How five key industries use AI, machine learning and the cloud to meet their sustainability goals

Technology innovation is helping some of the largest producers of greenhouse gases to drastically cut emissions. Here’s how they’re doing it.

While we can all take steps to cut our own carbon footprints through local efforts like reducing car trips and using LED bulbs, the nation as a whole won’t meet its obligations under the Paris climate agreement unless we start making gains at scale. To be sure, this will be challenging: The federal government last year announced a target of reducing emissions to half of 2005 levels by 2030 to meet the Paris goals.

For perspective, the dramatic covid-19 shutdowns of 2020 only reduced fossil fuel carbon emissions by about 5.6 percent, according to the United Nations. A similar reduction will be required every year through 2049 to keep the global average temperature increase below 1.5 degrees Celsius, notes Chrisopher Wellise, director of worldwide sustainability at Amazon Web Services.

“It’s going to require new technologies, large-scale efforts by business and governments, a transformation of the electrical grid and paradigm shifts in business models,” Wellise adds. “It’s a gargantuan effort on multiple fronts.”

The good news is that emerging technologies like artificial intelligence, predictive analytics and the cloud are giving the most carbon-intensive industries powerful tools to significantly reduce energy usage and emissions.

“It’s going to require new technologies, large-scale efforts by business and governments, a transformation of the electrical grid and paradigm shifts in business models.”
- Christopher Wellise,
AWS director of worldwide

Reducing carbon emissions usually comes down to achieving greater efficiency in the use of energy and materials. Improving efficiency requires detailed mapping and monitoring of very complex systems such as factories, electrical grids, HVAC systems and logistics routings. Increasingly, these environments are outfitted with sensors, and artificial intelligence excels at finding patterns in the huge amounts of data that these sensors collect, says Bratin Saha, vice president of machine learning services at Amazon.

The heavy computing work required can be done in the cloud, which requires less energy than computing done with on-premises servers.

“Finding efficiencies starts with collecting data about the system you’re trying to improve,” Saha says. “The cloud helps you collect, store and process that data at very low cost, and then you use AI and machine learning to gather insights. Ultimately, you’re becoming a lot smarter in your decisions, which translates into greater opportunities to realize those efficiencies.”

Here’s a look at emissions reduction targets for the five largest sources of greenhouse gases in the U.S. (based on EPA data*), and how companies in these industries are turning to the cloud for innovations like AI, machine learning and analytics to lower emissions.


Commercial Real Estate

Rise of the Smart Buildings
current emissions:
metric tons annually
2030 target emissions:
metric tons annually

The biggest emissions source for many businesses comes from their buildings, but 30 percent of energy used in commercial structures is wasted, according to the Environmental Protection Agency. The good news is that solution providers are stepping up with innovations that can help solve this problem. Carbon Lighthouse, for example, uses AI and the cloud to help building owners identify profitable energy efficiency projects. That identification process typically takes weeks, but Carbon Lighthouse can do it in minutes. The solution relies on data collected from over 100 million square feet of building space.

Carbon Lighthouse has also developed an ongoing operations solution that uses sensors to monitor ventilation, heating and AC systems at individual buildings. This system helps building owners take precise action when energy usage unexpectedly remains outside of normal bounds. Carbon Lighthouse recently helped one large office building that was already Gold LEED-certified save 25 percent of its energy spending.

“Access to the data helps you save the world. It’s all about scale, and the need to move quickly.”
- Nikhil Daftary,
executive VP,
Carbon Lighthouse

AI and machine learning help Carbon Lighthouse’s operations software react quickly to unpredictable changes in weather and building occupancy, says Nikhil Daftary, an executive vice president at the San Francisco-based company. When identifying energy efficiency projects, AI helps Carbon Lighthouse work through thousands of simulations in a few seconds. “Without access to the data, you can help buildings,” Daftary says. “But access to the data helps you to save the world. It’s all about scale, and the need to move quickly.”


Electricity Production

Intelligence for the Grid
current emissions:
metric tons
2030 target emissions:
metric tons

The old electrical grid transmitted steady streams of power from a small number of big plants to millions of consumers. The shift to renewables will require a much more flexible grid, drawing energy from diverse sources like home solar systems and EV batteries and intermittent sources like wind and solar. The “smart” grid requires sensors, to collect data on energy supply and demand; the cloud, to store that data; and AI and machine learning to analyze it.

UK-based utility Octopus Energy offers “smart tariffs,” which reward customers with lower rates if they shift their demand to off-peak hours. But that doesn’t mean they have to shower at midnight, says Greg Jackson, founder and CEO at Octopus Energy. That’s because Octopus is using neural networks—a form of machine learning—that are trained on billions of rows of smart-meter data in the cloud to examine energy consumption at different times of the day. The resulting insights allow Octopus to match up consumption patterns with highly variable grid pricing.

“Machine learning lets us look at real-time generation and then shift consumption around to meet demand.”
- Greg Jackson,
CEO, Octopus Energy

With its smart tariff ‘Intelligent Octopus’, Octopus is continuously figuring out the best and cheapest times to charge its customers’ electric cars, saving customers money by doing so while helping to balance the grid. What’s more, the system learns each customer’s usage patterns to optimize energy usage while also ensuring the car is charged when the customer needs it.

The impact can be significant: Businesses that charge up their delivery vans at optimal times can save 80 percent on their charging costs, says Jackson. “Machine learning lets us look at real-time generation and then shift consumption around to meet demand,” Jackson says. “You don't need to think about it. You just plug in your car, and we'll create an optimized charging schedule based on real-time forecasts for the grid price and renewable generation.”



Paring Emissions with Digital Twins
current emissions:
metric tons
2030 target emissions:
metric tons

One of the hottest developments in manufacturing is the digital twin—the software representation of a physical asset. It could be a single product, a machine, a factory or an entire supply chain. Digital twin simulations are constantly updated through information received via sensors monitoring the physical asset. “The model is being trained over time; it’s learning from experience,” says Saha.

Manufacturers can use digital twins to save energy and cut emissions by operating more efficiently. This can be done by simulating every aspect of operations and maintenance. By running simulations, digital twins predict when a machine will need servicing—reducing breakdowns, downtime and time spent prototyping and testing.

For example, a wind-turbine operator could use a digital twin to predict when a blade might fail without having to make time- and energy-consuming on-site inspections of the physical equipment, which could be located in a far-flung location like the North Sea. This could allow the operator to optimize preventative maintenance schedules to ensure uninterrupted supplies of wind energy to its customers.

“The model is being trained over time; it’s learning from experience.”
- Bratin Saha,
vice president of machine
learning services, Amazon

Manufacturing services provider GE—which uses the AWS cloud to power its digital twin services—estimates its clients achieve outcomes up to 75 percent faster using digital twins.

“Virtual twins drive sustainability and the circular economy at speed and scale,” according to a recent report by Accenture. “They help companies reduce their costs, resource use and carbon footprint.” Digital twins can be used for everything from buildings to manufacturing to transport, potentially saving 7.5 gigatons of carbon emissions by 2030, Accenture estimates.



Harvesting the Data
current emissions:
metric tons
2030 target emissions:
metric tons

Climate change poses a threat to agricultural productivity, and agricultural activity itself is a significant source of emissions and resource depletion. Many farmers lack insight into soil conditions, leading to suboptimal use of energy, water and resources.

Israel-based CropX, an agricultural analytics company, supplies soil sensors that provide real-time data on soil moisture, salinity and temperature. The soil data is combined with satellite imagery, crop models and other inputs. It’s all analyzed by AI-based algorithms to inform farmers if they’re using the right amounts of fertilizer, pesticides or water.

“We have seen savings on fertilizers, energy and labor.”
- Matan Rahav,
director of business development,

In experiments, CropX has demonstrated more than 40 percent water savings, with a 10 percent yield increase. “We have seen similar savings on fertilizers, energy and labor because farmers spend fewer resources on travel and equipment,” says Matan Rahav, director of business development at CropX.

Rahav says the cloud has proven indispensable to CropX’s ability to run a large data analytics platform with millions of data points each day, and for the company’s ability to expand rapidly.



Hitting the Road with AI
current emissions:
metric tons
2030 target emissions:
metric tons

Transportation logistics is a huge industry with massive inefficiencies. Every year, truckers in the U.S. travel more than 95 billion highway miles, but 40 percent of those miles are done with an empty truck, representing a costly waste of time and fuel.

AI is increasingly used in logistics, predicting product demand by location so that goods are warehoused near where they will be delivered. AI can also help plan distribution networks by finding the optimal location of warehouses and determining transportation routings for vehicle fleets.

With billions or trillions of routing options—and the best options constantly changing—vehicle routing optimization is perfectly suited for cloud-enabled AI. Anyone who has used a mapping app knows how AI can plan a driver’s route in real-time. Route optimization software expands that capability to entire fleets while factoring in depot and customer locations, vehicle load capacity, driver shift duration and more. The fuel savings translate directly into a lower carbon footprint.

“We needed a platform where we could learn a lot.”
- Moritz Siuts,
director of engineering,

Digital innovation also is helping to power eco-friendly ridesharing services. For example, Volkswagen-owned MOIA uses several AWS technologies, including IOT Greengrass, to manage its fleet of connected vehicles in cities across Germany. Usage data is collected from the vehicles and fed back to MOIA staffers, who are continuously optimizing operations to improve service and save energy. “We needed a platform where we could learn a lot,” says Moritz Siuts, MOIA’s director of engineering. “We can adjust the platform and our architecture as we learn how our customers behave.”


The Sustainable Cloud

Advanced technologies like AI, machine learning and predictive analytics can help organizations use resources more efficiently, but the computing required for these technologies can also be energy-intensive, notes Wellise. Fortunately, moving to the cloud can be a big energy saver.

According to a report by 451 Research moving on-premises workloads to Amazon Web Services can lower workload carbon footprint by 88 percent for the median surveyed U.S. enterprise data centers. That’s due not only to more efficient use of aggregated computing capacity, but also to AWS’s own innovations in areas such as data center cooling, Saha says.

“Sometimes people think about AI as the future, but people are using AI, machine learning and the cloud to become more efficient right now.”
- Bratin Saha,
vice president of machine
learning services at Amazon

“Sometimes people think about AI as the future, but people are using AI, machine learning and the cloud to become more efficient right now,” he adds. “It’s important to get started. You can do that now by moving your data assets to the cloud, unifying those assets, and then using AI to detect patterns in those data that allow you make smarter business decisions.”