Agility's Digit Humanoids Deploy in Warehouses, Target $2.5B SPAC IPO

Agility's Digit: From Lab to Warehouse Floor

Agility Robotics, founded in 2015 with over a decade of robot development experience, has deployed its Digit humanoid robots in real-world warehouse operations. At a Spanx facility in Georgia, the 5-foot-9-inch robots move boxes of shapewear. Agility plans to deploy hundreds of robots by year-end for logistics companies and manufacturers.

Training Approach: Demonstration and Simulation

Unlike traditional robots that require explicit programming for each motion and fail when conditions change (lighting, floor conditions, new obstacles), Agility combines two AI training methods to achieve dynamic adaptation:

Learning from Demonstration (LfD): Engineers teleoperate Digit via VR headset and controllers, seeing through the robot's cameras and precisely demonstrating tasks. Sensors capture camera feeds, joint positions, force readings, and end-effector poses. Repeated demonstrations under varied conditions yield a model that generalizes rather than executing rigid scripts. LfD excels at skills hard to codify but natural to demonstrate, such as subtle grasp-angle adjustments.

Large-scale Simulation: Using NVIDIA's open frameworks and libraries and Google DeepMind's MuJoCo physics engine, Agility generates precise virtual physical environments for massive parallel simulation. Reinforcement learning lets Digit explore options cost-effectively. Simulation runs unlimited parallel scenarios continuously, compressing months of real-world testing into hours. This builds the foundational whole-body control model — the "physical intelligence" base layer.

The company states that LfD handles skills nearly impossible to describe with rules but natural to demonstrate, while simulation builds the whole-body control foundation.

Commercial Traction and Economics

Customers include Amazon, GXO Logistics, and Toyota; NVIDIA is a key investor. Time magazine called Agility "the first company to make money with humanoid robots." Digit v5, slated for year-end deployment, has over $300 million in orders and is the first version designed to perceive human coworkers for safe collaboration, greatly improving practicality. Agility claims a five-year total cost of ownership 20% lower than hiring a human for the same work. The company is pursuing a SPAC merger at a ~$2.5 billion valuation.

Popping the Hype Bubble

Co-founder Jonathan Hurst argues that much embodied-AI narrative is disconnected from reality: "A humanoid robot that doesn't need to be caged is far more valuable." The firm positions Digit as a practical, reusable productivity tool rather than a demo spectacle.

References: https://www.agilityrobotics.com/content/agility-and-ai ; https://www.oregonlive.com/silicon-forest/2026/08/oregon-robotics-company-says-the-industry-needs-a-reality-check-i-really-want-to-pop-that-bubble.html

Code example

来源：ScienceAI
本文
约1600字
，建议阅读
5
分钟
Digit 是目前少有的有实际工作的人形机器人。