 at 1,000 robots a year and about $30,000 at 10,000 a year.1

Fact The same filing sets out two ways to buy. Under its Robots-as-a-Service model a customer pays a $8,500 monthly fee covering software and maintenance plus a one-time deployment fee of about $25,000. Under the ownership model the customer pays about $200,000 upfront, about $20,000 to deploy and about $36,000 a year for software and maintenance. Both models assume a five-year robot life.1

Fact Agility's model shows why the subscription is the preferred route. It expects a single RaaS deployment to bring in about $500,000 over five years and to recover its invested capital in less than a year, against about $400,000 for a robot sold outright. It also books about $15,000 a year per robot to deliver software and maintenance and about $15,000 to deploy each one.1

Our read Those prices only work if the BOM falls on schedule, and Agility's own risk factors say there is no assurance it will. Anyone selling a humanoid today is selling against a cost curve they do not fully control, and the steepest part of that curve sits inside the joints. A GTM team that cannot explain where its BOM goes from here will struggle to defend a five-year subscription price.

Actuators are most of the robot

Fact When Hyundai Mobis announced at CES on Jan 7, 2026 that it would supply actuators for Boston Dynamics' next-generation Atlas, it said actuators account for more than 60% of a humanoid's material cost. Each actuator stacks a motor, a reducer, encoders and bearings, and linear joints add a screw.2

Fact The same release tied the deal to Hyundai Motor Group's $26 billion U.S. investment plan, which includes deploying tens of thousands of robots over the next few years and a North American robot factory with annual capacity of 30,000 units. Mobis said it plans to build a large-scale actuator manufacturing system.2

Our read Hyundai's setup is the template incumbents will copy: the carmaker owns the robot maker, a group parts company builds the actuators, and the group's own plants are the first customer. A startup selling against that stack cannot win on hardware cost, so it has to win on deployment speed and on the tasks it handles.

Fact Costs have been falling faster than analysts expected. Goldman Sachs Research wrote in February 2024 that the cost to build a humanoid had dropped from an estimated $50,000 to $250,000 a year earlier to between $30,000 and $150,000. Its analysts had expected a decline of 15–20% a year and saw 40%, which they attributed to cheaper components, more supply chain options and better designs.3

Fact Goldman's base case at the time was more than 250,000 humanoid shipments in 2030, almost all for industrial use. Its analysts wrote that they expected further cost reduction in the coming years.3

Our read Put the Mobis and Goldman figures together and the commercial logic is simple. If actuators are more than half the bill, then the reducer and screw inside each actuator are the line items that move the robot's price. A buyer's procurement team will figure this out within a quarter, so the vendor should be the one to raise it.

Two reducer camps: Japanese incumbents and Chinese challengers

Fact Nabtesco says it holds about 60% of the global market for precision reduction gears, a business it has run for more than 20 years. Harmonic Drive Systems, the Japanese strain-wave gear maker, booked about ¥1.3 billion of humanoid-related orders in its fiscal second quarter and expected about ¥2.5 billion for the year to March 2026, Investing.com reported on Nov 19, 2025. The company counted more than 15 potential humanoid customers, and a Macquarie analyst called its ¥10 billion humanoid sales target for the following year aggressive.45

Fact On the Chinese side, Leaderdrive reported 2025 revenue up 47% to 570.7 million yuan and net profit more than doubled to 124.4 million yuan, VnExpress reported in April 2026. J.P. Morgan estimates it holds 30% to 40% of China's harmonic reducer market, and its domestic customers include Agibot and UBTech.6

Fact Harmonic Drive's parent-level sales were about ¥8.2 billion in the quarter, close to breakeven, with marginal profit above 50%, according to the same report. That means each extra order drops heavily to profit, which explains why the company is courting humanoid makers.5

Our read Humanoid orders are still small next to these companies' industrial robot business, which gives OEMs some leverage today. That leverage fades once one or two humanoid makers reach volume and lock up capacity. The supplier choice also sets tariff exposure: a Japanese reducer and a Chinese one carry different duty risk into a U.S. contract, and a buyer signing a five-year deal will ask which one is in the robot.

Screws, and the Tier-1s buying their way in

Fact Schaeffler agreed in August 2022 to buy Ewellix for about €582 million, excluding about €120 million of net debt, and closed the deal on Jan 3, 2023. Schaeffler described Ewellix as a market leader in electromechanical actuators, lifting columns and roller screws.78

Fact It has since turned that position into humanoid supply deals. In November 2025 it agreed to supply planetary gear actuators to Neura Robotics and to bring a mid-four-digit number of Neura humanoids into its plants by 2035. In April 2026 it agreed to supply Hexagon Robotics with rotary actuators built around strain-wave or planetary gears and to deploy at least 1,000 of its humanoids within seven years. The same month it became core joint-module supplier to China's ROKAE, through a Taicang unit founded in February 2024 whose portfolio includes planetary roller screws and precision reducers.91011

Fact Schaeffler is also on the buying side. Agility's filing lists Schaeffler among Digit's customers, alongside GXO, Toyota Motor Manufacturing Canada and Mercado Libre.1

Our read Schaeffler's deals pair supply with purchase: it sells actuators to a humanoid maker and commits to buy that maker's robots. For a startup, that is a reference customer and a parts source in one contract. It also means Schaeffler sees each partner's cost structure, so the startup gives up some pricing privacy in exchange.

Few suppliers, thin leverage

Fact Agility's filing says it depends on a limited number of suppliers, including single- or limited-source suppliers, for actuators, battery cells, sensors, vision systems and computing hardware. It also says it has less negotiating leverage with suppliers than larger companies and that tariffs and Buy American rules could raise input costs.1

Fact Agility also says it does not see cost reduction as mainly a function of volume, and credits engineering changes and supplier engagement for most of the savings. It describes a potential path to product gross margins above 70% as the business scales.1

Our read This is the cost wall in practice. A humanoid maker shipping hundreds of robots a year is a small account for Nabtesco or Schaeffler, and the reducer price it pays will look very different from Tesla's or Hyundai's. Startups that win on cost will do it through design choices, such as fewer high-ratio joints or a second-source reducer, more than through purchasing power.

What the joint supply chain means for pricing and sales

Our read The RaaS structure in Agility's filing is partly a hedge against the cost wall. At $8,500 a month the vendor keeps the hardware and its depreciation, and the customer pays for uptime. If the BOM falls from $150,000 toward $75,000, the vendor keeps the gain without repricing every contract. That is a strong reason to lead early deals with a subscription and keep outright sales for buyers who insist.

Our read The service line carries the joint risk too. Reducers and screws are the parts that wear, and a vendor who prices maintenance into a monthly fee owns that risk. Before quoting a multi-year contract, a GTM lead should get from engineering the expected service interval for each joint and the cost of the spare actuator stock behind it.