Today, two forces are accelerating faster than anyone predicted. The first is climate change, with the world hitting record-breaking temperatures, and the second is indisputably AI. For AI researchers who study climate change, it’s an extraordinarily humbling and simultaneously promising time. From the United Nations to scholarly journals, AI is being hailed as a game-changing and revolutionary tool in society’s efforts to address the climate crisis.
Part of why climate change is difficult to address is because it comes from so many sectors of our economy. Tackling the climate challenge will require changes to our transportation systems, our ecosystems, electricity and construction, and how we operate heavy industries like cement and steel – all while adapting to the changing climate that we find ourselves living in and to the intensification and increased frequency of natural disasters due to climate change.
AI makes it possible, for the first time, to model the extreme complexity of our climate solutions. The climate scientist Richard CJ Somerville has noted, “There is no silver bullet that solves all the challenges of climate change, but there is a lot of silver buckshot.” By this, he means that a combination of diverse methods—including emissions reduction, which remains by far the cheapest way of lowering carbon in the atmosphere—will be our best path forward. Rather than wait for long-term projects to bear fruit, we should take meaningful climate action right now.
That sense of urgency is what motivates companies like Google to set ambitious goals of reaching net-zero emissions across the company’s operations and value chain by 2030 and running on 24/7 carbon-free energy on every grid where they operate. Kate Brandt, Google’s Chief Sustainability Officer, and her team are responsible for turning that goal into reality. “I’ve committed my life’s work to help accelerate progress on climate action, “ says Brandt, “and I believe that AI is a complete game-changer for what we are trying to do and for the planet as a whole.”
She is not alone in that belief. Yossi Matias, Head of Google Research, who leads work in AI and state-of-the-art technologies, including many of the company’s climate initiatives, also believes that addressing these planetary-scale challenges requires, first, identifying the issues, and second, developing urgently needed tools to help mitigate, adapt to, or prepare for environmental crises.
Taking Big Swings
Google’s AI-powered climate solutions—virtuoso feats of coordination and data analysis—point toward entirely new clean-energy paradigms and hold potential for world-altering breakthroughs. “Today, we are in a golden age of research, where the world’s biggest challenges and opportunities are accelerating the pace of research and motivating the development of breakthrough technology,” says Matias. “With remarkable speed and efficacy, AI is enabling us to solve problems that in the past would have seemed impossible.”
Flood Forecasting
One of the most notable examples is the problem of flood forecasting. “When I asked the experts,” Matias says, “the general consensus was that predicting well where floods are going to occur is too difficult a problem to solve.” What makes flood prediction such a complex problem? It’s the number of variables involved, along with the ways in which they interrelate according to the laws of fluid dynamics. These variables begin with the action of precipitation on the portion of a river on which it falls—the quantity of rainfall; the volume, angle, and speed of its descent; the type of soil through which the river flows; the vegetation in and around it; and the topography of the river itself.
These predictions need to be reliable, both to earn the public’s trust and to spare people from costly evacuations of their families and their livestock. They are difficult even when there’s a well developed system for data recording. And in many regions, this historical data may be scarce or nonexistent. Google had to create a breakthrough global model that could analogize one place and situation to another. “In a way you can think about it loosely like a language model that generates new text based on all the text and examples it has seen, but in a very different domain,” Matias says. The model “enables a scalable solution, and adds a level of equitability,” he explained, because some data-scarce countries have suffered disproportionately from global warming.
A pilot program in India serving 10 million people was expanded to all of India and Bangladesh, and then further developed into a system that helps to predict flooding up to seven days in advance in over 100 countries encompassing 700 million people. Google’s success here, Matias says, “makes me really optimistic that there are other notoriously difficult problems for which we could actually drive research and do something about it.”
How to Reduce Emissions
Human activity creates 59 gigatons of carbon equivalent emissions annually, according to UN’s Intergovernmental Panel on Climate Change. Every single viable climate scenario that the IPCC has modeled—i.e., every scenario that would keep temperature rise below 2 degrees Celsius by 2100—depends in part on some form of carbon-dioxide removal.
AI could help reduce emissions by five to ten percent globally by 2030 simply by accelerating existing technology solutions, according to research co-produced by Google and Boston Consulting Group in late 2023. Five to ten percent of global emissions is roughly equivalent to the annual emissions of the European Union. “This is the decisive decade for climate action. We know the impact of climate change knows no borders – which is why our response needs to be global and it needs to be a collective effort,” says Brandt.
Project Contrails
Google’s Project Contrails is one of few active initiatives that mitigates global warming not on land or sea but in the stratosphere itself. Contrails—the white wake produced by airplanes—account for a surprisingly large proportion of air travel’s environmental impact: some 35 percent, according to a report from the IPCC. A certain level of humidity is needed for a contrail to form. That suggests a simple solution for air-traffic pollution: Avoid humidity. But that turns out to be difficult to predict. For Google’s engineers, the solution entailed manually creating a database from satellite imagery of existing contrails to teach a machine-learning algorithm to recognize new contrails and to predict contrail formation regions. In 70 American Airlines test flights, contrails were reduced by 54 percent—simply by adjusting altitudes a couple thousand feet.
Fuel-Efficient Routing
On the road, Google’s fuel-efficient routing initiative is an example of targeting the climate crisis in a way that yields immediate benefits. The system, which is indicated on Google Maps by a green leaf, presents drivers with the most fuel-efficient route to their destination. “In the background,” says Brandt, “there’s a very sophisticated AI model running that’s looking at all of the different variables, everything from traffic to the gradient of roads.” Companies are already using the function to streamline their supply chains. The initiative is now available in the U.S., Canada, Egypt, India, Indonesia, and more than 40 countries in Europe. Since launching in 2021, it has saved 2.9 million metric tons of greenhouse-gas emissions through the end of 2023—the equivalent of removing 650,000 fuel-based cars from the road for a year. With only a “slight change to how we get people from one place to another,” Matias says, “we can quite significantly affect global warming.”
Project Green Light
Another Google project that targets the impact of cars on the climate is known as Project Green Light, which emerged from the insight that intersections can have pollution up to 29 times higher than on open roads. “It reduces starting and stopping events, which is much more pleasant for the driver, and then it also reduces emissions by up to 10 percent,” Brandt says. Merely shifting a few seconds of green-light time from an east-west street to a north-south street can make a substantial difference in emissions levels. Green Light is now up and running in over a dozen world cities, with plans to scale to hundreds of others within the next few years. “The result has been surprising even to me,” says Matias, “that we can significantly reduce carbon emissions by a small change in scheduling—without any new infrastructure whatsoever and no dependency on people’s behavior.”
Green Light also overcomes a significant stumbling block to fighting global warming: the difficulty of engaging the general public in the developing and maintaining of sustainable habits. The technology does its work without requiring drivers to change their behavior in any way. Brandt compares it to her Nest smart home device, which maintains an energy-saving temperature in her home without the need for any human intervention. “The great thing about my Nest device is I don’t need to think about it, but it’s making an impact. In 2023, we estimate that Nest thermostats helped customers save more than 20 billion kWh of energy, enabling approximately 7 million metric tons of GHG emissions reductions,” says Brandt.
From Ambition to Action
Every day, there are new and innovative ideas brought forth to combat the climate crisis. The Department of Energy has supported research at Cornell University on a bacterium that, once incorporated into a biohybrid, can use solar energy to convert carbon dioxide into chemicals such as bioplastics and biofuels. Scientists working on solar power have proposed a project that would collect sunlight in outer space, where it’s both continuously available and 10 times more intense, and beam it down to Earth. In Georgia, researchers—having observed that older trees store carbon much more efficiently—are using gene editing to produce poplars that grow 50 percent faster.
“When we think about the opportunities to tackle the climate crisis,” Matias says, “obviously there are multiple approaches. One is to make our current systems more efficient. The other is to look for alternative approaches for energy.” To that end, Matias’s team is working on nuclear fusion—an alternative-energy source that has been a dream of physicists for a century.
Through a reaction known as ignition, fusion throws off more energy than is required to create it. In December 2023, scientists at Lawrence Livermore Laboratory in California achieved ignition for the first time in history by bombarding a peppercorn-sized fuel capsule made of diamond with 192 laser beams. In an August 2024 op-ed in The Washington Post, Steven Cowley, a professor of astrophysical sciences and laboratory director of the U.S. Energy Department’s Princeton Plasma Physics Laboratory, argued that AI’s predictive capacity has made possible many recent advancements in fusion. “I believe fusion might turn out to be AI’s ‘killer app,’” he wrote, “proving the value of AI to deliver truly world-saving innovations.”
While fusion is unlikely to happen at scale within the next decade, it’s worth noting that nuclear is another area in which Google is leveraging existing technology, having recently signed the world’s first corporate agreement to purchase nuclear energy from multiple small modular reactors (SMRs). “Of course we need to be thinking long term,” Brandt says, “but the world needs to get to net zero by 2050. Right now we have the opportunity to still really change the trajectory over these next five years.” The immediate goal, she says, is to reduce emissions for cities, individuals, and other partners by one gigaton—roughly the emissions of the nation of Japan—every year.
“I see promise in our collective ability to make progress,” Matias says of the moment we are in, “because great ideas can come from anywhere, the opportunity to address difficult problems is greater than ever. Human ingenuity is at the center of it all–advancing science and technology and applying it to the problems that matter most.” Project Green Light, for instance, grew out of a dinner-table chat between a Google engineer and his wife, while the team was sifting through many ideas to help address the climate crisis. “And research advancements in AI-based flood forecasting helped make significant progress in what seemed like an impossible problem to address,” says Matias. Through expert insight, intuition, and collaboration across fields, human beings are choreographing new steps to address climate change, leading a dance in which AI has become a crucial partner.
