When we covered Agility Robotics’ SPAC on June 25, the pitch was 65,000 logged hours across nine customer sites, more than $300 million in Digit v5 orders, and a $2.5 billion valuation. One number was missing from all of it: revenue.
On September 7, Agility filed its S-4 with the SEC and the number arrived. Net sales in 2025 were $1.8 million.
What the books say
Against that $1.8 million, Agility booked a $140 million operating loss. Operating expenses ran $111 million, up from $71 million in 2024, and the company burned roughly $100 million in cash over the year, per The Robot Report’s read of the filing. Set the $2.5 billion valuation against 2025 sales and the multiple is about 1,400 times revenue. The filing itself is on EDGAR.
The S-4 also tightens a number we reported loosely in June. The $300 million-plus in multi-year Digit v5 orders comes from one customer. Not a book of thirty. One.
A SPAC is the vehicle a company picks when it wants to be priced on projections rather than history. That is not a scandal; it is the product. But the registration statement still forces the books open, and Agility’s are now open.
The rate card is the real disclosure
The loss figures will get quoted for a week. The pages worth keeping are further in, where Agility spells out how Digit v5 is sold and for how much.
Under Robots-as-a-Service, Agility keeps the robot and charges roughly $8,500 a month — about $102,000 a year — covering the machine, the software and maintenance, plus a one-time $25,000 deployment fee. Over an assumed five-year life, that is about $535,000 of revenue per robot.
Under outright purchase, the customer pays about $200,000 up front, a $20,000 deployment fee, and roughly $36,000 a year for software and maintenance. Five years, about $400,000.
This is the first published price on a humanoid doing warehouse work in the United States. Every argument about whether a robot beats a wage has until now been conducted without one. The number is in an SEC filing. Anyone can run it.
The headcount math
We used a conversion in our July 30 piece: a full-time US warehouse worker logs roughly 2,000 hours a year, so Digit’s entire 65,000-hour fleet history amounts to about 32 human work-years. That is the operating record underneath $2.5 billion.
Run the conversion the other direction. Digit v5 is rated to run as much as 22 hours a day, a spec we cited on July 1. Discount it hard, to 60% utilization, and one robot still clears about 4,800 hours a year, or a bit over two human work-years. Spread $102,000 across two-plus work-years and you land in the low $40,000s per work-year displaced.
Read that however your own floor reads it. It is not a rout. It is also no longer absurd. What changed on September 7 is narrower and more durable than either: the comparison no longer requires imagination. It requires a calculator.
800, 7,000, 25,000
The S-4 lays out a deployment curve. Around 800 Digit v5 units in 2027. Seven thousand by 2030. Up to 25,000 by 2035.
Multiply by the rate card. One thousand robots is about $102 million in annual subscription revenue; 5,000 is roughly $510 million; 10,000 is about $1 billion; 25,000 is about $2.55 billion.
Two ratios fall out. To match the revenue of just 1,000 leased robots, Agility has to grow sales roughly 57-fold from 2025. And the 25,000-robot scenario sits at a bit more than 1,400 times 2025 revenue, which is exactly today’s valuation multiple. That is not a coincidence. That is the arithmetic that produced the $2.5 billion: price the company on the 2035 fleet, then look back at this year’s income statement.
Agility says plainly in the filing that the unit economics rest on assumptions, and that real contracts, timing, pricing and utilization may all differ. Worth copying down.
The cohort
In mid-August, Unitree listed on Shanghai’s STAR Market, closed its first day up 460%, and raised 6.1 billion yuan, about $905 million. We wrote up the turnover on that day.
Both companies are now public, but they are exposed to different things. Unitree sells machines it already ships in volume. Agility sells five years of future rent. Figure, Apptronik and 1X carry comparable valuations with closed books, so nobody can compute a revenue multiple for them yet. Agility lifted the lid first and caught the whole sector’s arithmetic in the process.
Who is exposed, and when
Digit does specific work: moving totes, machine tending, sortation, stocking. The nine sites sit inside Schaeffler, GXO, Toyota’s Canadian plant, Mercado Libre and Amazon. This is not pressure on “warehouse workers” as a category. It is pressure on the handful of stations in a building that are the most repetitive, the hardest to schedule, and the hardest to staff.
On timing: 800 units in 2027, spread across dozens of facilities, is a few stations per site. That is not a labor event. The number to watch is 7,000 by 2030. At that volume a single large distribution center stops running a pilot and starts carrying Digit as a standing line on the shift plan.
The exposure mechanism has also changed shape. It used to be “a robot may replace you someday.” It is now “your station has a published price.” Once a workstation carries a dollar figure attached to a machine, its headcount becomes an arithmetic exercise in a budget meeting you do not attend.
Then there is the concentration. That $300 million order sits with one customer, which means the first real deployment wave lands inside one company’s network rather than spreading evenly. Whichever network that is, its stations get re-planned first.
Last, watch whether $8,500 holds. Hardware prices move one direction. At low-$40,000s per work-year the comparison is still a fight. Take 30% off the monthly rate and it stops being one.
