On 15 September, Agility Robotics unveiled Digit 5, the fifth generation of the humanoid that already works in GXO warehouses and Schaeffler plants. Most coverage led with the spec sheet: a 50 lb payload, 40% more than the last model, a 9-minute charge, 20+ productive hours a day.
The line that actually matters for money is a different one. Digit 5 is Agility's first humanoid built to work right next to people with no physical safety barrier. Its predecessor, Digit 4, had to operate inside a fenced area. This one does not.
That sounds like a safety footnote. It is closer to a pricing announcement. In industrial automation, the robot is often the cheaper part of the project, and the cage, with everything the cage forces on your building, is a large share of what you really pay. A humanoid that has to live behind a fence throws away the one advantage a humanoid is supposed to have, which is fitting into spaces designed for humans. Here is why the cage matters so much, what Digit 5 does and does not change, and what the $2.5 billion SPAC behind it means if you are thinking about buying robots or buying the stock.
What Agility actually launched
The facts from the launch, including the dates that most headlines skipped.
| Detail | Digit 5 |
|---|---|
| Payload | 50 lb (22.7 kg), up 40% on Digit 4 |
| Runtime / charge | 90 minutes of work, 9-minute charge, a 10:1 ratio |
| Productive hours | 20+ per day |
| Height / weight / reach | 5'11", 284 lb, 7.2 ft reach |
| Safety | AI human detection on multiple sensors, visual and audio intent cues, independent safety controller that avoids, stops, or sits the robot down |
| Standards | First humanoid through an independent OSHA field evaluation on a production line; Agility is contributing to ANSI/A3 R15.108 and ISO 25785-1, the first international humanoid safety standard |
| Orders | $300M+ in multi-year contracts (as of May 2026), subject to contractual milestones |
| Track record | 65,000+ operating hours across nine customer sites; 100,000 totes at GXO with roughly 98% accuracy |
| Availability | Early access H1 2027, general availability by end of 2027 |
| Factory | RoboFab, Salem, Oregon, designed for 10,000 units a year; about 75% of parts sourced in the US |
Read the availability line twice. Digit 5 is announced, not shipping at scale. The orders are real contracts, but they are tied to milestones, and the robot most customers will actually receive lands in 2027. None of that makes the launch less significant. It does tell you what kind of significance it has: this is a statement about where the economics are heading, not a change to anyone's warehouse this quarter.
Why the cage costs so much more than the fence
To understand why removing a barrier is a money story, look at what an automation project really costs. Integration specialists are consistent on one number: the robot hardware is typically only 25 to 40 percent of the total investment needed to get a working cell running reliably. The other 60 to 75 percent is integration: safety systems, fixtures, controls, floor work, and the engineering time to tie it together.
The fence itself is the cheap, visible part of that. Safety hardware for a fenced cell commonly adds $5,000 to $15,000, and facility work such as power, floor preparation and compressed air can add $10,000 to $50,000 more. The expensive part is what the fence forces you to do.
| Hidden cost of a caged robot | What it means in practice |
| Floor space | The robot's zone is reserved, even when it is idle; the aisle around it is lost |
| Process redesign | Work has to be routed into and out of the cage, usually through conveyors or handoff stations |
| Safety engineering | Interlocks, light curtains, laser scanners, safety PLCs, validation and sign-off |
| Rigidity | Moving the robot to a different task means moving the cage, which means a new project |
| Exception handling | When something goes wrong, a person has to stop the cell, enter, fix, and restart |
Integrators often put it bluntly: once fencing, safety controllers, extra floor space and longer integration timelines are counted, a traditional fenced cell can cost three to four times an equivalent collaborative solution. That comparison comes from robot arms, not humanoids, but the logic carries over, and for humanoids it bites harder.
The whole commercial pitch of a humanoid is that you do not redesign your building for it. It walks your existing aisles, lifts from your existing shelves, and hands totes to your existing conveyors. Put that humanoid in a cage and you have bought a very expensive, relatively slow industrial arm, and then paid the full integration bill anyway. The fence was not a minor safety requirement sitting on top of the business case. For a humanoid, it quietly cancelled a large part of the business case.
The exception problem, and why cage-free fixes it
There is a second reason the cage matters, and it is hidden inside a number Agility is rightly proud of. Digit has moved 100,000 totes at GXO with roughly 98% accuracy.
That is a strong result for a humanoid in live operations. But run the arithmetic the way an operations manager would. Two percent of 100,000 is about 2,000 exceptions: totes that need a person to look, fix, or redo. In a caged cell, each exception is a small shutdown. A human walks over, the cell stops, the worker enters, fixes it, leaves, the cell restarts. Multiply that by a few dozen a shift and the stoppages become a real line item.
In a cooperative setup, the person who fixes the exception is already standing nearby, and the robot simply slows, steps aside or waits while they work. Exception handling goes from a stop-and-restart event to an interruption measured in seconds. The accuracy number does not change. What it costs you does.
This is the underrated truth about robots in 2026: most real deployments are not full replacements, they are mixed teams where the robot does the repetitive volume and the human handles the edge cases. The cheaper it is for those two to share a space, the more of the robot's work you actually get to keep.
The hourly math, honestly
Now the number everyone wants. Agility mostly does not sell Digit outright; buying one has been estimated at around $250,000. It rents robots through Robots-as-a-Service, which on the figures we have tracked works out to roughly $30 per robot-hour, fully managed. Investor models put fully burdened warehouse labour at about $30 an hour as well.
Put those side by side and a lot of lazy robot coverage falls apart.
| Human team | Digit on RaaS | |
| Cost per working hour | ~$30 fully burdened | ~$30 per robot-hour |
| Hours covered per day | One person per shift | 20+ hours from one robot |
| Hiring, training, turnover | Constant, and expensive in warehouses | Not your problem |
| Peak season | Temporary hiring, often late and short-handed | Rent more robots, return them after |
| Absences and no-shows | Real, unpredictable | Charging and maintenance windows, predictable |
| Exceptions and judgement | Handled well | Needs a human nearby |
On a straight hourly basis, a warehouse humanoid on RaaS is not cheaper than a person. It is roughly the same price. Anyone who tells you Digit halves your labour bill today is doing the math on automotive wages, where Figure's roughly $25 an hour against $50 to $70 of human labour genuinely does come close to half, not on warehouse wages.
So where is the saving? It is in everything the hourly rate leaves out. One robot covering 20 hours replaces the coordination of multiple shifts. Warehouses have some of the highest turnover in any industry, and every departure means recruiting, onboarding and a productivity dip. Peaks are brutal and temporary labour is unreliable. A robot you can rent for November and return in January is worth a premium on those grounds alone. The case for Digit in a warehouse is not "cheaper labour", it is labour that shows up, does not leave, and scales on demand, at a similar price.
That also tells you where the real economic break comes from. Agility's factory is designed for 10,000 units a year, and hardware costs fall steeply with volume, which is exactly the pressure Chinese makers are putting on the whole category. If the robot-hour drops from $30 toward $20 while human wages keep rising, the calculation stops being about reliability and becomes about raw cost. Cooperative safety is what lets that cheaper robot-hour get deployed without the integration bill cancelling it out.
Safety standards are a sales tool
One more item from the launch deserves attention from anyone who buys or sells robots: the standards work.
Digit 5 is described as the first humanoid to pass an independent OSHA field evaluation on a production line, and Agility is helping write ISO 25785-1, the first international safety standard for humanoids, along with the North American ANSI/A3 standard for dynamically stable mobile robots. It is also pursuing CE marking for Europe and the UK.
That can read like compliance homework. It is actually procurement. A plant manager can believe a robot is safe and still be unable to deploy it without a document their insurer, their safety officer and their works council will accept. Standards are how "we think it is safe" becomes "we are allowed to put it next to employees". Whoever shapes the standard gets to define what "safe enough" means, and gets there first. It is the same dynamic that makes Neura win on procurement approval in Europe rather than on specs.
The SPAC behind it
Digit 5 does not arrive alone. In June, Agility agreed to go public by merging with Churchill Capital Corp XI, at a $2.5 billion pre-money equity value, under the expected ticker AGLT.
| Deal term | Figure |
| Announced | 24 June 2026 |
| Valuation | $2.5B pre-money equity value |
| SPAC trust | $420M, assuming no redemptions |
| PIPE | ~$200M at $10 a share, led by Foxconn |
| Total gross proceeds | more than $620M |
| Disclosed in the announcement | orders, hours, customers; no revenue or loss figures |
If you are tempted by the stock, keep three things in view. First, "assuming no redemptions" is carrying a lot of weight: SPAC shareholders can take their money back before the merger closes, and in many deals a large share do, which can shrink that $420 million considerably. Second, the headline $300 million in orders is multi-year and milestone-dependent, for a product whose general availability is late 2027, so it is not the same thing as revenue. Third, the announcement did not give revenue or losses; the details sit in the registration statement, and those are the numbers to read before the story. I laid out the wider pattern of these listings in the 2026 AI SPAC wave.
Foxconn leading the PIPE is the most interesting signal in the table. A contract manufacturer does not usually write that cheque for a financial return alone. It is positioning to build robots at volume, which is a quiet vote that humanoid manufacturing is about to become a real industrial business rather than a lab project.
What this means for you
If you run a warehouse or a plant, Digit 5 does not change your 2026 budget, but it should change how you evaluate every robot proposal in front of you now. Stop comparing the robot's price with a worker's wage. Compare total installed cost, including cage, floor space, safety engineering, process redesign and exception handling, against the hours you actually cover. On that basis, a cooperative robot at the same hourly rate as a caged one is a much cheaper robot. And if a vendor quotes a humanoid that still needs a fence, ask why you are paying humanoid prices for arm economics. The fuller cost comparison is in the warehouse robot ROI breakdown.
If you sell services around robots, the money is moving to exactly the work cages used to hide. Cooperative zones need risk assessments, layout design, workflow mapping between humans and robots, staff training and ongoing safety validation against new standards. That is consulting and integration work, and it is the physical-world version of what the AI labs just discovered when they built consulting arms to close the implementation gap: the product is ready before the customer is. If you understand operations and safety, you are positioned for the cooperative-robotics wave before most integrators retool.
If you invest, separate the technology from the listing. Cooperative safety is a genuine unlock for the whole humanoid category, and it will benefit every maker that gets through the standards process, not only Agility. The SPAC is a specific bet with specific risks: redemptions, a 2027 product, and financials you need to read in the filing rather than in the press release. If you want exposure to the theme without that concentration, the robot boom investing guide covers the broader options, and the humanoid price comparison shows who else is competing for the same buyers.
What to watch
Early access in H1 2027. Whether the cage-free claim holds up in real mixed shifts, and whether customers actually remove fences or keep them "just in case". If the fences stay up, the economics described above do not arrive.
ISO 25785-1. Whoever gets certified first against the first international humanoid safety standard gets an enormous procurement advantage in Europe and with insurers everywhere.
SPAC redemptions and the filings. The redemption rate tells you how much the market believes in the deal. The registration statement tells you revenue, losses and how much of the $300 million backlog converts.
The robot-hour price. At $30, a warehouse humanoid competes on reliability. At $20, it competes on cost. Watch the RaaS rate as RoboFab scales.
The honest take
The cleverest thing about Digit 5 is that it attacks the part of the cost that nobody puts in the headline. Robotics coverage loves payloads, speeds and demo videos, because those are visible. The fence, the lost floor space, the conveyor you had to add to feed the cage and the stoppage every time a tote goes wrong are invisible, and they are where many automation business cases quietly die. A humanoid that can share an aisle with a person turns those invisible costs into much smaller ones, and that matters more than an extra 14 pounds of payload.
It is also worth being clear about what it does not do. It does not make robot labour cheaper than warehouse labour per hour today. It does not ship at scale until 2027. And the stock you might soon be able to buy is a SPAC with the usual SPAC risks attached. Believing in the first thing should not make you careless about the other three.
The pattern that carries beyond this launch is one that shows up across the whole AI economy: the model or the machine is rarely the expensive part. The expensive part is everything you have to rebuild around it. Whoever shrinks that rebuild wins the market, and the missing cage on Digit 5 is one of the clearest examples yet.
So here is the question worth asking of the next robot quote that lands on your desk: how much of that price is the robot, and how much is the cage you have not been shown yet?
This article is general information, not financial advice. Do your own research and consider a licensed professional before making investment decisions.
Sources: Agility Robotics: Digit 5 announcement; Forbes: Agility launches Digit 5, no more safety cages; Agility Robotics: merger with Churchill Capital Corp XI; AMD Machines: total cost of ownership for robotic systems.
