Agility Robotics helped shape the modern humanoid industry before most of its competitors even existed.
The Oregon-based robotics firm got its AI-powered, two-legged machine into the Guiness World Records years before China’s humanoids started going viral for speed. Its influence can be seen in today’s robots that are smashing human records (and into walls). Agility also opened what it calls the world’s first dedicated humanoid robot factory, with a production capacity that matches the leading Chinese companies pushing humanlike machines into the workforce.
Despite its pioneer status, Agility is struggling to keep pace with what it helped create. China is scaling much faster, and Western rivals are raising more money and dominating the hype cycle.
So now Agility is heading into what could be its harshest stress test yet: the public markets. Its plan is already divisive, and the disclosures behind it show just how much pressure is building beneath the surface.
As Agility finds itself in its biggest spotlight yet, can it prove that being early still matters? Or will it become a cautionary tale of the physical AI boom?
Agility’s Origin: ATRIAS
Agility Robotics grew up out a research project that began around 2008 at Oregon State University. That’s where Jonathan Hurst established the Dynamic Robotics Laboratory after getting his PhD from Carnegie Mellon University (CMU). Similar to MIT’s Leg Laboratory that spawned Boston Dynamics, the program sought to build walking machines that could navigate the world as efficiently as people and animals.
One of the lab’s early creations was known as ATRIAS, an acronym for Assume The Robot Is A Sphere. It’s a riff on the old spherical cow joke about physicists oversimplifying complex problems.
The team built ATRIAS to test a simplified model of walking that treated the body like a spring-supported mass. It used two spring-loaded legs and lightweight joints to mimic the basic physics of human and animal locomotion. The robot stored and released energy through its legs as it stepped, while its control system constantly adjusted balance and foot placement.
Though its aesthetics were very unhuman-like, ATRIAS could move more naturally and efficiently than leading humanoids of the era like Honda’s ASIMO. The breakthrough established the design philosophy that Agility would later build upon. It attracted the attention of DARPA, which funded further research into more capable and efficient walking machines for places where wheeled robots struggled.
OSU Spinout
As the technology improved, Hurst and his colleagues saw the commercial value for work in environments build for humans.
He formed Agility Robotics as a small university spinout in 2015 with Damion Shelton and Mikhail Jones. Shelton, who hurst knew from Carnegie Mellon, brought business experience from cofounding threeRivers 3D, a Pittsburgh area 3D-imaging startup. Jones had worked closely with Hurst as a grad student and contributed to the control systems behind the lab’s robots.
They secured almost $800,000 in early funding from The Robotics Hub, a CMU-linked startup accelerator. They were among a small cluster of legged robot startup founders that formed in the mid 2010s.
In Philadelphia, researchers spun Ghost Robotics out of the University of Pennsylvania to develop fast and agile quadruped robots that would eventually find their way into military and security operations. The ETH Zurich spinout ANYbotics was starting to commercialize its research into four-legged robots. The work would evolve into its flagship robot, the ANYmal, which is widely used to inspect dangerous industrial sites.
And in China, hardware engineer Wang Xingxing founded Unitree Robotics in Hangzhou to bring high-performance quadrupeds to the mass market. Unitree’s founder was heavily influenced by Boston Dynamics, which was entering the mainstream during the era with early prototypes of its iconic Spot robodog and Atlas humanoid.
First Product: Cassie
Agility’s founding team had its first major win in 2016 when it secured a million-dollar DARPA contract to turn its work from ATRIAS into a more robot named Cassie. Its name was inspired by the cassowary, a large flightless bird found in the rainforests of New Guinea and northeastern Australia known for its powerful spring-like legs.
Introduced in 2017, Cassie had a pair of robotic legs similar to its namesake’s with a compact torso and exposed mechanical joints. The original Cassie stood about 1.1 meters (3 feet 7 inches). At about 31 kg (68 lbs), the robot was half the weight of ATRIAS.
It had 20 degrees of freedom (DoF) with 10 DoF per leg. Of those, five joints in each leg were actively powered, while others were passive and spring controlled. The mix helped Cassie stay lightweight and move more efficiently than a fully motorized design.
Cassie became famous for its running capabilities. In 2021, the biped finished a 5K in about 53 minutes on a single battery charge. And in 2022, a Cassie robot set a Guiness World Record for the fastest 100-meter sprint by a two-legged robot, finishing in 24.7 seconds.
Though slower than human runners, Cassie completed both feats autonomously using machine learning control. Like Neo in The Matrix, the robot’s digital twin trained through thousands of simulated runs before transferring what it learned to a real Cassie through a process called Sim-to-Real, short for Simulation to Reality. On the physical robot, onboard sensors constantly measured its body position and movement as the controller adjusted its motors many times per second.
Agility ultimately sold just over a dozen Cassies to leading research institutions for around $250,000 per unit. Though it never reached mass production, Cassie earned the robotics firm recognition far outside the usual tech echo chambers. Its influence is also obvious in the designs of newer high-performance Chinese bipeds like Honor’s Lightning, whose top speeds exceeds human capabilities.
Digit the Robot Worker
Cassie’s successor brought Agility a new level of attention. By the time the company introduced its flagship robot worker, Digit, in February 2019, it had raised nearly $9 million to date.
The OG Digit stood about 155 cm (5 foot 1) and weighed 42 kg (93 lbs). It inherited Cassie’s proven leg design, but added a larger torso, arms, sensors, and much more onboard computing. Each arm had 4 DoF, which Digit used to carry objects and to help keep balance and protect itself when it stumbled. It could carry and stack boxes weighing up to 18 kg (40 lbs).
Digit V1 used lidar and depth cameras to perceive its surroundings, plan its footsteps, and navigate around obstacles. Its walking speed stopped out at 1.5 m/s (3.4 mph). Like Cassie, it could climb stairs and traverse uneven terrain.
Agility sold early Digit prototypes to research labs and corporate R&D teams. Among them was the CLEAR Lab at Shenzhen’s Southern University of Science and Technology, better known as SUSTECH. The research was led by Professor Wei Zhang, who’d go on to found LimX Dynamics, one of the perceived leaders in China’s humanoid robotics scene.
Agility refined Digit for its first commercial customers.
Ford became the first buyer in 2020, taking two Digit robots to explore autonomous last-mile delivery. Ford envisioned a delivery system that it could potentially sell to commercial fleets. Digit would ride in the self-driving van, unfold itself, and carry a package to the recipient’s door.
The concept grabbed headlines when Ford announced its intentions, but the automaker quietly abandoned the project as Agility shifted its focus toward warehouse and factory work.
Digit had its biggest redesign in 2023. Agility funded it after securing around $150 million in a Series B round the prior year, its biggest raise led, led by DCVC and Playground Global. Amazon also participated through its billion-dollar Industrial Innovation Fund.
The new and improved Digit had a much larger and more industrial build than the first iteration, standing 175 cm (5 feet 9 inches) and weighing 65 kg (143 lbs). It gained a more andromorphic head with animated eyes and a clearer face to communicate with human coworkers. Agility also gave Digit specialized hands engineered to reliably transport totes in logistics facilities.
The company had begun integrating LLMs into the humanoid’s higher-level intelligence, though its walking and balance still used traditional robotic software.
Under the leadership of CEO Shelton and Hurst as CTO, Agility also changed its marketing strategy. Rather than pitching Digit as general purpose, the company locked in on warehouse work, especially moving totes between stations and conveyors. However, Agility still said Digit could expand into broader jobs over time, excluding lethal applications that it pledged to never support.
Agility began livestreaming its robots during this era at events like Promat in Chicago and MODEX in Atlanta. The Digits used QR-like markers called AprilTags to determine their position and line up precisely with objects and workstations. Aside from a few viral mishaps, the streams got little attention compared to ones years later by rising competitors like Silicon Valley-based Figure AI and Shanghai’s Agibot.
In October 2023, Amazon announced that it was piloting Digit robots at its R&D center south of Seattle. Digit’s first role was tote recycling, picking up and moving the empty plastic containers after human workers removed all the inventory from them. Amazon stressed that the Digit test was a small pilot, not a commercial deployment, but the robot’s photogenic design made it a media symbol for the ecommerce giant’s automation push and the transformation of the future of work in general.
Agility’s Current Era
Agility entered its current era in early 2024. Peggy Johnson became its new CEO in March, replacing Shelton, who moved into the role of president.
She had spent more than two decades at the chipmaker Qualcomm, first as an engineer before moving into business development. Then she joined Microsoft in 2014 as an executive vice president overseeing partnerships and corporate development.
In 2020, she became CEO of Magic Leap, where she attempted to shift the struggling augmented reality company from consumers and toward enterprise customers. Under Johnson, the Florida-based company moved into next-gen applications within Agility’s orbit, including the industrial metaverse, connected-worker training, digital twinning, and factory automation. She kept Magic Leap afloat, but it remained commercially weak when she left in October 2023.
At least from the outside, Agility appeared to be in a much stronger position when Johnson came aboard. The company’s workforce had grown to more than 200 employees and it had just opened what it called the world’s first humanoid robot factory in Salem, OR. The facility, called RoboFab, covers about 6,500 square meters (70,000 square feet) and is designed to manufacture more than 10,000 robots per year once fully ramped.
Though the numbers appeared staggering on paper, there wasn’t enough demand to justify anything close to full production capacity. There was also a lot more competition and much higher expectations compared to when Agility first unveiled Digit.
In 2019, Agility pretty much only had to worry about Boston Dynamics. But the Massachusetts robotics leader appeared nowhere close to commercializing Atlas. The hydraulic-powered robot was viewed more like an advanced research platform, like NASA’s Valkyrie, that demonstrated what humanoids might eventually do.
Tesla ignited the modern humanoid race as Elon Musk said he believed Optimus would become its best-selling product by far. The hyperbole, along with the explosive growth of generative AI, helped pull huge amounts of attention and capital into the field.
By the time Johnson joined Agility, Figure AI had just raised a then-unprecedented $675 million in a Series B found and was preparing to send its humanoids to BMW’s plant in Spartanburg, SC. Another rising firm, Austin-based Apptronik, announced a commercial agreement with Mercedes-Benz to test its Apollo humanoids in its factories. And Boston Dynamics was preparing to retire its hydraulic Atlas, only to replace it with a fully electric humanoid engineered for the industrial workforce.
But China!
While competition was stiffening in the US, a tsunami was coming from China after Beijing set a national goal to mass manufacture humanoids by 2025 and to lead the market by 2027. Chinese humanoids from companies like Unitree, Agibot, and Booster Robotics would soon become ubiquitous at research labs worldwide with prices Agility couldn’t compete with.
As China flooded the market with hardware below $20,000, Agility became a leading voice pushing for tightened restrictions on their imports. Cofounder Shelton testified before a US Senate subcommittee in March 2026 about the competitive threat.
“To be as blunt as possible, they’re doing a good job,” he said. “Their technology is well-designed, highly capable, and backed by a formidable supply chain.”
He added that he was concerned after reading two recent research papers that “identified a critical security vulnerability in a Chinese humanoid robot,” referring to Unitree’s G1 without naming it, “allowing for remote takeover as well as a phone-home data-logging mechanism that sends data to a remote server.” The robot, he noted, was available for less than $20,000 in the US.
“Combined with open-source and open-weight AI models, China is creating a compelling value proposition for early adopters of general-purpose humanoid robots,” he said. “This is a siren’s call that would be well served to take seriously or risk ceding the future of automation and, by extension, economic growth to others.”
Under Johnson’s leadership, Agility aligned itself with the Trump Administration’s push to protect the domestic robotics sector. She publicly backed restrictions on Chinese competitors, and in July 2026, the FCC effectively shut the door on new foreign-made imports.
Going Public
The decade-old company, which now simply calls itself Agility, hopes the public markets will help it compete in the humanoid race.
It plans to merge with Churchill Capital Corp. XI, a SPAC incorporated in the Cayman Islands in June 2025 by investor Michael Klein. The longtime Wall Street dealmaker has taken several companies public through SPACs, which stands for Special-Purpose Acquisition Companies.
A SPAC is basically a public shell company built to raise money and then merge with a private company. Investors buy in before knowing what company it will merge with. Once the merger closes, the private company takes the SPAC’s place on the stock market. It’s essentially a shortcut to going public by bypassing the lengthy IPO process, though critics say SPACs mostly favor dealmakers, encourage inflated valuations, and push companies onto the stock market before they’re ready.
SPAC stocks often surge on hype. Then reality sets in. Supporters say they still have a purpose. They can be useful in fast-growing industries, like humanoid robotics, that require companies to raise a lot of capital before they’re profitable.
It’s a route that Figure founder Brett Adcock chose when he took Archer Aviation public with Atlas Crest Investment Corp. in 2021. The deal brought Archer $857.6 million in gross proceeds as its shares began trading on the New York Stock Exchange. Archer, which Adcock exited in 2022, is now trading at about half of its post-listing peak, though it has more than tripled its all-time low.
Churchill Capital’s SPAC record is considered weak-to-mixed for long-term investors. One of his biggest disappointments has been Skillsoft, a New Hampshire-based corporate training and online learning company that went public through Churchill in 2021. Its stock collapsed the following year, shedding 86 percent of its value. Skillsoft is now executing a plan to avoid being delisted.
Agility hopes to follow a better example: Symbotic. The Massachusetts warehouse robotics company also went public through a SPAC, backed by Softbank, in June 2022. As of September 2026, its shares are trading at about 4x their original price. Symbotic also grew into a major warehouse automation company, reporting more than $1.1 billion in contracted orders soon after going public.
“It’s the most efficient, flexible way to tap into the public markets and there is pent-up demand right now across investors,” Johnson told CNBC’s Sqawk Box
Agility’s deal values the company at $2.5 billion and is expected to bring in more than $620 million in fresh funding. Of that, $200 million would come from outside investors through a PIPE (private investment in public equity) led by Foxconn, the Taiwanese electronics manufacturing giant.
Agility says the money will help scale Digit production and improve its software and AI. It also wants to support much larger customer deployments.
As part of the SPAC process, Agility filed a Form S-4 with the US Security and Exchange Commission (SEC). The filing gave the public its clearest look yet at Agility’s finances.
The numbers show a very early-stage business. It reported generating $1.8 million in net sales in 2025. It posted an operating loss of $140 million and also burned through nearly $100 million in cash that year. So, the $2.5 billion valuation is about 1,400x Agility’s 2025 revenue.
The target is much more aggressive than Unitree’s pre-IPO valuation. The Chinese robotics leader initially targeted a raise of about $620 million, which is almost identical to Agility’s planned raise. Its early valuation was around $6.2 billion but strong investor interest pushed it to $9 billion.
Unitree’s stock exploded when it entered the Shanghai STAR Market. Shares jumped 460 percent on the first day of trading, but quickly fell around 45 percent, erasing about $30 billion in market value. Still, Unitree remained worth nearly 5x its original expected valuation. Unitree also entered the public markets from a much stronger financial position as it reported turning a profit and generating hundreds of millions of USD in revenue shipping thousands of humanoids and quadrupeds worldwide.
Agility’s auditor questioned whether the company could continue for more than a year without significant funding. Its valuation is highly based on assumptions about future growth.
The filing names five major companies that are using or committed to using Digit: Amazon, the global supply chain company GXO Logistics, the German industrial automation leader Schaeffler, Toyota Motor Manufacturing Canada, and Mercado Libre, a Latin American ecommerce and fintech company. At least two of the customers, Amazon and Schaeffler, also have equity in Agility.
The filing also details Agility’s biggest announced deal, valued at more than $300 million. The order, for up to 1,000 Digit V5 robots, is apparently from an existed related party that’s also being offered the ability to buy AGLT stock as it takes delivery of the robots.
The paperwork also peels back the curtain on Agility’s pricing structure. Its robots-as-a-service (RaaS) model runs about $8,500 with a $25,000 fee for deployment. Buying Digit outright costs about $200,000. The robots are expected to cost around $150,000 to build. Deployment adds $20,000, while software and maintenance add another $36,000 per year for each robot.
That annual fee alone is close to what many workers earn doing the same narrow tasks that Digit handles alone. The real cost is also likely much higher because companies also need trained robotics staff to supervise and maintain the systems who can earn into the six figures.
Agility says it’s charging a premium for a system built for real industrial work, and that Digit’s capabilities are growing. The company isn’t targeting home use in the near-term, but insists the path is through industry. By the end of 2026, the company says it expects to have the first humanoid robot certified to work outside caged areas in factories and warehouses.
Digit v5
Agility revealed the fifth-generation version of Digit in mid-September 2026, which Agility says is designed to “work in close proximity to people without the physical safety barriers required by traditional automation.”
Digit v5 has a much bulkier design and a more humanlike silhouette that sheds the bird-like legs that defined earlier iterations. Agility says the more conventional design improves lifting and squatting. It also supports interchangeable hands for different material-handling jobs beyond tote transport. An animated visual in Agility’s launch video teases a wheeled variant shown working in a datacenter.
At 180 cm (5 feet 11 inches) and 129 kg (284 lbs), it’s almost 3 percent taller and more than double the weight of v4. According to Agility, Digit 5.0 can lift loads up to 22.7 kg (55 lbs), which is a 43 percent improvement over v4’s payload capacity. In its investor materials, Agility rounds the improvement to 40 percent.
Its battery life is just 90 minutes, but Agility says recharges autonomously in 9 minutes for up to 20 hours of work per day. The approach differs from companies like UBTECH, Hexagon, and Boston Dynamics, whose latest humanoids have hot-swappable batteries that they can change themselves for near-24/7 operation.
But Agility is playing up the safety features the most. The company says Digit v5 has specialized sensors that can detect when a person is too close so it can slow down, stop, or shut off.
Agility has previously said Digit v5 would eliminate the need for safety cages. However, its press release stopped short of announcing that Digit 5 had received a third-party certification validating cage-free operation. This is definitely something they would mention CEO Johnson has said she expects to unveil the first humanoid certified to operate outside of a cage by the end of 2026.
Instead, the company says it’s “expanding Digit 5’s safety architecture by partnering with industry leaders is the first launch partner for NVIDIA’s Halol for Robotics platform.” The system, powered by NVIDIA IGX Thor, is marketed as a dedicated safety layer for monitoring a robot’s surroundings and responding to potential hazards.
The certification is expected to involve an OSHA-recognized NRTL and compliance with standards including ISO 10218, ANSI/RIA R15.08, and ISO 13849. Agility is also helping develop the forthcoming ISO 25785-1 standard that’s specific to industrial robots like humanoids.
The rollout seemed rushed. Here’s what I posted on X about it:
Obviously, Agility wanted to announce the safety certification since its launch video includes a scene where workers literally remove the barriers separating it from people.But they also need to maintain investor confidence as it moves toward a SPAC IPO with Churchill Capital Corp. XI.
$CCXI initially surged after the Agility merger plan was announced on June 24 and reached a peak of $19.69 on July 6. But then enthusiasm faded and shares pulled back toward $12.
Agility's Digit v5 unveiling seems rushed and I think I know why.
— Mike Kalil (@mikekalilmfg) September 15, 2026
The Oregon-based company is playing up safety hard with the unveiling, calling the 129 kg (284 lb) Digit 5 its "first humanoid robot engineered for cooperatively safe work at scale."
Since Peggy Johnson took… pic.twitter.com/8TOsbgRN51
Doubts about its $2.5 billion valuation grew after its S-4 SEC filing revealed heavy losses and Digit’s high costs. The robots are priced around $200,000 to buy or $8,500 per month to rent, with tens of thousands of dollars in annual software and maintenance fees per unit. A Bernstein analysis said costs need to fall 30 to 50 percent to make economic sense for customers.
In the media, Agility about $300 million in contracted orders, but the S-4 filing indicates it comes from one 3-year RaaS agreement for 1,000 robots. It also identifies the customer as a related party who’s getting shares in exchange for each robot it deploys.
Agility isn’t gonna come out and say they rushed the announcement to boost CCXI’s share price and restore enthusiasm around the merger. But the rollout is unusual for them.
Agility teased Digit 5 the night before and scheduled it for 7:00 AM ET for before markets opened. For comparison, they never even formally announced Digit v4.
They needed to get something out. Now when they actually get the certification, it won’t hit the same because the Digit 5 launch is already being reported as if the cage-free breakthrough has happened.