billion Industrial Innovation Fund in 2022, signaling Amazon's serious interest in bipedal automation. Figure has emerged as a leading competitor, with its robots now undergoing testing in operational Amazon warehouses.
What Figure Robots Actually Do
The Figure robots you'll find in Amazon warehouses aren't replacing humans entirely—yet. They're handling specific tasks that represent the "last mile" of automation challenges: tote recycling, picking and placing individual items, and operating in spaces too narrow or dynamic for traditional automation.
The robots stand approximately 5'6" tall, can carry up to 45 pounds, and operate for up to 5 hours on a single charge. They use stereo vision cameras and neural networks to perceive their environment, identify objects, and plan grasping strategies. When a robot encounters something it can't handle—an oddly shaped package, a cluttered bin—it flags the item for human attention and moves to the next task.
This human-in-the-loop approach characterizes the current deployment phase. The robots aren't fully autonomous generalists; they're specialized workers that handle routine tasks while escalating exceptions to humans. Over time, as the robots encounter more edge cases and their neural networks improve, the exception rate drops, and they handle more independently.
The Economics of Humanoid Labor
The business case for humanoid robots rests on a simple calculation: if you can build a machine that does human work for less than human wages, and you can deploy it at scale, you've unlocked enormous value.
Current humanoid robots cost between $50,000 and
00,000 per unit. With a five-year depreciation schedule, that's
0,000 to
0,000 per year—comparable to or slightly higher than a human worker's annual wages in many markets. But robots work 24/7, don't require benefits, never take sick days, and don't form unions.
The real breakthrough, however, isn't just cost—it's scalability. During peak seasons like holidays, Amazon currently hires hundreds of thousands of temporary workers. With humanoid robots, they could theoretically deploy additional robots on demand, scaling capacity without the massive HR operation required for seasonal hiring.
The Challenge of Human Spaces
The hardest problem in humanoid robotics isn't building a robot that can walk—it's building one that can navigate spaces designed for humans. Warehouses have varying shelf heights, uneven floors, narrow aisles, staircases, and doorways. They're filled with unpredictable obstacles: fallen boxes, loose packing material, other workers moving erratically.
This is why the humanoid form factor matters. By matching human dimensions and capabilities, these robots can use existing infrastructure without modification. They can climb the same stairs, reach the same shelves, and operate the same tools as human workers.
Figure's robots use a combination of LIDAR for mapping, cameras for object recognition, and proprioceptive sensors for balance. When a robot encounters a staircase, it doesn't need a ramp—it just walks up, using real-time balance adjustments to handle varying step heights.
Competition Heats Up
Amazon isn't betting exclusively on Figure or Agility. The company is actively exploring multiple approaches to mobile manipulation, including combining traditional wheeled AMRs with robotic arms—essentially creating wheeled humanoids that sacrifice bipedal locomotion for stability and payload capacity.
Meanwhile, competition among humanoid startups is intensifying. Figure recently announced partnerships with BMW for automotive manufacturing applications. Tesla continues to develop Optimus, with Elon Musk claiming it will eventually be larger than the electric vehicle business. 1X, backed by OpenAI, is focusing on household applications.
If Digit doesn't seamlessly integrate into Amazon's workflow, it may not mark the end of bipedal robots. The ongoing pilots have the potential to reshape how we perceive and utilize these robots in various industries.
The Human Question
Inevitably, the rise of humanoid robots raises concerns about employment. Amazon employs over 1.5 million people globally, many in warehouse roles that could theoretically be automated.
The company's official position emphasizes augmentation rather than replacement. "We're not trying to eliminate jobs," an Amazon Robotics executive explained. "We're trying to eliminate the parts of jobs that people don't want to do—the repetitive lifting, the walking miles every day, the physical strain. Our vision is humans and robots working together, with robots handling the dull, dirty, and dangerous work while humans focus on problem-solving and exceptions."
History supports this view to some extent. When Amazon introduced Kiva robots, warehouse employment actually increased—the robots made fulfillment more efficient, which grew the business, which required more humans for tasks robots couldn't handle. But humanoid robots are different. They're designed to handle exactly the kinds of tasks that previously required humans.
What Comes Next
The next five years will determine whether humanoid robots become as ubiquitous as forklifts or remain niche curiosities. Key developments to watch:
Cost reduction: Can manufacturers drive prices below $30,000 per unit, making them accessible beyond mega-corporations?
Generalization: Can robots handle increasingly broad ranges of tasks without reprogramming?
Safety: Can they work safely alongside humans without cages or restricted zones?
Battery life: Can they operate full shifts without recharging?
If Figure and its competitors succeed, the warehouse of 2030 will look very different—staffed by teams of humans and humanoids working side by side, each doing what they do best. And that quiet whirring sound you hear? It's the future, walking in on two legs.