# The First Open Books on a US Humanoid

> Markdown twin of https://roboterminal.app/posts/agility-s4-us-humanoid-books · https://roboterminal.app/posts/agility-s4-us-humanoid-books.md

**Published:** 2026-09-08

Agility's S-4 is the first public ledger from an American humanoid company, and it puts 65,000 hours of robot work on the same page as $1.8 million in 2025 sales and a $2.5 billion price.

Agility's S-4 is the first public ledger from an American humanoid company, and it puts 65,000 hours of robot work on the same page as $1.8 million in 2025 sales and a $2.5 billion price.

**Companies:** [Agility](https://roboterminal.app/companies/agility)

**Tags:** humanoids, warehouse, funding

On September 4, 2026, [Agility Robotics](https://roboterminal.app/companies/agility) filed an S-4 with the SEC as part of its merger with Churchill Capital Corp XI (https://www.sec.gov/Archives/edgar/data/2074973/000121390026097764/ea0297114-04.htm). Most of the document is standard merger boilerplate. Risk factors, share counts, lockup terms. But near the back is something no American humanoid company has put in public before: an income statement with a real revenue line on it.

Until this filing, the US humanoid market ran on three inputs. Demo videos and private round marks. And warehouse footage of a bipedal robot moving totes at a Schaeffler plant or a GXO site. The word "commercial" got attached to all of it, and commercial got read as something close to revenue. Figure, Apptronik, 1X and the rest still trade on that reading, because as private companies none of them has to disclose otherwise.

Agility now does. The books show a deployment company with 2025 net sales of $1,781,967, an operating loss near $140 million, and a pre-money equity value of $2.5 billion (https://www.therobotreport.com/agility-robotics-reports-18m-revenue-ahead-of-humanoid-spac/). Divide the second number into the first and the multiple comes out around 1,400 times trailing revenue. The operating hours are real, and so are the multi-year orders, with conditions attached. The revenue does not yet carry the price. This filing is the first chance to see exactly how far apart those two columns sit, and to read them without a walk video playing in the corner.

## The books, without the walk video

Start with what the filing reports. Net sales were $1,781,967 in 2025 against $310,301 in 2024. That is roughly 5.7x growth, which sounds strong until the base is put next to the company's own pricing: the whole of 2025 revenue is a bit more than three Digits on a five-year RaaS contract. The operating loss was about $140.2 million, up from about $69.7 million the year before. Cash used in operations came to $99.8 million. Research and development ran near $91.6 million, and selling, general and administrative near $45.8 million (https://www.therobotreport.com/agility-robotics-reports-18m-revenue-ahead-of-humanoid-spac/).

GeekWire's earlier read of the preliminary numbers had operating expenses climbing to about $111 million in 2025 from about $71 million in 2024, with roughly $100 million of cash burn (https://www.geekwire.com/2026/digit-maker-agility-robotics-to-go-public-in-2-5b-deal-heres-what-the-filings-say-about-its-finances/). The S-4 lands in the same order of magnitude. Nothing in the S-4 softens the picture.

Those are the numbers. None of them is shocking for a company building bipedal robots and a factory to make them. What matters is that they exist in public at all. Figure has raised billions in private rounds and has never shown a revenue line. Apptronik and 1X are in the same position. Every valuation in the US humanoid category has been set by lead investors looking at pitch decks, and the rest of the market has inferred a P&L from the size of the round. Agility's filing breaks that pattern and gives the category its first real denominator.

The one other public humanoid ledger belongs to Unitree, which listed on the Shanghai STAR Market in mid-August 2026, raised about 6.1 billion yuan, roughly $905 million, and surged on its first day of trading (https://www.therobotreport.com/agility-robotics-reports-18m-revenue-ahead-of-humanoid-spac/). Unitree is a different animal, with a quadruped and component business that predates its humanoids by years, so its books are not a clean comparison. The relevant point is timing. A Chinese humanoid company opened its books first and the market bid it up. Agility is the first American one to do it, and the first pure play anywhere to show what a warehouse-humanoid business looks like on paper before scale.

## Hours are not dollars

The commercial story in the S-4 and the company's announcement is built around one figure: Digit has logged more than 65,000 operating hours across nine customer sites (https://www.agilityrobotics.com/content/agility-robotics-to-go-public-through-merger-with-churchill-capital-corp-xi). The customers have names. Schaeffler, which is also an investor. GXO, the contract logistics operator. Toyota Motor Manufacturing Canada. Mercado Libre in Latin America. Amazon was an early investor and has run multiple pilots, and an Agility spokesperson told GeekWire the company looks forward to working with Amazon again on the next-generation, cooperatively safe Digit (https://www.geekwire.com/2026/digit-maker-agility-robotics-to-go-public-in-2-5b-deal-heres-what-the-filings-say-about-its-finances/).

Sixty-five thousand hours is a serious number for a biped. It is roughly seven and a half robot-years of continuous operation, and it means Digit has been doing repetitive tote and machine-tending work in real buildings long enough for the failure modes to show up and get fixed. The current machine stands about 5 feet 9 inches and lifts up to 35 pounds. Digit v5 adds swappable hands, a 50-pound lift, and a safety architecture meant to let it work alongside people without cages or light curtains. The company claims up to about 20 hours a day on repetitive tasks.

Then there is the backlog. The announcement cites more than $300 million in multi-year Digit v5 orders. The S-4 language around that figure is careful: the orders are subject to contractual milestones and are "not a measure of current period revenue." GeekWire pulled the detail out of the fine print. The $300 million is a three-year contract for 1,000 robots from a single undisclosed customer. One buyer, one machine that has not yet shipped in volume, and a set of milestones the filing does not fully describe. Behind that, the company says more than 30 potential customers are evaluating large-scale deployments.

Put the hours and the orders next to the revenue line and the ledger gap is the whole story. **Operating hours prove a robot can work. They do not prove a business at this valuation.** Sixty-five thousand cumulative hours sit next to $1.8 million of 2025 sales, and even crediting every one of those hours to a single year, recognized revenue comes to under $30 per operating hour before asking what it cost to keep the robots running. The $300 million order, if it converts on schedule, would be about 170 times last year's revenue. Whether it converts is a question about one customer's site readiness and one robot's manufacturing ramp, and the filing is honest that neither is done.

## Two price tags for the same machine

The S-4's unit economics section is where the company shows its hand on how it plans to close that gap, and it offers two ways to buy the same robot (https://www.therobotreport.com/agility-robotics-reports-18m-revenue-ahead-of-humanoid-spac/).

Under Robots-as-a-Service, Agility keeps the machine on its own balance sheet and charges about $8,500 a month, roughly $102,000 a year, plus a deployment fee near $25,000. Over an assumed five-year life that comes to about $535,000 of revenue per robot. Under the ownership model, a customer pays around $200,000 upfront, about $20,000 to deploy, and something like $36,000 a year for software and maintenance. Five years of that is roughly $400,000. Both figures are company assumptions and the filing says so repeatedly.

The RaaS number is the one that scales the story. At $8,500 a month, 1,000 robots on subscription produce about $102 million a year. Five thousand produce $510 million. Ten thousand, about $1 billion. Twenty-five thousand, about $2.55 billion. Those are Robot Report's arithmetic on the company's own rates, and they line up with the illustrative deployment curve in the S-4: around 800 Digit v5 units in 2027, around 7,000 by 2030, and up to about 25,000 by 2035. The filing flags every step of that curve as an illustration built on assumptions the company cannot promise.

Read the curve against the reported line and the size of the leap is clear.

The illustrative revenue from 1,000 robots on subscription is about $102 million a year. Getting there from the reported 2025 line means growing net sales roughly 57 times, and that is before the curve reaches 7,000 or 25,000 units. Nine customer sites today, a factory built for 10,000 units a year, and a five-year illustration that ends with a fleet larger than most warehouse robot deployments of any kind. **The deployment curve describes a business that the reported statements do not yet contain.** The company is not hiding that. It labels every step an assumption, and the labels are the honest part of the document.

The operator's voice on the same gap comes from Jonathan Hurst, the co-founder, in an interview with the Chronline published August 31, 2026 (https://www.chronline.com/stories/oregon-robotics-company-says-the-industry-needs-a-reality-check-i-really-want-to-pop-that,407333). Hurst put the lifetime cost of a Digit at around $400,000 to buy and operate, and said the company's figures show it running about 20 percent cheaper than employing a person for the same job over five years. He described the company as "just about at that cost," the cusp where a sale is profitable for Agility and still saves the customer money. That is a narrower margin than the walk videos suggest, and Hurst seemed to want it read that way. "There's so much hype and so much storytelling that is not connected to reality," he told the paper. "And I really want to pop that bubble a little bit." The Chronline also cited a Bernstein view that humanoid prices need to fall 30 to 50 percent before the category is commercially viable. Set that analyst estimate next to Hurst's 20 percent advantage and the cusp looks like a cusp. It is not a cliff in either direction.

## What the $620M is for

The merger is expected to deliver more than $620 million in gross proceeds (https://www.agilityrobotics.com/content/agility-robotics-to-go-public-through-merger-with-churchill-capital-corp-xi). About $420 million of that is the Churchill trust, and the filing's own condition on that figure is the one to remember: it assumes no redemptions. The other $201 million is a PIPE priced at $10 a share and led by Foxconn (https://techcrunch.com/2026/06/24/agility-robotics-plans-to-go-public-via-spac-in-a-2-5b-deal/). Existing Agility shareholders roll all of their equity and sit under a 180-day lockup. Before this deal the company had raised more than $390 million in private equity from a list that includes DCVC, NVIDIA, Amazon, SoftBank Vision Fund 2, Foxconn, Schaeffler, Abico and Playground Global (https://www.geekwire.com/2026/digit-maker-agility-robotics-to-go-public-in-2-5b-deal-heres-what-the-filings-say-about-its-finances/). The SPAC roughly doubles the capital that has ever gone into the company, in one transaction.

Where it goes is easy to read off the company's own materials. RoboFab in Salem, Oregon is designed to build up to 10,000 Digits a year, and the company says about 75 percent of the robot's parts are sourced in the United States (https://www.agilityrobotics.com/). Digit v5 carries the cooperative safety architecture that lets the robot share floor space with people instead of working behind barriers, and Agility is a launch partner for NVIDIA's Halos safety program for physical AI. Agility Arc is the cloud platform that schedules and monitors the fleet, which is the software layer a RaaS business lives or dies on once robots are in dozens of buildings instead of nine. Headcount is around 400, with a new software office planned in Fremont, California for roughly 200 people and another near Carnegie Mellon in Pittsburgh (https://www.chronline.com/stories/oregon-robotics-company-says-the-industry-needs-a-reality-check-i-really-want-to-pop-that,407333).

At the 2025 rate of cash use, $99.8 million from operations, the proceeds cover roughly six years. That arithmetic is misleading in a specific way. A company that plans to ship 800 robots in 2027 and keep most of them on its own balance sheet under RaaS will spend far more cash than one that shipped a handful in 2025, because every subscription robot is capital Agility fronts and recovers over five years. The money is runway for a factory and a fleet. **The proceeds buy the capacity to build the robots, and the fleet math still depends on customers the filing does not name.** Foxconn's lead on the PIPE says a manufacturer believes in the manufacturing. Product-market fit at scale is a different claim, and the S-4 does not make it.

## Objection docket

The strongest case against reading anything into this filing goes like this. SPACs have carried plenty of companies to public markets at prices the next few years did not support. The trust figure assumes no redemptions, a condition many SPACs have failed. The $300 million backlog is one customer, one contract, one robot generation that has not shipped in volume, and the filing itself says it is not a measure of revenue. The multiple is 1,400 times a trailing number smaller than a single Digit factory's annual parts bill. Every forward figure in the document is labeled an illustration. On that reading, the S-4 is a marketing document with financial statements attached, and the statements are the part nobody is meant to look at.

Most of that is fair, and some of it is exactly what makes the filing useful. The complaint that the backlog is contingent is true because the company said so in the fine print, which is more than any private peer has disclosed about any order. The complaint about the multiple is true because the revenue line is real and public, which is why the multiple can be computed at all. Figure's private valuation cannot be divided by anything. The SPAC structure deserves its skepticism, and redemptions are a real risk to the trust, but the structure is also why the public statements exist. A Series F would have produced a press release and a number.

The one objection that stands without an answer is the concentration. A single undisclosed buyer accounts for the entire disclosed backlog, and the filing does not describe the milestones in enough detail to judge the odds of conversion. Hurst's 20 percent cost advantage and Bernstein's 30 to 50 percent price gap are two estimates of the same unknown, and only shipping robots will settle it. The company's own co-founder went to a regional paper to say the industry's storytelling has outrun reality. The books he just published support him. That is the reason to read them closely rather than dismiss them, and it points to a short list of numbers that will show which way the ledger moves.

## Three numbers to watch

First, recognized revenue against the RaaS fleet math. The company's own rate is about $102,000 a year per subscription robot, so every 100 Digits on RaaS should show up as roughly $10 million of annual revenue once they are deployed and running (https://www.therobotreport.com/agility-robotics-reports-18m-revenue-ahead-of-humanoid-spac/). The 2027 illustration of around 800 units, if all of them went out on subscription, would imply an annualized run rate in the low $80 millions. Quarterly reports will show how far the actual line sits from that. A large gap between the number of robots the company says are in the field and the revenue they generate would mean pilots, discounts or ownership sales dominating the mix, and each of those tells a different story than the RaaS curve does.

Second, whether the $300 million converts. The contract covers 1,000 robots over three years for one customer (https://www.geekwire.com/2026/digit-maker-agility-robotics-to-go-public-in-2-5b-deal-heres-what-the-filings-say-about-its-finances/). The signs to look for are recognized sales tied to Digit v5 shipments, deferred revenue building on the balance sheet, and any change in how the company describes the milestones. If the number drops from the backlog language or the customer is named, either event is informative. So is silence past the point when v5 units were supposed to be moving.

Third, cash burn against RoboFab output. Operations used $99.8 million in 2025 with a handful of robots shipping. The factory is built for 10,000 a year. Burn per robot delivered is the number that says whether the Salem plant is filling up or sitting idle, and it is the number that determines how long $620 million lasts. A rising burn with a rising unit count is a company building a fleet. A rising burn with a flat unit count is a company paying for a factory it cannot yet use.

Agility filed first, so its books set the template. The other US humanoid companies are still raising private money on hours, videos and orders that have never been audited. At some point one of them will go public and the same three numbers will appear on the same kind of page. The question the Agility filing raises for the whole category is whether the shape rhymes: real hours, a real backlog with conditions attached, a real factory, and a revenue line that has not yet caught the price. The first open set of books says that is what a US humanoid company looks like on paper in 2026. The next set will say whether it was a stage or a pattern.

## Sources

- [Agility Robotics S-4 filing](https://www.sec.gov/Archives/edgar/data/2074973/000121390026097764/ea0297114-04.htm)
- [Agility Robotics merger announcement](https://www.agilityrobotics.com/content/agility-robotics-to-go-public-through-merger-with-churchill-capital-corp-xi)
- [The Robot Report](https://www.therobotreport.com/agility-robotics-reports-18m-revenue-ahead-of-humanoid-spac/)
- [GeekWire](https://www.geekwire.com/2026/digit-maker-agility-robotics-to-go-public-in-2-5b-deal-heres-what-the-filings-say-about-its-finances/)
- [TechCrunch](https://techcrunch.com/2026/06/24/agility-robotics-plans-to-go-public-via-spac-in-a-2-5b-deal/)
- [The Chronicle](https://www.chronline.com/stories/oregon-robotics-company-says-the-industry-needs-a-reality-check-i-really-want-to-pop-that,407333)

## Citation

```
"{quote}"
Source: Roboterminal · {url}
```

---

Source: Roboterminal · https://roboterminal.app/posts/agility-s4-us-humanoid-books.md