Agility Reports Hours, Not Units — And That's the Tell
65,000 operating hours across nine facilities is the only humanoid metric that can't be staged. It also reveals how early the whole industry still is.
Humanoid robotics has a disclosure problem. Every company in the category reports the metric that flatters it most — units shipped, robots per hour off the line, valuation, a demo video shot in one take if you are lucky. Almost none of it survives contact with the question a buyer actually asks, which is whether the machine works for a full shift without someone standing next to it.
Agility Robotics reports a different number. As it moves toward a public listing, the figure it puts forward is more than 65,000 operating hours for Digit, accumulated across commitments at nine customer facilities. That is an unusual thing to lead with, and it is worth understanding both why it is the honest metric and why it is smaller than it sounds.
Do the division
65,000 hours is roughly 7.4 robot-years of continuous round-the-clock operation. Spread across nine facilities, it averages about 7,200 hours per site — under a single year of 24/7 running, and in practice considerably less than that per robot, since sites run multiple units on partial shifts.
That is a real number and a small one. It is also, as far as anyone outside these companies can verify, the largest disclosed body of commercial humanoid operating experience in existence. Both facts are true simultaneously, and holding them together is the entire analytical exercise in this sector right now.
The reason hours matter more than units is that hours are the only metric that captures failure. A robot that ships is a manufacturing achievement. A robot that accumulates hours is one that kept working after the integration team went home — through edge cases, floor changes, lighting shifts, the specific chaos of a real logistics facility. Units shipped tells you what a company built. Hours tell you what a customer tolerated.
It is also the metric that is hardest to inflate. You can stage a demo. You can book an order. You cannot fake a cumulative runtime figure across nine third-party sites where the customers know what their own uptime was.
The listing, and what it is priced on
Agility has a definitive business combination agreement with Churchill Capital Corp XI, at a $2.5 billion pre-money equity value, expected to close before the end of 2026 under the ticker AGLT. The transaction is structured to deliver more than $620 million in gross proceeds, including a $200 million PIPE at $10.00 per share, with all existing Agility shareholders rolling their equity.
The order book behind it: more than $300 million in multi-year Digit v5 orders, subject to contractual milestones, against a customer pipeline of more than 30 companies, with manufacturing designed for up to 10,000 units annually.
Two of those phrases deserve weight. "Subject to contractual milestones" means the $300 million is contingent revenue, not backlog in the sense a manufacturing investor would normally use — it converts if Agility hits delivery and performance gates, and not otherwise. And "designed for up to 10,000 units annually" is a capacity statement, not a production statement. Designed capacity is what a factory could do; it tells you about ambition and capital commitment, not output.
That every existing shareholder is rolling equity rather than taking cash off the table is the more encouraging structural detail. It is the closest thing to a conviction signal a SPAC structure offers.
The Silicon Valley half
In mid-July, Agility opened a 60,000-square-foot facility in Fremont as its Bay Area physical-AI hub — the software and capability site where engineering teams train and test the models that let Digit acquire new skills and handle more complex tasks in customer environments. The company plans to hire close to 200 people across AI/ML software engineering and field operations.
The split is the strategy made legible. Manufacturing and the robot itself stay where they were; the learning problem moves to where the machine-learning labour market is. It is an admission that the bottleneck has migrated. Agility does not appear to believe its constraint is building Digits. It believes the constraint is teaching them, and it is buying its way into the talent pool that does that work — in the same city where Tesla is standing up Optimus production, which will not make the hiring cheaper.
Two theories of the same market
Set this against Unitree, which is finalising a Shanghai STAR Market listing at roughly $6.2 billion, on 2025 revenue near $250 million, $41 million of net profit, and more than 5,500 humanoid units shipped.
These are not the same business wearing different flags. Unitree is a profitable hardware manufacturer that has driven cost down and volume up, selling into research, education and early industrial buyers, with humanoids crossing half of revenue. Agility is an unprofitable systems company selling a narrower proposition — a bipedal robot doing specific logistics work inside a customer's operation, priced on outcomes rather than units — and it is going public on deployment evidence rather than earnings.
The market will eventually decide which metric it pays for: units shipped and margin today, or hours logged and the operational data that comes with them. Unitree's answer scales the way manufacturing scales. Agility's answer scales only if the hours compound into capability faster than the competition can buy their way to the same place.
What the number needs to become
The honest read on 65,000 hours is that it is a strong disclosure about a weak base. It is the best evidence anyone has offered that humanoids do productive work in commercial settings, and it represents an amount of operating experience that a single mid-sized warehouse's conveyor system exceeds in a quarter.
The figure to watch is not the total. It is the rate. If the next update reports a number that has doubled without the facility count doubling, Digit is getting more useful per site and the deployments are deepening. If the hours grow only in step with new logos, the company is selling pilots, and pilots have a well-documented habit of not renewing.
Agility deserves credit for publishing a metric that can indict it. Very few of its competitors have offered anything that could.