Digit Out of the Cage

The next time you find yourself on EDGAR, consider using my SECSift reader instead!

It turns SEC filings into clean, readable documents. You can search every word, export anything, merge tables, highlight what matters and gray out the rest, and compare against previous filings.

Here’s a video I made recently demonstrating how you can use SECSift to compare EDGAR filings (with a free account):

Not familiar with the reader? Here’s a 3-minute interactive demo (viewable on your computer) and another video I made walking through the entire tool.

🌾 Welcome to StableBread's Newsletter!

Unitree, the Chinese maker of humanoid robots and robot dogs, listed in Shanghai on August 19, opened 629% above its IPO price at a ~$66B valuation, and still closed the day up 460% at a ~$48B market cap on just ¥1.7B (~$240M) of 2025 revenue. That's ~200x sales!

The market isn't just betting on humanoid robots—it's treating them as an inevitability.

Just look at what the biggest names in the space are saying and actively doing:

Wall Street has followed with their own optimistic forecasts:

But the reality today is a humanoid market that barely exists.

The leading companies in the space have only deployed a small number of robots in controlled pilots, mostly behind safety barriers, and every forecast above assumes a humanoid that's capable, safe, and cheap enough to run at scale (which is still largely unproven).

Even the bulls put the real volume in the 2030s. Bank of America's own curve doesn't pass 1.2 million a year until 2030, with industrial and service uses carrying it until then, and Morgan Stanley expects adoption to stay "relatively slow until the mid-2030s."

The question is whether today's handful of pilots can connect to that promised scale, and if you believe they eventually can, what are the best pure-play stocks to buy?

The one investors are flocking to, especially after Unitree's IPO craze, is Agility Robotics, which will soon be the only pure-play humanoid company listed in the U.S.

It's going public via a SPAC led by Churchill Capital Corp XI (Nasdaq: CCXI) at a $2.5B pre-money valuation, and the combined company will trade as AGLT once the merger closes.

CCXI closed at $10.42 the day before the June 24 announcement, peaked at $19.10 on July 2, and now sits at $13.40/share.

Agility makes Digit, a two-legged robot that already does paid work in a handful of U.S. warehouses and factories.

Its management sizes the opportunity at $1 trillion across U.S. manufacturing, distribution, and logistics, and CEO Peggy Johnson said $1.25 trillion by 2032 on the June 24 announcement call, calling it "the TAM that our robot Digit can address."

If you've been following my research, you'd know I've written up three pre-merger SPAC setups in 2026, and the scorecard is mixed:

The key difference with CCXI as a SPAC setup is that it's already jumped pre-merger. Its trust held $10.16 at June 30 and accrues ~$0.09 a quarter, so it’ll hover ~$10.25 by a fall vote. At $13.40/share you're paying a 30.7% premium to that floor, and redeeming would cost you 23.5%.

In this 7.8k-word deep dive, I do the work almost no one else has. I discuss what Agility has today, the safety certifications, the $300M order book and 30-customer pipeline, the actual economics, the geopolitical angle and lobbying, the SPAC terms, and the stock's realistic valuation range from here.

Agility builds Digit, a 5'9", 140-pound bipedal robot with two arms, a head, and a face. The current version, Digit v4, lifts up to 35 pounds, runs ~4 hours on a charge, and docks itself to recharge.

Its job today is simple material handling, meaning it picks up plastic totes and bins and moves them between shelves, conveyors, and the wheeled autonomous mobile robots (AMRs) that already roam large warehouses.

Customers can get Digit two ways:

Agility's own software is Agility Arc, a cloud platform that assigns tasks and monitors the fleet. Arc plugs into warehouse management systems from Manhattan Associates and into AMRs from two vendors, MiR and Zebra.

MiR is the Danish robot unit of Teradyne, the chip-testing equipment maker, and Zebra is the barcode-scanner company that bought Fetch Robotics in 2021 and then sold that AMR business to Skild AI in April 2026.

Most of the rest is outsourced. Ricoh, the office-equipment company, maintains and repairs Digits in the field, and two warehouse-automation integrators, Zion Solutions Group and Tompkins Solutions, install them at customer sites.

One supplier is publicly traded. Ability Enterprise (2374.TW), a Taiwanese camera-module maker in the Abico group, is co-developing vision modules for Digit, has put $10M into Agility with another $10M planned, and holds sales agency rights for Agility's robots in parts of Asia.

All of that infrastructure serves a short customer list. Today it's only five names:

Although Amazon bought Fauna Robotics in March 2026, it's not a direct competitor to Agility since Fauna’s building a 3.5-foot, $50K soft-bodied humanoid for homes and research rather than warehouse work.

Beyond the five, the S-4 says four unnamed customers had joined its Customer Acceleration Program as of May 2026, paying ~$500,000 each for a trial that runs from a proof of technology to a RaaS pilot.

The slide above is the deck's proof that Digit works, and it comes down to four numbers: (1) Nine committed facility deployments, (2) 65,000 hours of operation, (3) 125,000 totes moved, and (4) 98% accuracy.

The S-4 repeats all four as of May 2026 and shows the totes came from two customers, 100,000 at GXO and 25,000 at Schaeffler.

What the slide doesn't show is how few robots are behind those numbers.

In March 2025, CEO Peggy Johnson said Agility expected to ship hundreds of Digits that year. Co-founder Jonathan Hurst now puts that two years out.

He told the Portland Business Journal in August, in an article Churchill filed with the SEC, that the RoboFab factory is "currently producing fewer than 100 a year" and should reach "triple digits" in 2027.

The S-4's own risk factors go further and say Agility doesn't "currently have the facility capacity, employees, or equipment needed to manufacture its products in high volume."

That's the same company whose deck rates the factory for 10,000 a year, which gives you an indication of just how difficult it is to scale a humanoid production line.

Agility spun out of Oregon State University's robotics lab in 2015, founded by professor Jonathan Hurst, Damion Shelton, and Mikhail Jones. Its first product was Cassie, a pair of ostrich-like legs built on DARPA-funded research, sold to university labs.

Digit arrived in 2019 with a torso and arms, and Ford bought the first two production units in January 2020. Playground Global led an $8M round in 2018, and DCVC and Playground led a $20M round in 2020. The $150M Series B in April 2022 brought in Amazon's Industrial Innovation Fund.

The commercial version, Digit v4 with a head and hands, launched in March 2023. That September Agility announced RoboFab, a 70,000 square-foot factory in Salem, Oregon, with a stated capacity of 10,000 robots a year.

Amazon's test began the next month, and the GXO pilot that became its first RaaS contract started near the end of 2023.

Peggy Johnson took over as CEO in March 2024, with Shelton moving to president. Here are the commercial milestones since then:

That Fremont hiring plan is a large add for a company of 393 full-time employees at June 30, and it lines up with the deck's R&D guide jumping from $74M in 2025 to ~$115M in 2026. The S-4 shows the Fremont lease was signed on February 10, four months before the deal, at $27.5M over 10 years.

The S-4 puts gross equity raised at $395.8M through 2025. The Series C round that ran from November 2024 to February 2026 accounts for ~$223M of it ($51.0M of C-1 at $36.86/share, $13.9M of C-2 from converted notes, and $157.9M of C-3 at $66.15/share), and the July 2026 SAFEs add $100M on top.

Interestingly, the 2025 round came after a takeover approach. The Information reported that SoftBank's Masayoshi Son met Agility in late 2024 and discussed buying the whole company for at least $900M.

Agility raised money instead, with SoftBank participating. The Series C-3 round valued the company at ~$1.985B post-money, Churchill's first letter of intent on March 2, 2026 proposed $2.5-3.0B, and the deal signed at the bottom of that range.

So the SPAC values Agility 25.9% above its last round and 2.8x the SoftBank number.

Agility’s bench is strong on people who have commercialized new hardware categories, and light on people who have scaled a factory to thousands of units:

The S-4 also shows who's paid what. Johnson earns a $475,000 base and holds 550,000 options struck at $5.10 from March 2024. Hunter and Chief Legal and People Officer Ana Lang got 350,000 and 300,000 options at $26.31 in November 2025, plus $250,000 retention bonuses each, paid quarterly through 2026.

Johnson has also put her own money in. She bought $2.5M of Series C-3 stock at $66.15/share in August 2025, the same price as the outside investors.

After the close the board has seven seats, six chosen by Agility and one by the sponsor, with co-founder Damion Shelton as chairman.

Three notable people have left:

Wise's exit matters for a reason beyond pedigree. A month after leaving, she told IEEE Spectrum: "I don't think anyone has found an application for humanoids that would require several thousand robots per facility."

That’s the former head of product questioning the density the order book assumes. If no site requires thousands of Digits, growth means winning new sites rather than filling existing ones, and Agility has won nine since Digit went commercial in 2023.

Today, every humanoid working in a factory does so behind physical safety barriers, Digit included. Management is candid that those barriers are what's stopping the company from scaling.

On the announcement call, Hurst called safety "the biggest blocker" and said Digit v5 "will be the scaling moment for humanoid robots."

The reason is regulatory, not technical. A robot arm bolted to the floor can be certified under decades-old standards. A 140-pound machine that balances on two legs can fall, and no published standard covers that risk.

The IEEE's humanoid study group wrote in September 2025 that existing robot safety standards assume a base that "is either fixed or has a statically stable base," and that "none of these specialized safety standards apply to humanoids."

A December 2025 scoping review of 121 studies in the journal Electronics found "a significant lag in standardization and regulation compared to the rapid pace of technological advancement."

Without a standard, a customer's insurer has no basis to cover a humanoid walking among workers. So the robot stays behind barriers, and once it's confined to one spot, it's competing with conveyor belts and fixed robot arms that do the same job for less.

Agility's answer has two parts.

The first is Digit v5, designed to detect people, slow down, set its load on the floor, sit, and power off before anyone can touch it. It also lifts 50 pounds, reaches 7.2 feet, charges at a 10-to-1 ratio instead of 2-to-1, and runs ~20 hours a day.

Two days before the SPAC was announced, NVIDIA named Agility the first partner for Halos, its new safety package for robots. That's why v5's safety computer, the chip that decides when the robot stops and sits, is NVIDIA's IGX Thor.

The second part is the standard itself. The rulebook for dynamically stable robots, ISO 25785-1, is being drafted by an ISO working group whose project leader, per the A3 robotics association, is Agility's own safety engineer Kevin Reese.

Reese is also the lead inventor on Agility's patent for hazard response in a collaborative environment, granted in February 2026. So the company helping write the rulebook also holds a patent on one way to comply with that rulebook.

The deck's wording is generous here. Slide 27 says "New ISO safety standard approved," with a footnote pointing to 25785-1.

What ISO's own site shows is a committee draft, an early stage where member countries comment on the draft before voting on it, and that comment window closed on July 8, 2026.

What was approved is the project to write the standard, which is possibly what the slide means. The standard itself isn't published, and Digit v5 has no third-party certification yet.

The S-4 says it plainly. Its risk factors list ISO 25785-1 among the standards "under development," and the only safety sign-off it claims for Digit is that v4 deployments "have passed OSHA recognized Nationally Recognized Testing Laboratory (NRTL) field evaluations," which is what lets a caged robot run today.

For v5, NVIDIA's Halos lab does a pre-assessment against IEC 61508 and ISO 13849, the functional safety standards insurers recognize, and a separate certifier signs off after that.

Now look at how the timeline has moved:

That December 2026 date is the one to watch.

Any robot with radios needs an FCC equipment authorization before it ships, and the only authorization under Agility's name is for Digit v4, with its last filing in September 2025. When a v5 grant shows up, the December delivery date Diez gave Forbes holds up.

The number Agility leads with everywhere, from the deal announcement to the deck, is "$300M+ of Digit v5 orders." The footnote on slide 11 spells out the details.

The figure "relates to 1,000 Digit v5 robots with three-year term RaaS contract, which includes warrants issued to purchaser vesting proportionately to robots deployed," as of May 2026, "subject to the realization of certain contractual milestones," and "not a measure of current period revenue."

Read that footnote carefully: it's a single three-year RaaS contract with a single unnamed customer, not an order book. 1,000 robots over 36 months works out to $8,333/month, right at the $8,500/month illustrative RaaS price.

What the customer gets besides robots is equity. The S-4 says the buyer is "a customer, also a related party," meaning an investor, and that Agility will issue it 453 warrants per robot, on the condition that it takes at least 1,000 robots (453 x 1,000 = 453,000 warrants). The filing adds that Agility "is still evaluating the accounting for this transaction."

The S-4 doesn't name the buyer, but it narrows the field. Two of Agility's investors are also customers in the filing. Schaeffler did $91,500 of business with Agility as a customer in 2025, and Amazon bought five Digits for $1.05M.

The S-4 also shows a warrant from October 2022 on 2,888,182 shares that vests 358 shares per robot purchased, and 1,790 of those shares had vested by year-end.

That's exactly five robots' worth (1,790 / 358), and Amazon is the only customer in the filing that bought five robots. So Amazon already holds a warrant from Agility that vests per robot, and the 1,000-robot order comes with the same setup, a per-robot warrant to a buyer the S-4 calls a related party.

Therefore, the $300M+ of Digit v5 orders is most likely from Amazon.

Agility has done this before. In July 2021 it gave "a potential customer" warrants on 358,000 shares that vest at 358 per robot, which is 1,000 robots' worth. Five years later that customer hasn't bought a single robot, so none have vested, which is what a 1,000-robot warrant commitment can turn into.

Then there's the ramp. Agility builds fewer than 100 robots a year today, v5 hasn't shipped, and the first early-customer units arrive in December.

To deliver 1,000 v5 units the factory has to produce more than 10x its current rate, and management's own deployment schedule, the one it gave Churchill's board and the fairness-opinion firm, has ~800 robots in service in 2027. So the contract doesn't finish deploying until 2028 at the earliest, and its three-year revenue clock runs into 2031.

Beyond that one contract, there are the four Customer Acceleration Program customers from earlier, paying ~$500,000 each for a program that walks them from a proof of technology to an on-site proof of concept and a RaaS pilot, with more than 30 prospects behind them. Management calls the pipeline "multiple times" the $300M.

None of the four are named, and a CAP fee is a paid trial, not an order. So being realistic, the v5 order book is one signed customer, four paid trials, and a healthy pipeline.

I can't treat the $300M as backlog until v5 ships and the milestones start clearing, and the S-4 says the same in its own words.