AI start-ups take aim at climate change

While growing up in Normandy, northern France, Yanis Souami became enamoured with the ocean. One of his earliest memories was “going to the sea, snorkelling and marvelling at the fish and the wildlife”, he says.

And, today, the now 42-year-old entrepreneur runs Sinay, a French start-up that uses artificial intelligence to analyse ocean data — on shipping movements, weather patterns, and air and water pollution — to help the maritime industry streamline its operations and reduce its environmental impact. One of its services enables users to monitor both underwater noise, from sources such as shipping or pile-driving, and the presence of marine life that might be adversely affected by it.

“There is a lot of impact of human business on ocean biodiversity,” Souami says. “We can combine biodiversity conservation with business efficiency using the best available information, partly thanks to AI.” He points out that preserving the world’s oceans — which produce about half of the Earth’s oxygen — is a crucial part of the fight against climate change.

Sinay is part of a constellation of European start-ups that are using AI and machine learning to help businesses both cope with climate change, and minimise their contribution to it.

Applications such as emissions monitoring, recycling management, and predictive infrastructure maintenance are attracting significant interest from big companies — and investors. Last week, for example, Danish start-up Electricity Maps, which tracks the carbon intensity of corporate electricity usage and numbers Google and Samsung among its customers, closed a €5mn financing round.

Electricity Maps, Sinay and their peers are benefiting from a broader surge in funding for so-called climate tech. According to data platform Dealroom, European climate-tech start-ups — in subsectors as diverse as electric mobility, nuclear fission and alternative proteins — raised $20.2bn last year, only slightly below the record $20.4bn they achieved in 2022. In total, they accounted for 43 per cent of global venture capital investment in climate tech, up from 29 per cent in 2022.

When it comes to AI, Lynn Kaack, who leads the AI and Climate Technology Policy Group at Berlin’s Hertie School, a graduate university of governance, says it pays to specialise. “AI is not a silver bullet against climate change,” she says. “Successful start-ups are using machine learning as one element in pretty technical applications, like predicting the evolution of clouds for better operating a power grid with high shares of solar photovoltaic.”

One such company is Dutch start-up Overstory, whose technology analyses satellite imagery and other remote sensing data to spot when vegetation is encroaching on power lines. The information enables electric utilities to cut the offending greenery back before it causes power outages or wildfires.

When it was founded, in 2018, Overstory’s aim was to use machine learning to detect deforestation. But, after two years, it decided to focus on serving utilities, which already spent heavily on vegetation monitoring. “We give them a network-wide view of risks caused by vegetation,” says chief executive Fiona Spruill. “Companies never had this technology. They had to have someone walk the lines and understand the risks.”

Satellite imagery, by contrast, allows them to “target the risks quicker and hopefully in a more cost-effective way”, Spruill says — adding that those risks are increasing because “utilities are facing more extreme weather events”.

Spruill notes, however, that AI is only as good as its training data. The company relies on experts to help label images, with visits to particular sites still sometimes needed.

“We’ve been powered by AI from the beginning so we’re not jumping on the bandwagon,” Spruill says. “We have six years under our belts, but identifying the species of a single individual tree from space is hard.”

Sinay’s Souami makes a similar point. Training AI models to recognise marine life, such as dolphins, turtles and seabirds, he says, requires “low-tech” inputs. “You need to have experts in species recognition physically tagging them and then training the machine. That’s low-tech.” 

Sometimes, too, Souami adds, analysing historical information is more valuable to his clients than, say, AI-powered predictions of weather patterns. “We need to focus on where we provide value and then choose the best tool to provide this,” he says.

Other AI entrepreneurs are similarly modest about the technology. Alex Marti, chief executive of Mitiga, a Spanish start-up that uses high-performance computing to evaluate the risk of natural disasters, warns that AI systems may embed biases found in the data they are trained on. “It’s very important to have transparency about how those models are built,” he says. “If the information is not there, you cannot use AI to make it up.”

Although some worries about AI are overblown — such as “if you let it run wild, then it turns into Terminators” — the real danger is overhyping it, Marti argues. “AI will help you, but it has limitations,” he says.

Some researchers, though, believe AI has significant environmental downsides — including heavy energy and water consumption by data centres, and the potential to accelerate the spread of climate disinformation.

Kaack at the Hertie School says that evaluating the technology requires a broad perspective. “It is important to not only look at the impacts of AI applications in the climate space,” she says. “Instead, we need to look everywhere where AI is being applied, and ask: in which way is this application reducing or increasing emissions? . . . It is a multipurpose tool, and it can also be applied to hurt the climate.”

Climate Capital

Where climate change meets business, markets and politics. Explore the FT’s coverage here.

Are you curious about the FT’s environmental sustainability commitments? Find out more about our science-based targets here