How Amazon scientists are reinventing delivery

These teams are using AI and machine learning to reduce our carbon footprint, all without compromising speed.

Ever wondered what happens when you click “Buy Now” on Amazon.com? You probably expect your stuff to arrive quickly, but that speedy dispatch is also part of an advanced strategy that helps Amazon reach millions of customers every day, all while reducing our carbon impact. 

Reducing carbon emissions in Amazon’s transportation network is key to curbing our operational footprint. As a result, we set a goal to put at least 100,000 electric delivery vans on the road by 2030, and we've rolled out more than 35,000 across our global operations. It’s exciting progress, and it wouldn’t be possible without the teams supporting the charging infrastructure and routing technology.

That’s where Amazon Principal Research Scientist André Snoeck comes in: Every day, he and other scientists on the Amazon Delivery team use operations research—a branch of math to help optimize complex systems—and other methods to scale Amazon's electric vehicle (EV) infrastructure and improve how packages get to your door.

Scaling EV infrastructure  
 

Snoeck and his team help manufacturers design vans tailored to Amazon's routes, including the right battery size for different geographies. To get those specs right, the team uses machine learning—a subfield of AI that finds patterns in data—to predict how much energy a route will use reflecting route patterns, geography, and seasonal variations in weather.
 

“Our team has spent years on one deceptively simple question: How much energy will an EV actually use on a delivery route?” Snoeck said. “Two routes of the same distance can use different amounts of energy. Getting that right is what makes electrifying a fleet this big possible.” 

Snoeck and team also developed a model that makes suggestions about when and where vehicles should charge based on the region. This helps ensure vehicles are ready to go when they deliveries begin each day.

Optimized route planning  
 

The team also uses modeling to create better routes for the Amazon vehicles.
 

Scientists refer to the “last mile” as the segment of the transportation network that involves bringing your package from a delivery station directly to a customer's doorstep. Planning these last mile delivery routes at Amazon's scale is another staggering math problem.

"For a single delivery station, we're assigning thousands of packages to hundreds of vans. The number of possible combinations is larger than the number of atoms in the universe," Snoeck said. "That's the scale of the problem our team of scientists works on every day."

The work that the Amazon Delivery teams are doing has real results. By the end of 2024, Amazon had reduced carbon emissions per shipped unit by approximately one-third since 2019. The teams’ efforts contributed to this achievement.

“We’ve spent years working to make Amazon deliveries more sustainable,” said Julian Pachon, global science director and chief scientist for Amazon Delivery. “Our team of scientists, who genuinely care about the outcome, are creating impact beyond simply building routes—we’re proving that a delivery network this large can run on clean energy. And if what we learn helps the rest of the industry, that’s a huge win for the planet.” 
 

Learn more about Amazon’s sustainable transportation strategy.  
 

Sign up for our monthly newsletter to get Amazon sustainability updates sent directly to your inbox.