Almost a year ago to the day, Figure AI CEO Colby Adcock responded to some gentle social media taunting from rival robotics company Agility with a prediction: “You’ll be bankrupt in <12 months.”
Instead, Agility is going public through a reverse-merger that will see it become a publicly-traded company and raise $465 million in new capital from retail investors and a tranche of private funding. The transaction will take place through a special purpose acquisition company, or SPAC.
“You don’t see a lot of dancing robots or backflipping robots from Agility,” CEO Peggy Johnson told TechCrunch ahead of the company’s market debut. “We stuck to our strategy, which [are] these manual, very repetitive jobs, that it’s hard to find humans to do.”
Agility predates the recent boom in physical AI investment. The company was spun out of Oregon State University in 2015 by a group of researchers who made headway in figuring out how to get a robot to walk safely on two legs. The company has focused on deploying a handful of its Digit robots in industrial settings with nine different customers, including Schaeffler, GXO, Toyota Motor Manufacturing Canada, and Mercado Libre. Digits are generally used to move heavy totes, including more than 100,000 of them at a GXO logistics facility.
Compared to humanoid robots built by Figure, Tesla, 1X, or Boston Dynamics, Agility’s Digit can seem plodding, but the robot is actually doing real work, and earning money — albeit just $174,000 in the first six months of 2026, and mostly from related parties. The company lost $362 million in the same time period, which means the clock is ticking for it to successfully deploy its infusion of new capital.
The good news is that Agility has $300 million in planned contracts from 30 customers for its new generation of robot, Digit 5. The new model can safely operate around humans, not just in an isolated work cell like its previous model (and indeed most humanoid robots). That deal was “one of the driving factors” behind the decision to take the company public, Johnson said, but could not yet reveal who the customers or say if they are investors.
To close the deals, Johnson said that Digit 5 will have to hit undisclosed targets along several metrics, including throughput, uptime, and accuracy of placement. The company has yet to show off the prototypes of Digit 5, but Johnson said they will be demonstrated at CES 2027 and widely available at the end of that year.
The company’s pitch is that a robot can save $100,000 a year for companies under a robot-as-a-service model and even more if the customer purchases the robot outright. Whether those unit economics work out in the end will depend on Digit 5’s capabilities, both in terms of task accuracy, and how long it can work without pauses for recharging or maintenance.
Johnson gave the example of a single person lifting a 50-pound object — an informal threshold some workplaces use to avoid injury. “Digit can do that over and over and over,” she said. “You don’t want to give a human that job. You want to give that to the robot.”
Agility faces two key questions: How quickly can it find new customers for Digit 5, and how quickly can it improve on the robot’s capabilities? Prospective customers need to spend $500,000 for Agility to do an on-site proof of concept before a deployment is fully scaled up; there were four customers signed up in May, 2026. And while Agility can pick up heavy loads and place them with 98% accuracy, it still — like most of today’s robots — lacks the ability to do complex manipulation tasks.
It’s still not clear how quickly generative AI techniques will unlock the kind of dexterity that could allow robots to handle a wide variety of items. Training systems based on video and teleoperation data have not led to reliable general autonomy, though adapting the transformer architecture to different modalities appears promising to physical AI researchers.
“There’s no internet of physical data of robot movements, and that’s why you see so many people trying to simulate the data using a variety of things, and we use all of those techniques as well,” Johnson said. “They have their pros and cons, but we also have … 65,000 hours of real operational data from real customers in real facilities…that helps us accelerate the training of Digit.”
Johnson said the capital raised in the SPAC transaction would meet the company’s current needs, but wouldn’t project that it would be sufficient for the company to be self-sustaining. She wouldn’t share production targets for Digit 5, but the company has said its production facility could produce 10,000 of the robots annually.
“What you’ll see from us going forward is continuing to listen to our customers, understanding what they need in these facilities,” she said. “We’re taking our cue from them because we’re building a business.”
For all the open questions, Agility still finds itself further ahead of most humanoid builders by dint of simply putting its product in the field.
