Inside Amazon's Robotics Race and What It Means for the Future of Human Work

The world's second-largest corporate employer has bet on large-scale deployment of robotic workers.

Amazon has more than 1 million robot workers. What is the future of human resources?

The second-largest corporate employer is betting on robot workers at scale.

Wheeled mobile robots move tall inventory pods full of items inside an Amazon fulfillment center.

No organization in the world uses robots as much as Amazon. The company already has more than 1 million robots working around the clock in warehouses to deliver tens of millions of same-day and same-day packages daily — and Amazon is poised to significantly expand its robotic fleet in the coming years as it builds new factories to attract more capable robot workers.

Despite deploying so many robots, Amazon is also the world's second-largest corporate employer, with more than 1.5 million full-time and part-time workers, according to its 2025 annual report. Amazon executives say the growing fleet of robots has reduced the burden on its human workforce by handling the most demanding fulfillment center tasks, which carry the greatest risks of physical fatigue and injury. But labor advocacy groups and independent researchers point to evidence that AI-driven warehouse automation isn't all positive for humans working alongside robots.

Bryan: "In the last 100 years, most of the automation that we've seen has tended to happen away from the hands of the workers — you're taking this high-risk work away from the worker and you're doing it in a way that's not physically close," Greenwood, a business professor at George Mason University, told Ars. The advent of collaborative robots really changes this dynamic for two reasons: one, it is physically close, and second, the robot has the ability to dictate the pace of work.

Many jobs in fulfillment centers also often involve "dull, dirty, and dangerous" tasks that people may prefer to outsource to robots. According to the US Occupational Safety and Health Administration, injury and illness rates for workers in the US warehouse sector are typically more than twice the average for all industries.

"You have almost 2 million people working in the warehousing sector in the United States, and it's characterized by very high turnover and a very high rate of workplace injuries," Greenwood said.

But robots are still a long way from being able to perform many different tasks that human workers can do. So Amazon still has a huge workforce working alongside the company's growing fleet of wheeled robots and robotic arms — and when humans are expected to keep pace with their robotic counterparts, the workplace can get complicated.

Amazon has been rapidly building its robotics fleet since its landmark decision to buy robotics company Kiva Systems in 2012. Kiva was one of the first companies to pioneer wheeled mobile robots that bring shelves of items to warehouse workers, rather than requiring people to constantly pull items from shelves.

Since the acquisition, Kiva has been rebranded as Amazon Robotics and has focused on deploying a suite of mobile robots and robotic arms that Amazon manufactures primarily in-house at its Massachusetts facility. This year, the company also announced that it will build additional manufacturing plants in Austin, Texas, and Greenwood, Indiana, to produce more robots in the future.

Scott Dresser, Amazon's vice president of robotics, told Ars: "There's a theme of how we can help our colleagues and employees with challenges, and how we can make our customers' lives better through these technologies." "Every step of the way, [we're] tackling a tough or deep set of technical challenges to achieve that goal."

Most of Amazon's million-strong robot fleet still consists of those mobile, wheeled robots. The company claims that the Sequoia robotics system can identify and store items in fulfillment centers up to 75 percent faster by coordinating groups of these robots. The Hercules and Titan robots, about the size of a pool table or a baby grand piano, respectively, roam the warehouse aisles to retrieve tall fabric shelving units called pods packed with plastic boxes and bring them to human employees at various workstations.

Such mobile robots typically follow the lines of a grid of tags placed on the ground and are placed in enclosed areas away from human employees. But the newest mobile robot, Proteus, has sensors that allow it to safely roam around humans and other objects in a more free-roaming role.

Those moving robots that navigate the narrow lanes between rows of storage pods can sometimes create robotic traffic. That's why Amazon started coordinating its bots using a set of AI-based models called DeepFleet, which can help predict potential bot congestion in advance.

Recently, Amazon has deployed a growing number of robotic arms that can sort items stored in inventory pods and place them in carts or on top of mobile robots, handling the many different products that Amazon customers order when they shop online.

The company expects to more than double its existing fleet of robotic arms this year, Amazon CFO Brian Olsavsky said on a July earnings call. When pressed for an explanation, Ares described the current number of robotic arms deployed as thousands in the single digits.

Amazon first purchased third-party robotic arms to help lift and stack heavy pallets and bins in its fulfillment centers. But the first robotic arm, built by Amazon Robotics to separate envelopes and mail boxes from clutter, began testing in 2021 and will begin operating in 2022 at a factory in Lakeland, Florida.

With Robin, it was really the first attempt to take a bunch of customer orders and build advanced robotics to identify the items in that scene, to strategize about how to use them. To take them and make a motion plan that would be successful in putting it on one of our mobile drive units, Dreser explained.

This robotic arm uses eight suction cups to pick up individual packages. But before any action, the robot software analyzes the top-down view of the conveyor belt full of packages and quickly ranks the possible selection actions.

In a 2023 research paper, Amazon's robotics team explained that it had developed several selection strategies for the bots to initially follow—but each had its own limitations. One strategy that prioritized the selection of larger packages encountered challenges when smaller packages were placed on top of or overlapped with larger packages. Another strategy, which sought to pick packages near the top of piles, failed in practice to recover "inaccessible picks" and partially buried packages.

Improvements came from replacing such manual algorithms with an artificial intelligence approach called machine learning. The machine learning model was first trained on more than 394,000 choice actions from deployed robotic arms to estimate the probability of success, and then used an online choice ranking to prioritize the optimal actions for the robin robots.

The best machine learning method in that experiment achieved a success rate of 93.73% compared to 92.28% success for the best performing traditional algorithm. This may seem insignificant at first glance, but it meant that the machine learning method resulted in about 16,500 fewer failed selections than the traditional method over a work period of more than 1 million total selections. This is important for an operation like Amazon that prioritizes speed and efficiency at scale.

Amazon's robotics team also demonstrated how such a machine learning approach could reduce the failed pick rate by 20 percent in a test involving 2 million picks by robotic workstations that resembled Amazon's Robin fleet operation. They presented this work at the International Symposium on Experimental Robotics in July 2025.

Cardinal was a robotic arm that could place 50-pound packages into wheeled carts that would eventually be brought to Amazon trucks at loading docks. The robots first began operating in November 2022 at a facility in Nashville, Tennessee.

"Think of it as a game of Tetris: You sort packages, and you put them in the cart, and you want to be able to pack them as densely as possible," Dresser told Ars. Dense containers are great from a sustainability perspective and from a shipping cost perspective.

Cardinal builds on lessons learned from Robin's advancements in package picking and handling, while adding the ability to densely pack customer orders to carts. Amazon has just started deploying Cardinal in larger numbers and will continue to do so over the next few years, Dresser said.

Robin and Cardinal can handle multiple package changes. But the newer Sparrow robotic arm has been developed to tackle a more challenging task - managing the hundreds of millions of products of various shapes and sizes that can fit into customer orders.

In particular, Sparrow focuses on merging items from partially packaged containers to create dense packaging devices. As items are selected while fulfilling customer orders, Sparrow continuously refills containers.

"When we think about how artificial intelligence is applied to the physical space in robotics, Sparrow was really the first place where we really applied the basic models that enable us to deal with this high diversity of products," Dresser told Ars.

Dreser explained that such basic models have helped Amazon's robotics team effectively identify and classify hundreds of millions of items in order to develop new robotic strategies to understand them. They also eliminated the need to manually code each item.

Recently, Amazon's robotics team developed a basic model for robotic perception that can handle new and never-before-seen products that may end up in Amazon's fulfillment centers. The team also developed a generalized action model compatible with different types of robotic arms that may use different combinations of suction and finger grasping.

"If you need to build ad-hoc perception models for every task and visualization like this, it's not scalable," said Kapil Katal, senior director of applied science at Amazon Robotics, during an Ars presentation at Amazon Robotics. The Robots and Intelligent Systems conference was held Sept. 29 in Pittsburgh.

"So one of the initiatives that we did in our team was to design a basic model that could work for any new incarnation of the robots that we were going to use, and for all of the perceptual tasks that we would have to perform on them," Cathal said.

Most robotic arms rely primarily on vision. But Amazon's newest robotic arm, the Vulcan, is the first to combine vision with force feedback sensors capable of providing a sense of touch. This enables the robot to accurately pick or place items in fabric-covered inventory pods.

"You have to be able to bring an item in, create space or pull something out," Dreser said. "It's really hard to do these things today with traditional non-touch systems, like the suction-based systems that are part of Sparrow."

Vulcan will initially operate at Amazon's fulfillment centers in Spokane, Washington, and Hamburg, Germany, beginning in 2025. It's not yet widely deployed, but Dresser said the company expects to ramp up use of the robotic arm in 2027.

According to the company, each new robot has been beneficial to Amazon's workforce. For example, Vulcan focuses on picking and storing items that are in the top rows of inventory pods, which would otherwise require one to use a stairwell, along with moving items that are just above the floor. Enables employees to focus on picking and stocking items that are more easily accessible.

But investigations by labor advocacy groups, journalists and the US Congress have painted a less rosy picture of Amazon's automation drive over the years. The Strategic Organizing Center (SOC), a coalition of labor unions based in Washington, DC, released reports in May 2021 and April 2022 showing higher injury rates in Amazon's robotic warehouses compared to non-robotic warehouses. The analysis is based on injury data reported by companies to the US government from 2017 to 2020.

"Amazon likes to talk about its latest robots because it doesn't want to talk about the tens of thousands of Amazon workers who are injured every year because of Amazon's obsession with maximizing speed and minimizing costs," Joanne Moriarty, director of strategic research and campaigns at SOC, told Ars in an email.