AI is helping Amazon buildings improve energy and water efficiency

At one of Amazon's logistics facilities in Glasgow, Scotland, a single faulty valve in an underground pipeline belonging to the local water supplier was leaking.

Although the leak was unnoticeable to the human eye, Amazon’s AI-powered utility efficiency tool, FlowMS, was able to detect it.

Built using Amazon Web Services (AWS), FlowMS traced the underground water leak after it analyzed metering data and noticed the building was using more water than usual. It then alerted Amazon employees to the anomaly, and a team of engineers discovered the leak was likely coming from piping underground. Fixing the valve helped prevent 9 million gallons of water from being lost per year.

FlowMS is part of a suite of AI and science-based tools that are improving how Amazon makes its buildings and utility management systems more efficient. These innovations have already delivered impressive results—from catching utility meter issues, to reducing food waste and identifying energy leaks at loading dock doors.

“At Amazon, we’re innovating with AI to help us find new ways to decarbonize even faster, including inventing new solutions that continue to make our buildings more energy- and water-efficient,” said Kara Hurst, Amazon’s chief sustainability officer. “This is just one example of how Amazon is leveraging our decades of experience in AI development and sustainability to think big about decarbonizing our business and operating more efficiently.”

Using AI to make our buildings more sustainable

Buildings and construction account for 40% of the world's greenhouse gas emissions, according to the Carbon Leadership Forum. As part of our Climate Pledge, Amazon is focused on making our buildings more sustainable, including using practical AI-powered solutions like FlowMS, in addition to other technologies and innovations being developed within the company.

Monitoring key building systems with AI

Amazon BBAM system detects malfunctioning AC units in warehouses

In addition to FlowMS, Amazon's Decision Science & Technology team—comprising research and applied scientists working to improve maintenance and equipment operations—has built the Base Building Advanced Monitoring (BBAM) tool. It leverages AWS's machine learning tools Amazon SageMaker and Lambda to monitor our HVAC systems. This includes analyzing our HVAC operational data, energy consumption, and local weather to determine anomalies in the system's behavior. The tool can also identify issues like clogged filters, which can cause compressors to consume more energy and can even determine if a site is consuming more energy than expected due to shifting weather patterns.

Amazon sites that have onboarded FlowMS and BBAM are seeing strong energy efficiency improvements. At a fulfillment center in New York, FlowMS detected that the building appeared to be using five times more energy compared to other Amazon sites nearby. Employees discovered the utility meter was miscalibrated, and the building was actually using much less energy than the meter indicated. And in Spain, BBAM detected a malfunctioning air conditioning unit by comparing its cooling output to the expected demand based on weather conditions, like outside temperatures, in real-time. This early detection allowed the site management team to proactively address and resolve the issue before employees were impacted.

We’re starting to deploy the BBAM tool at fulfillment centers and delivery site dock doors—where our trucks load and unload—to alert employees when the dock doors are accidentally left open. This allows us to reduce energy loss, which at scale can lead to substantial energy savings.

FlowMS and BBAM are now being used at 120 Amazon sites, and by the end of 2025, we aim to expand the tools to more than 300 of our buildings worldwide.

Optimizing our grocery refrigeration

Diagram showing ARM detecting faulty refrigeration equipment through defrost cycle changes

Advanced Refrigeration Monitoring (ARM) is another AI tool Amazon developed that monitors and analyzes refrigeration units in real time in our fulfillment centers and helps maintain optimal temperatures for perishable goods. ARM analyzes the refrigeration units’ energy meters and complex data patterns to monitor product temperature levels and predict potential operational issues.

For example, if a piece of equipment is malfunctioning or using excessive energy, ARM flags the issue in real time to associates via instant message, and predicts which part of the equipment is involved—such as a clogged compressor, fan motor failure, or ineffective insulation.

In one of our sites in Spain, ARM detected a change in the defrost cycle pattern, which led our teams straight to the faulty equipment. This early detection prevented both significant food loss and an estimated 1,000 hours of equipment downtime.

We are currently using ARM throughout our grocery network in North America and across Europe, and by the end of 2025, we aim to expand to more than 150 sites, including in India.

Leveraging AI tools to make our buildings more efficient is one of many ways we’re working toward our Climate Pledge commitment of reaching net-zero carbon emissions by 2040.

Discover the full scope of Amazon's sustainability efforts and how we're working to decarbonize our global energy use.