Agility Robotics Shipped 300 Digits and Nobody Wrote About It

Somewhere in an Amazon fulfillment center, a robot with no head is picking up a yellow tote, walking it across a warehouse floor, and placing it on a conveyor belt. It does not give interviews. It does not appear in Super Bowl commercials. Its parent company has not raised $1.85 billion and does not have a $39 billion valuation.

That robot is Digit, built by Agility Robotics in Corvallis, Oregon. And Agility has shipped 300 of them.

Three hundred deployed humanoid robots makes Agility Robotics the fifth-largest humanoid shipper in the world. Ahead of Figure AI at 200. Ahead of Apptronik at 50. Ahead of 1X Technologies at 20. Ahead of Sanctuary AI at 15. In a race where most American competitors are still counting units in the double digits, Agility crossed into the hundreds without anyone outside the logistics industry particularly noticing.

The disparity between Agility’s output and its media profile is not a minor curiosity. It tells you something fundamental about how the humanoid robot industry is being covered, which companies are being rewarded, and what kind of progress actually matters in the early stages of a hardware revolution.

The Oregon origin story

Agility Robotics was founded in 2015 by Jonathan Hurst and Damion Shelton, spinning out of Oregon State University’s Dynamic Robotics Laboratory. Hurst is a professor of mechanical engineering who has spent over two decades studying how animals move, specifically the spring-like dynamics of bipedal locomotion. He is the kind of academic researcher who measures success in published papers and working prototypes, not in press coverage or fundraising rounds.

The company’s first product was Cassie, a legs-only bipedal robot that looked like an ostrich and walked like one too. Cassie was not trying to be a general-purpose humanoid. It was a research platform designed to prove that legged locomotion could work outside the lab, on real terrain, in unpredictable conditions. The robot set the Guinness World Record for the fastest 100-meter dash by a bipedal robot and completed a 5K run at Oregon State in just over 53 minutes.

Cassie mattered because it forced Agility’s engineering team to solve the hardest problem in bipedal robotics: walking reliably in the real world. Not on flat concrete under perfect laboratory conditions. On uneven warehouse floors, over small obstacles, through doorways, around humans. Cassie’s lessons are embedded in every Digit that rolls off the production line today.

That founding date matters. Agility started in 2015. Figure AI started in 2022. That is seven years of additional engineering iteration, seven years of walking data, seven years of learning what breaks when bipedal robots operate outside controlled settings. In an industry where everyone talks about AI breakthroughs and foundation models, Agility’s biggest advantage might simply be time.

What Digit actually is

Digit is unusual among humanoid robots because of what it lacks. There is no face. No head in the traditional sense. No attempt to look human. Where most humanoid robots place an expressive head with cameras and sensors arranged to suggest eyes and a mouth, Digit has a compact sensor cluster that sits low between its shoulders. The design choice is deliberate and revealing.

Digit stands 5 feet 9 inches tall and weighs approximately 140 pounds. It can carry up to 35 pounds in its two-fingered grippers, walk at roughly 3.5 miles per hour, and operate for several hours on a battery charge. It has 16 degrees of freedom, significantly fewer than the 40-plus that full-dexterity humanoids like Tesla’s Optimus or Figure’s 02 offer. Its hands are functional grippers, not articulated fingers capable of fine manipulation.

Look at that chart and you might conclude Digit is technically inferior. That would be missing the point entirely.

Agility made a series of engineering tradeoffs that prioritize reliability and deployability over generality. Fewer degrees of freedom means fewer actuators, fewer points of failure, less weight, lower cost, and simpler software. Two-fingered grippers cannot fold laundry, but they can pick up and move the standardized totes that make up the backbone of warehouse logistics. The absence of a humanoid head means less material cost, less weight, and no uncanny valley problem when workers share a floor with the robot.

These are not compromises born from inability. They are strategic decisions about what the first generation of commercially viable humanoid robots needs to do. Agility is not trying to build a general-purpose household robot. It is building a warehouse worker that walks on two legs, and it has optimized Digit for exactly that job.

The Amazon partnership

Amazon is both an investor in Agility Robotics and the company’s largest deployment partner. Understanding the structure of this relationship explains why Agility ships more robots than companies with ten times its funding.

Amazon has over 750 fulfillment centers worldwide. These are not quiet storage facilities. They are high-throughput logistics operations where millions of items move through complex sorting, packing, and shipping workflows every day. The physical infrastructure was designed for human workers, which means narrow aisles, stairs, varied floor surfaces, and constantly changing layouts. Traditional wheeled robots and fixed automation cannot navigate these environments. Legged robots can.

Amazon began testing Digit in fulfillment centers in 2023, initially in a single facility in Washington state. The pilot focused on a specific task: moving empty totes from one location to another. This is exactly the kind of work that is brutally repetitive for human workers but straightforward for a bipedal robot with basic gripper capabilities. Pick up tote. Walk. Set down tote. Repeat.

The pilot expanded throughout 2024 and 2025. By early 2026, Digit units are deployed in multiple Amazon facilities across the United States, handling tote movement and basic logistics tasks. The 300 units that Agility has shipped are overwhelmingly going to Amazon and to a smaller number of third-party logistics companies like GXO Logistics.

Here is what makes the Amazon partnership structurally different from most humanoid deployment deals. Amazon does not need convincing that warehouse automation is valuable. Amazon has invested billions in robotics over the past decade, acquiring Kiva Systems in 2012 for $775 million and deploying over 750,000 mobile robots in its facilities. The company understands unit economics, deployment logistics, and total cost of ownership better than almost anyone in the world.

When Amazon tests Digit, it is not running a publicity stunt. It is running a rigorous operational evaluation with clear metrics for throughput, reliability, error rates, and cost per task. The fact that Digit survived this evaluation and expanded to multiple facilities is a stronger validation of the technology than any demo video or conference presentation could provide.

RoboFab: the factory nobody talks about

In late 2023, Agility Robotics opened RoboFab in Salem, Oregon. It was the world’s first purpose-built factory designed specifically for manufacturing humanoid robots.

Read that sentence again. The world’s first purpose-built humanoid robot factory was not in Shanghai, not in Austin, not in Silicon Valley. It was in Salem, Oregon, a city of 180,000 people better known for being the state capital than for advanced manufacturing.

RoboFab is designed to produce up to 10,000 Digit units per year at full scale. That number is important because it means Agility was planning for mass production before most competitors had finished their first prototype. When Figure AI was still iterating on the Figure 01, Agility was building the factory that would produce Digit at scale.

The facility spans approximately 70,000 square feet and includes dedicated areas for subassembly, final integration, testing, and quality assurance. The production line is designed for a humanoid robot that is intentionally less complex than its competitors, meaning fewer parts, fewer assembly steps, and faster throughput per unit.

RoboFab is currently operating well below its full capacity. Producing 300 units over the course of a year and a half from a factory designed for 10,000 annually means the facility is running at roughly 5-10% utilization. That seems wasteful until you understand that the capacity exists for a reason. When Amazon decides to scale Digit deployment from dozens of fulfillment centers to hundreds, Agility needs to be able to meet that demand without a two-year construction delay. The factory is an option on the future, pre-paid.

The coverage gap

Figure AI has raised $1.85 billion. Agility Robotics has raised $179 million. That is roughly a 10x funding difference. But the media coverage gap is far larger than 10x.

Search for “Figure AI humanoid robot” in any major technology publication and you will find hundreds of articles from the past 12 months. Profiles of founder Brett Adcock. Analysis of the $39 billion valuation. Coverage of every new demo video. Breathless reporting on the BMW partnership. Deep dives into the OpenAI collaboration.

Search for “Agility Robotics Digit” and you will find a fraction of that coverage. A few articles when RoboFab opened. A brief mention in stories about Amazon’s warehouse automation strategy. An occasional update buried in the back pages of a robotics trade publication.

Agility has shipped 50% more humanoid robots than Figure AI. It opened the world’s first humanoid factory a year before Figure opened BotQ. It has a deployment partnership with the world’s largest logistics company. And yet, in the media landscape that shapes investor perception and public understanding of the humanoid race, Figure AI occupies roughly 100 times more space.

Why does this happen? Several factors compound.

Funding announcements generate headlines. Figure AI’s $675 million Series B and its $1 billion Series C were each major financial news events. Agility’s $150 million Series B was significant for a robotics company but did not hit the threshold that generates coverage from the Financial Times or Bloomberg’s front page. In a media environment where funding is treated as a proxy for importance, companies that raise less get covered less, regardless of what they ship.

Celebrity investors attract attention. Figure AI’s investor list includes Jeff Bezos, Microsoft, NVIDIA, Intel, and OpenAI. Each of these names comes with its own media ecosystem. When Figure raises a round, the story is not just about Figure. It is about what Bezos thinks about robots, what NVIDIA’s strategy means, how OpenAI is expanding beyond language models. Agility’s investors, while credible, do not have the same name recognition.

Demo videos outperform deployment data. A 90-second video of a humanoid robot doing something impressive generates more engagement than a press release about 300 units deployed in fulfillment centers. Figure AI and Tesla have mastered the art of producing visually compelling demos. Agility’s Digits working in Amazon warehouses are doing real work, but warehouse work does not look exciting on social media.

The narrative favors ambition over execution. Figure AI is building a general-purpose humanoid that will eventually do everything. Agility is building a warehouse worker that carries totes. The first story is more interesting to write and more interesting to read, even if the second story represents more real-world progress.

The case for boring robots

Digit is, by design, a boring robot. It does not have dexterous hands. It cannot have a conversation. It does not look human. It does one thing in one environment, and it does that thing reliably enough that the world’s most demanding logistics operation keeps deploying more of them.

This is not a failure of imagination. It is an expression of something Jonathan Hurst has argued since Agility’s founding: the path to general-purpose humanoid robots runs through special-purpose humanoid robots. You do not build a machine that can do everything by trying to build a machine that can do everything. You build a machine that can do one thing well, deploy it at scale, collect data from real operations, and gradually expand its capabilities.

The Digit that Agility ships today is dramatically more capable than the Digit that shipped in 2023. Each generation incorporates lessons from thousands of hours of warehouse operation. The walking gait is smoother. The tote manipulation is more reliable. The obstacle avoidance is more robust. The battery management is more efficient. These improvements come from real-world deployment data that no amount of simulation can replicate.

This approach has historical precedent. The iPhone in 2007 could not even copy and paste text. It had no app store. It could not record video. Apple did not ship a fully featured smartphone on day one. It shipped a phone that could do a few things exceptionally well, then expanded capabilities over time based on real user data. The analogy is imperfect, but the principle holds: ship, learn, improve, repeat.

Where Agility sits in the global race

When you zoom out and look at the full humanoid shipment leaderboard, Agility’s position is both impressive and precarious.

At 300 units, Agility is the second-highest shipping American humanoid company behind Tesla’s 500. It leads all American pure-play humanoid startups. But the Chinese leaders are operating at a completely different scale. Unitree has shipped 5,500 units. AgiBot has shipped 5,200. Even UBTECH and Fourier, at 1,000 units each, are more than three times Agility’s volume.

The gap is real, but the context matters. Unitree’s volume leader, the G1, is a compact humanoid priced around $16,000. AgiBot’s fleet includes many lower-cost models deployed across diverse industrial settings. Agility’s Digits are full-size humanoids deployed in one of the most demanding operational environments on Earth. The per-unit commercial value and the operational data generated per deployment are not directly comparable.

Tesla, at 500 units, is the only American company shipping more humanoids than Agility. But Tesla’s units are deployed exclusively within Tesla’s own factories, which limits both the diversity of operational data and the commercial validation. Agility’s deployments at Amazon represent genuine third-party commercial adoption.

The timeline tells a story of patience

Agility’s journey from academic research to 300 deployed units spans over a decade. That timeline is worth examining because it reveals an approach to company-building that is fundamentally different from the venture-backed speed runs that dominate current humanoid discourse.

Compare this to Figure AI’s timeline. Founded May 2022. First prototype demonstrated 2023. BMW deployment 2024. 200 units shipped by early 2026. Figure moved fast by any measure, but it has existed for less than four years. Agility has been working on bipedal robots for over a decade.

The question is whether Agility’s decade of accumulated knowledge translates into a durable competitive advantage, or whether well-funded competitors like Figure AI and Tesla can compress that learning curve with massive capital investment and advanced AI capabilities.

What happens next

Agility’s trajectory depends almost entirely on one variable: the speed at which Amazon scales Digit deployment.

If Amazon decides that Digit has proven itself and begins deploying the robot across hundreds of fulfillment centers, Agility could quickly become the highest-volume American humanoid shipper. RoboFab is built for 10,000 units per year. Amazon has over 750 fulfillment centers globally. The math works.

If Amazon moves slowly, testing for another year or two before committing to large-scale deployment, Agility risks being overtaken by competitors who are scaling faster. Tesla is targeting 50,000-100,000 Optimus units in 2026. Figure AI’s BotQ is built for 12,000 units per year. AgiBot is targeting 10,000-plus annual production. Agility cannot afford to remain at 300 units for long.

The company also faces a strategic decision about capability expansion. Digit’s current form factor is optimized for warehouse logistics with basic grippers. But Amazon’s warehouses require more than tote movement. There is picking, packing, sorting, stacking, and dozens of other tasks that would benefit from a bipedal robot with better manipulation capabilities. Agility will eventually need to add dexterity without sacrificing the reliability that makes Digit commercially viable today.

The uncomfortable conclusion

Three hundred humanoid robots are walking around Amazon warehouses today, picking up totes and moving them where they need to go. These robots were built in the world’s first purpose-built humanoid factory by a company founded by a bipedal locomotion researcher from Oregon State University. The company has shipped 50% more humanoid robots than the $39 billion startup that dominates every robotics headline.

Nobody wrote about it.

That gap between execution and attention is not just a media curiosity. It reflects a deeper distortion in how the humanoid robot industry is being evaluated. We are measuring companies by their fundraising, their valuations, their demo videos, and their founder profiles. We should be measuring them by their shipped units, their customer retention, their real-world reliability data, and their manufacturing capability.

By those measures, Agility Robotics is one of the most quietly successful humanoid companies in the world. It is not the flashiest. It is not the most funded. It is not building the most versatile robot. But it is building robots that work, shipping them to the world’s most demanding customer, and manufacturing them in a factory designed for scale.

In a market that rewards hype, Agility chose execution. Whether that choice will be rewarded in the long run depends on whether the industry eventually catches up to the simple truth that shipped robots matter more than demo videos.

Three hundred Digits would disagree with the current allocation of media attention. But they are too busy working to say so.