Amazon said it’s applying a new artificial intelligence (AI) model, called DeepFleet, to coordinate its growing fleet of warehouse robots, which now tops 1 million.
The ecommerce giant introduced DeepFleet, which is a generative AI foundation model designed to optimize how its robots move across Amazon’s fulfillment facilities. Amazon said the technology should reduce congestion and improve routing, boosting robot travel efficiency by an estimated 10%. That, in turn, could accelerate order processing and shorten delivery times for Amazon’s millions of customers, who have come to increasingly expect rapid fulfillment.
What is Amazon’s DeepFleet AI model?
“Think of DeepFleet as an intelligent traffic management system for a city filled with cars moving through congested streets,” said Scott Dresser, vice president of Amazon Robotics, in the blog post. “Just as a smart traffic system could reduce wait times and create better routes for drivers, DeepFleet coordinates our robots’ movements to optimize how they navigate our fulfillment centers.”
The DeepFleet rollout coincides with another major milestone: Amazon said it recently deployed its one millionth robot, which arrived at a fulfillment center in Japan. That brought the number of robots closer to Amazon’s total workforce, which stood at 1.56 million full-time and part-time employees as of December 31, according to its most recent annual report.
Amazon ranks No. 1 in Digital Commerce 360’s Top 2000 Database. The database is how Digital Commerce 360 tracks the largest North American online retailers by their annual ecommerce sales. Amazon is also No. 3 in Digital Commerce 360’s Global Online Marketplaces Database. That database ranks the 100 largest such marketplaces by third-party gross merchandise value (GMV).
How DeepFleet was developed
Amazon says it developed DeepFleet using internal warehouse and inventory data, along with tools from Amazon Web Services (AWS). That includes Amazon SageMaker, the company’s cloud platform for developers to build and deploy machine learning models.
Dresser described DeepFleet as a practical application of AI focused on solving real-world logistics challenges, not just pursuing technology for its own sake.
“By reducing robot travel time by 10%, we’re not just improving efficiency — we’re creating tangible benefits: faster delivery times, lower operational costs, and reduced energy usage,” he said.
In addition to improving robot coordination, Amazon says DeepFleet allows it to store more products closer to customers — a shift that supports faster shipping and, according to Dresser, can help reduce costs.
Because DeepFleet continuously learns and adapts, Amazon expects ongoing efficiency improvements across its warehouses. Robots already assist in roughly 75% of Amazon customer orders, according to the company, and that number is likely to rise as the technology advances further.
“As DeepFleet learns from more data, it will continue to get smarter — driving deeper efficiencies, unlocking more selection closer to customers, and reimagining what’s possible in robotic logistics,” Dresser added.
More than a decade of robotics investment at Amazon
The DeepFleet rollout builds on Amazon’s 13-plus years of robotics investment.
Amazon acquired Kiva Systems in 2012 for $775 million, later rebranding the company as Amazon Robotics. At the time, the company introduced automated guided vehicles (AGVs) that carried shelves to human pickers.
“We started in 2012 with a single type of robot that could move inventory shelves across warehouse floors,” Dresser noted. “Today, we operate a diverse fleet of robots, designed to make our employees’ jobs easier and safer, and our operations more efficient.”
Among those robots:
- Hercules, which can lift and move up to 1,250 pounds of inventory
- Pegasus, which uses conveyor belts to handle individual packages
- Proteus, Amazon’s first fully autonomous mobile robot, which navigates around employees in open areas
- Vulcan, introduced in May 2025, which uses force-feedback sensors and AI-driven tooling to stow and retrieve items from tight spaces
According to Amazon, Vulcan is its first robot with a sense of touch. The company said it is especially useful for retrieving items from areas where workers would typically need to crouch or climb.
Robots and workforce development
While Amazon’s robotics efforts have raised concerns about automation replacing workers, the company said its investments are creating new kinds of jobs.
Since 2019, Amazon claims to have upskilled more than 700,000 employees through various training programs, many of which are focused on working with advanced technology.
Dresser said that Amazon’s new robotics-powered fulfillment center in Shreveport, Louisiana — part of its “next-generation” warehouse model announced in October 2024 — requires 30% more employees in reliability, maintenance, and engineering roles compared to older facilities. Powered by AI and equipped with 10 times more robots than a typical site, the center is designed to cut processing times by up to 25%, according to the company.
“The technology we’re building does more than move products—it’s transforming workplace safety and creating new career opportunities,” he said. “Our robots handle the heavy lifting and repetitive tasks, reducing physical strain.”
Amazon manufactures its robots in the United States and works with local suppliers, Dresser said. That allows it to maintain high-quality standards while creating a feedback loop between designers, manufacturers, and front-line workers.
Looking ahead, Amazon said it plans to deploy these next-generation robotics systems across existing facilities to modernize its network and drive efficiency at scale.
Do you rank in our databases?
Submit your data and we’ll see where you fit in our next ranking update.
Sign up
Stay on top of the latest developments in the online retail industry. Sign up for a complimentary subscription to Digital Commerce 360 Retail News. Follow us on LinkedIn, X (formerly Twitter), Facebook and YouTube. Be the first to know when Digital Commerce 360 publishes news content.
Favorite