• Amazon invests billions in warehouse robotics, deploying more than 750,000 robots globally.
• Tye Brady, Amazon Robotics's chief technologist, believes robots help people work more efficiently.
• AI significantly boosts the potential of robotics in warehouses, he told BI in a recent interview.

Amazon has long been a leader in warehouse robotics, investing billions of dollars to automate some of the picking and sorting jobs at its fulfillment centers.

But its work in robotics is just getting started, Amazon Robotics Chief Technologist Tye Brady told Business Insider in a recent wide-ranging interview.

Amazon now has at least 750,000 robots roaming around its sprawling network of warehouses. It has been deploying robotic arm systems called Sparrow and Robin that can quickly lift and move millions of products and boxes.

This wave of automation is helping Amazon deliver packages faster, and it could make the company more profitable. Morgan Stanley recently estimated this robotics push could save Amazon up to $10 billion a year by 2030.

According to Brady, the goal is to help Amazon's frontline workers do their jobs more safely and efficiently.

"We do technology with a purpose. And if that purpose makes sense in e-commerce and our material handling fulfillment systems, then we will do that as long as it improves the safety of our employees and their performance," he said.

Brady, who has been at Amazon for almost a decade, saw the potential for robots in the logistics industry early on. He was part of a group that formalized the Amazon Picking Challenge, a grassroots competition that helped bring together a community of robotics experts. Before that, he worked at MIT and Draper Laboratory.

Brady said breakthroughs in artificial intelligence give Amazon more reason to invest in this space. In 2022, the company launched a $1 billion Industrial Innovation Fund focused on supply chains and logistics. In August, Amazon hired the founders of startup Covariant, a robotics AI startup last valued at $625 million.

"The physical AI part is hugely important," Brady said. "AI has really revolutionized and transformed robotics because it allows us to have the mind and body as one."

Brady doesn't dispute these advancements will change jobs in the future. But he argues that it will benefit the workforce, creating more skilled jobs and giving people more time to focus on things that "really matter."

"Our future is in people and technology working together," Brady said. "If we have technology that allows us to be more capable in our jobs, that's a win."

This Q&A has been edited for clarity and length.

BI: Amazon has more than 750,000 warehouse robots? Will all these be Amazon's own designs or a mixture of internal and external machines and systems that you get from other places?

"What's really awesome is that 750,000 is for our drive units alone. And I'm very proud that those have not only been designed inside Amazon but also manufactured by Amazon. So we manufacture every one of those drive units in Massachusetts in one of our two facilities.

We have our own manufacturing lines where we actually build those locally, and we ship those globally. And we've created a lot of jobs because of that manufacturing capability. We have a great talent pool around here. We actually source a lot of material even locally as well, proving that you can do this cost efficiently and very performantly here in the United States, which is great. And that's the world's largest fleet of industrial mobile robots out there."

BI: Any sense of where that number's going in the future?

"I would imagine it's only going north. I can't share the number yet.

If you went into a large fulfillment building, you'd see 3,000 to 4,000 different drive units working in coordination with each other. You would see 10, 20, 30 Sparrow arms working in coordination, picking up things. You'd see tens of Robin work cells picking up boxes. You'd see a couple hundred Proteus drives moving what we call go-karts around. They're everywhere. And you're also going to see a lot of people as well. So it's the idea of people and machines working together because when we do our robotics, we're extending and augmenting human capability."

BI: Why name your robots after birds?

"The designers get to pick whatever names they want, but it started with the first bird, which was Robin, which stands for 'robotic induct.' So our team said, hey, we're going to call this Robin. And then a whole bunch of birds started to come on."

BI: How much resources and money is Amazon putting into automating warehousing and logistics? Is it all focused on the kind of e-commerce operation, by which I mean warehousing and logistics, or also delivery and other parts of Amazon that we haven't thought of?

"To put a figure on it, would be nearly impossible for me to do. You could see the scale is if we're going to open an industrial fund of $1 billion, then we're serious about this. And we're seeing the real-world benefits for our employees today.

AI has changed the game of what products need to go where. We start even before the customer makes their order on Amazon.com. And then we offer a delivery promise that's backed by our amazing people and our robotic systems. And that delivery promise is a function of what we call 'first mile,' where we store our goods, our 'middle mile,' where we help sort those goods to get them to the right locations. And then our 'last mile' deliveries. And robotics is affecting all three of those phases significantly. And AI has changed the game for us and the robotic systems that we have continue to introduce more and more efficiencies in that process."

BI: Is there any task inside the warehouse and logistics process that will never be automated because they are too complex or something?

"Just plain old, simple common sense will never be automated. Humans have an amazing ability to adapt and understand what's going on and an amazing ability to use tools as long as they're available. As long as they're intuitive and natural and actually provide value, we'll use tools to do the job better and more efficiently. And for our case, it's safer.

I'm a systems engineer, and what we want to do is we want to leverage what people are good at. We're good at problem-solving, using common sense and reasoning, generalization, and creative thinking. We're amazing at that. I don't want to throw that away. Machines are really good at crunching numbers, assessing databases, and moving with precision. Let's bring those together in order to do the task at hand. And for us, that's moving our packages for our customers."

BI: Any thoughts on the resurgence of investment in robotics lately?

"The 'physical AI' part is hugely important. I think we've done a pretty good job of building the body in robotics. But let's think about the mind and the body, and AI has really revolutionized and transformed robotics because it allows us to have the mind and body as one.

We're doing things not in the digital world. We're doing real-world things applied in the analog world, the world that we live in. And there's no better example of that than what we're doing in Shreveport, Louisiana. We have 10 times the amount of robotics under that roof. The processing time is 25% faster. We can reduce what we call the cost to serve by 25% as well. We can pass those cost savings along to our customers and we're creating brand new jobs. That's a win all the way around because it's great for customers first and foremost, and it helps people do their jobs better. So, it's not a replacement strategy; it's an augmentation strategy.

We do technology with a purpose. And if that purpose makes sense in e-commerce and our material-handling fulfillment systems, then we will do that as long as it improves the safety of our employees and their performance.

That's what we want to do, and we've done that. I can give you lots of examples, whether it's in our Sequoia system, our Proteus system, our Sparrow system, or our Robin system. All those systems allow for safer employees and more efficiency in our fulfillment processes."

BI: What's the next challenge? What's the next frontier specifically in your e-commerce logistics world?

"First of all, physical AI is here and proving very useful, but we're just getting started. And I am really excited about the scale at which we operate, which is mind-boggling. And that scale allows us to advance the technologies of physical AI. That's really exciting. When we ship billions and billions of packages every year, that makes you have to do it right because you can't be 99% good because a 1% exception rate over billions of items is quite a lot of things that you would then have to specially handle."

BI: Would Amazon ever consider providing a lot of this physical AI robotic technology as a service to other companies? Or are you already doing that?

"We are doing that in a sense through AWS. Our physical AI systems have the same toolkits that hundreds of thousands of our customers have available to them, and they're using them, so we're seeing a lot of growth there. So it's an incredible system, how AWS has structured it and made it available to all [through Bedrock].

We're always finding ways to evolve retail. That's what we do. That innovation mindset, that experimentalist mindset, allows us to achieve what we have. Whether it's helping our productivity, this idea of mind and body being unified inside our robotic systems is really exciting. These advancements that we're seeing, especially with Covariant, helping drive the mind, that is really exciting. How automation is empowering people. Our mission continues to be, I want to eliminate the menial, the mundane, and the repetitive. I want to eliminate those tasks and allow people to focus more on what matters."

BI: Tell us more about Covariant.

"It's a really broad brush approach to foundation models and how they should be applied, but with specific needs. Whether it's path planning or perception systems for robots to understand their world, we're seeing some significant advancements in those areas.

Generative AI is fueled by data and we have lots and lots of data from our objects. We offer the world's largest selection of goods to our customers out there and are the generative AI models that really changed the game when it comes to efficient planning. The movement of robots when it comes to understanding the scenes that the robots are in. And systems like Proteus, for example, these are 'in the wild.' They are around people, they're not behind fences, they're moving in concert with people, and understanding the environment is really important in a quick manner. And their generative AI systems are changing the game for us."

BI: And that example for Proteus, that's in a warehouse, not following a preordained path, it's just out there?

"That's right. It's out there around people. People move around it. It moves around people. It's really, really cool to see, and I can show you just one example of how state-of-the-art it is. Typically a robot, when it sees maybe three, four, or five people gathered together, it'll just stop and wait for the folks to kind of disband. Or if you're at a cocktail party, when you're trying to navigate from one side of the room to the other, you know how hard that can be. The Proteus system actually is capable of that. So it doesn't just stop. What it does is it kind of nudges its way, never run into anything, but it has its way of signaling to people 'I would like to get through,' and then it can kind of navigate its way through that cocktail party in a very safe manner. That's really hard to do.

It's all about intent. We need to build in our models ways for people to understand the intent of the robot. I want to take a left turn, I want to move. So we have eyes and lights that allow a person to understand the robot's intent. But we also program the robotics to understand the person's intent. If they're walking, are they going to take a left turn, so we want to make sure that we keep the right buffer around them? Maybe they're going to stop really quickly. We have to be mindful of that as well. And our physical AI systems that are doing that today have really changed the game. When it comes to understanding and perceiving those scenes that are immediately around it."

BI: And Covariant helps with all these things?

"Absolutely. So it's a balance between what we call edge computing. So that's the software that we run directly on the machine itself, and then also our cloud computing that we have through AWS where we have these foundation models that are helping us extract what's a door and what's a fence and what's the dock look like and what's a truck, and all that knowledge we are assembling.

They're great. I mean, they're a leading robotics company for sure. Really excited about it. The way that they have done their models is kind of unique and we think that they can really help assist people. So that's a great partnership."

BI: Do you think robotics will at least slow down the massive growth of this human warehouse workforce from the past 10, 15 years or not?

"Inside our Shreveport building, we have 35% more skilled jobs. So the labor force is changing. We have also committed more than $1.2 billion in our upskilling pledge. We offer things like career choices. We offer things like apprenticeships and prepaid tuition programs for our employees inside our fulfillment centers for upskilling. That's really important.

It is really important for us to do this because we know jobs will change, and we want to keep those skills local in the community where we're creating those jobs. Our future is in people and technology working together.

And the perception of technology really matters. It really matters to people because if you are more willing to adopt technology and use it, then you get the benefits of technology.

And people, when they have technology, can do really amazing things. When you're in India, if technology allows you to have cleaner water, that's a win. That's a loss for nobody. That is a win. If you have technology that makes you healthier and safer, that is a win. Let's embrace that. If we have technology that allows us to be more capable in our jobs, that's a win.

It's exactly the way I feel about robotics. But instead of being in the digital world, it's in the physical world. Let's allow people to do great things to be more efficient. It's why I got into robotics. When you do robotics right, it allows us to be more human. And I mean physical AI, I mean smart, capable technology systems that allow us to be more human, more capable. If I can have a robot do my dishes for me, which I do, it's called a dishwasher, and that's great. It's such a good robot that we don't really talk about the robot, but what it does is very simple. It allows me more time to connect with my loved ones. The menial and mundane, go ahead and do it. Mow my lawn, do my pool, whatever it is, give me more time to be more human.

I'm not going to run out of things to do on the weekend. I can guarantee you that. There's no replacing me over the weekend. But if I can have machinery and physical AI systems doing these tasks for me that allow me to focus more on what matters at home, that's my loved ones, that's a total win.

I definitely believe that what we're doing inside Amazon, applying robotics and physical AI in many, many systems, is helping master the fundamentals of robotics, whether it's a movement system, a manipulation system, a sortation system, a storage system, an identification system, or a packaging system.

That's what we're focusing on. When we achieve those things, what's going to happen is that the world will see in this particular industry that robotics can really help people do their jobs, and that will start to transform other industries. You're going to see the healthcare industry pick up on physical AI. You're going to see agriculture pick up on physical AI.

And we are pioneering that field, which is as simple as storing goods, picking goods, moving goods, and bringing them to customers. But that context allows us to accelerate the development of robotics."

Do you work at Amazon? Got a tip?

Contact the reporter Eugene Kim via the encrypted messaging apps Signal or Telegram (+1-650-942-3061) or email ([email protected]). Reach out using a nonwork device. Check out Insider's source guide for other tips on sharing information securely.

• Amazon's warehouse robots have come a long way since it acquired Kiva Systems in 2012.
• Robots can now perform a variety of tasks alongside employees in fulfillment centers.
• They can transport packages, sort individual items, and lift heavy objects.

Amazon's fleet of warehouse robots has swelled to more than 750,000 — and it continues to grow both in size and complexity.

Robots now perform a variety of tasks in Amazon's fulfillment centers, transporting packages around busy workspaces, sorting and consolidating items into storage systems, and making fit-to-size packaging.

The company's robotics efforts started when it acquired Kiva Systems for $775 million in 2012. Kiva's automated guided vehicles navigated by following barcode stickers placed on the floor of a warehouse. More than a decade later, Amazon now has more than 16,000 people working on robotics as its technology has become more sophisticated, including Proteus, a mobile robot that can move autonomously.

Investing in robotics helps Amazon accomplish its goal of getting packages to customers as quickly as possible, Tye Brady, chief technologist at Amazon Robotics, said in a recent interview with Business Insider. Robots also create efficiencies that will help Amazon save money — some $10 billion a year by 2030, Morgan Stanley estimated in a recent research note.

"We can have faster delivery times because of the work that we've done in robotics," Brady said. "We can also pass on a lower cost. And we're creating thousands and thousands of jobs because of the work that we've done in robotics."

In August, Amazon hired three of the robotics startup Covariant's founders and licensed some of its foundation models to bring flexibility and fluidity to its robotics. The company also launched the Amazon Industrial Innovation Fund in 2022, investing in companies innovating in emerging tech like robots. That includes an investment in Agility Robotics, which makes a bipedal robot called Digit that Amazon is testing in fulfillment centers.

"We have a commitment for more than $1 billion for our startups, in order to help the startups and the community raise the capital that they need in order to do some of these great ideas that we think are going to help our customers," Brady said.

Here are some examples of the most advanced robots working in Amazon's fulfillment centers, and a bit about what they do:

Proteus

Unlike Amazon's earlier mobile robots, Proteus is fully autonomous. It uses sensors to navigate around objects in its path. It works more closely alongside Amazon employees and is not restricted to fenced-in areas.

Proteus travels under package carts and then transports them to the fulfillment center's loading dock. Brady compared its skills to how humans make their way around a crowded cocktail party.

"Typically a robot, when it sees maybe three, four, or five people gathered together, maybe they're all talking in a circle or something like that, it'll just stop and wait for the folks to kind of disband," he said. "When you're trying to navigate from one side of the room to the other — the Proteus system actually is capable of that. It doesn't just stop."

It launched in a Nashville fulfillment center in 2022.

Sparrow

Sparrow is Amazon's first robotic arm to handle individual items rather than packages. It uses computer vision and AI to pick items from containers and place them into totes. It's roughly the size of an elephant trunk and capable of picking up more than 200 million different items, Brady said.

It was first introduced in a fulfillment center in Richmond, Texas, in 2023.

Sequoia

Sequoia combines AI, robotics, and computer vision into one storage solution. Totes containing items are stored in a vertical platform system called a gantry. Robots then move those totes to ergonomic workstations where employees pick products to be shipped to customers. Once an employee picks items from the totes, Amazon's robotic arm, Sparrow, retrieves any remaining items and then consolidates them so that full totes can be returned to storage.

Amazon says Sequoia allows the company to identify and store products as much as 75% more quickly than before. It was first introduced in a Houston fulfillment center in 2023 and is at the center of Amazon's "next-generation" fulfillment center in Shreveport, Louisiana, where it can handle more than 30 million items.

Hercules

Introduced in Sumner, Washington, in 2017, Hercules is one of Amazon's earlier mobile robots. It moves pods of items around a fenced area of the fulfillment center by using its 3D camera to reference markers on the floor. That camera also helps it differentiate between the various things that may be in its way, like people, pods, or other robots.

It can lift pods that weigh up to 1,250 pounds.

Titan

Titan operates similarly to Hercules but can lift double the weight, up to 2,500 pounds. It made its debut in San Antonio in 2017 and is typically used for bigger items.

Pegasus

First introduced in 2018, Pegasus is a cousin to the Hercules robot that incorporates a conveyor belt on top of the drive unit.

It takes finished packages from employees to a sorting area that's determined by the recipient's ZIP code. Using a route provided by Amazon's centralized planning system, it navigates until it finds the correct location, then drops packages through a chute in the floor to a loading dock below.

Pegasus works in conjunction with Amazon's robotic arm Robin.

Robin

Robin was Amazon's first robotic arm, introduced in Lakeland, Florida, in 2022. It works in tandem with other robotic systems, picking up packages from conveyor belts and placing them on Pegasus mobile robots that bring them elsewhere in the fulfillment center. It also handles damaged packages.

Cardinal

Cardinal, a robotic arm that was introduced in Nashville in 2022, lifts packages and sorts them into the appropriate cart before they are brought out to be loaded onto a truck. Robin and Cardinal both use suction to handle packages weighing up to 50 pounds.

Xanthus

Xanthus is essentially a lighter version of Pegasus. It also has upgraded sensors that allow it to navigate obstacles from further away than Pegasus can.

Also called "X-bot," Xanthus is less expensive to produce than its previous iterations.

Xanthus was initially used for sorting in 2019, but Brady has previously said the company sees it being used elsewhere due to its flexibility.

Packaging Automation

The packaging automation system uses sensors to measure an item and then create packaging that is properly sized for that item, eliminating excess materials. It previously made plastic bags but now makes paper bags. It was first introduced in Euclid, Ohio, in 2023.