Inside the Geospatial AI Boom Reshaping Climate Risk Analysis

In early May, Xoople — a Spanish geospatial data and AI start-up — emerged from stealth with a €115mn (US$131mn) funding package backed by the Spanish government. The self-titled “Earth Intelligence” pioneer offers businesses insights on climate impacts, land use, and other environmental factors that could affect their bottom lines.

It’s not the only geospatial company with a climate spin to hit the headlines.

Back in April, Seattle-based Earth Finance announced its acquisition of Climate Engine, a trailblazer in “spatial finance” — a nascent corner of the geospatial market where satellite imagery, economic data, and financial analysis intersect. The deal gives Earth Finance, a boutique advisory firm, access to tech that helps corporate clients map climate risks to their physical assets and supply chains. 

Then there’s LGND AI, a buzzy San Francisco-based geospatial AI start-up which Climate Proof understands is raising a seed round.

But not all geospatial news has been rosy. Germany’s Maya Climate — a three-year-old startup attempting to layer generative AI on top of geospatial data for corporate clients — is winding down operations, its founder said on LinkedIn. Elsewhere, Pachama, a similarly tooled, nature-focused geospatial company is cutting staff “in light of macro challenges affecting sustainability markets”, according to its CEO.

These moves illustrate both the promise and fragility of the geospatial market, where investor enthusiasm and the growing demand for climate risk insights is running up against commercial realities.

“It's the wild west,” says Priscilla Cole, a product engineering and strategy consultant at Circle Risk Consulting. “Commercial buyers are overwhelmed at the sheer number of products, and rapidly disappointed at how not ‘turn-key’ most of them are. Few have really figured out how to break into commercial sales on their own.”

THE ADAPTATION MARKET PULL

Past clients for Earth Observation (EO) data and analytics have originated from the defense and government intelligence sectors. But today, interest is being propelled by corporates and financial institutions anxious about their climate-related physical risk exposure and eager to implement adaptation strategies.

“[Asset owners] increasingly acknowledge that physical risk is becoming a lot more prominent, it’s becoming more damaging, it’s becoming more financially material — and they need to get their act together in how they’re measuring, assessing, mitigating these risks,” says Christophe Christiaen, head of the UK’s Spatial Finance Initiative.

This has accelerated demand for asset-level intelligence, the kind that EO provides the raw ingredients for, and spatial finance — which companies like Climate Engine and Sust Global are pioneering.

“Spatial finance [is about] applying financial — and we’re [also] adding policy-related —layers onto [geospatial] data so that instead of having a descriptive observation of what's happening or what might happen, [we] create an insight of why does this matter to your business, and what do you need to do to change or to adapt,” says Garrett Kephart, CEO of Earth Finance.

THE EVOLVING TECH STACK

Spatial finance is one of many niches in the burgeoning geospatial arena. The market can be roughly organized across three tiers, with the climate adaptation imperative a force within each.

First, there are the EO satellite operators, such as Planet Labs, ICEYE, and Maxar, which anchor the data layer by supplying raw imagery from hundreds of satellites. Public providers like NASA and ESA also pipe open source data to commercial entities.

One rung down the ladder, tech giants like Google and Microsoft increasingly dominate the platform layer, processing and storing petabytes of geospatial data in their cloud environments and creating immensely powerful foundation models for the world to play with. 

Then there’s the application — or intelligence — layer, populated by companies that build sector- or risk-specific analytics, integrating EO data with other datasets and injecting their own AI magic to deliver actionable information to clients. Start-ups like Floodbase, Fathom, Climate X, Sust Global, LGND AI and others are all operating in this space.

“Intelligence is where a lot of value happens,” says Aravind Ravichandran, Founder and CEO of geospatial strategy and market analytics firm TerraWatch Space. “That’s where the relationship with a lot of the enterprise happens.”

The advent of mighty foundation models in recent years has been a key driver of activity in this intelligence layer. These large-scale, deep learning models are pre-trained on vast quantities of geospatial data and optimized for mapping and prediction tasks. With their capabilities, it has become possible to deliver detailed climate risk and exposure data to users faster and more accurately than ever before. 

Google (PDFM), IBM (Prithvi), Microsoft (Aurora), and the nonprofit start-up Clay have all released — or are releasing — models that vastly reduce the compute power required to process geospatial data, democratizing access for developers and downstream users.

“What’s cool about these models is they’re able to make use of the breakthroughs and insights from the LLM space and apply them to other modalities,” says Avery Cohn, Strategy Lead at Clay and a member of the founding team at LGND AI. “The way that we’re seeing it applied in the satellite data space is … they take the raw data and then they translate it into model speak — and it makes it really easy to use that compressed data for downstream applications at a fraction of the cost and compute effort that you would have had if you had to use raw satellite imagery,” he explains.

But while foundation models like these help spur innovation, they also risk commoditizing geospatial data and analytics to such a degree that it becomes tough for companies to differentiate themselves, or maintain competitive advantages.

“The thing that these geospatial foundation models do is they effectively remove a lot of the defensive moat … all of a sudden people are able to do it [analytics] a lot more seamlessly,” says Josh Gilbert, CEO of Sust Global. “But it’s actually fucking awesome for the market. I’m hopeful that with these foundation models, it kind of allows geospatial AI and spatial finance to actually achieve the things that people have been saying they can achieve for probably the best part of a decade now.”

However, it also makes the application layer a more volatile and challenging place to build a business. Maya Climate and Pachama have their own reasons for shutting down and scaling back, but their stories are likely to become more common as a stampede of capital and entrepreneurs rush in.

“The smaller companies are just going to fall apart,” says Tee Barr, Director of Product, Geospatial, at risk modeler Verisk. “The engineers who love technology, but don’t understand the business use cases [will collapse], and the reverse will also be true — the business people who don’t understand the technology.”

Gilbert agrees. He says the rise of the “hyperscalers” — Google, Microsoft and their ilk — will reveal those companies “that didn't really know what they were talking about when they talk about geospatial AI.”

He predicts successful future businesses in the intelligence layer may resemble artisanal shops — small teams of domain experts combining datasets and analytics for narrow verticals like insurance, infrastructure, or agriculture. Instead of raw data capabilities, it will be the “taste and discernment” these providers offer that differentiate them.

THE RIGHT TOOLS FOR THE JOB

Artisanal providers in the climate risk and adaptation space may be particularly well-placed to flourish.

A number of institutional investors are already using geospatial tools, or building them in-house, to get a handle on climate-related financial risks and opportunities. For example, asset management giant BlackRock has built an internal geospatial investment capability that works to turn “data into alpha”, said Joshua Kazdin, a Managing Director at the firm, on a recent podcast. Elsewhere, reinsurance behemoth Swiss Re has partnered with UK-based Fathom to refine its flood risk models.

Christiaen at the UK Spatial Finance Initiative says appetite from governments, development banks, and corporations to understand how adaptation measures can yield returns will boost the geospatial intelligence market further.

“If you can track that [the adaptation benefit], then you can start to incentivize it and link it into your financial system theory of change,” he says. “I think that's still a piece of the puzzle that is missing,” he explains.

But growing this segment also requires geospatial firms to find ways to seamlessly integrate their data and insights into client workflows. “Tables and data feeds are the native formats of finance. It's a pipedream to think that financial firms will magically convert into geospatial users, as they will likely never drop their Excel addiction,” says Cole.

Large Language Models (LLM) may be a big help here. Indeed, one of the most hyped use cases is applying generative AI so that querying geospatial data is as simple as typing in a question on a search bar. This could dramatically lower technical barriers for bankers, insurers, and corporate risk managers.

“If I could write a prompt that said, ‘Convert this from Mercator to Peters projection and intersect all cities within 50 miles of the US border,’ and just run the analysis for me... a large language model could do that much more effectively than an analyst could,” says Barr.

Yet others warn that while AI can simplify interfaces, it doesn’t eliminate the need for domain expertise — especially when data gaps or model limitations risk producing misleading outputs.

“AI helps build more models to get more insight or data where you wouldn’t have it now, but if that is not actually underpinned with some sort of validation or good ground truth data, then you don’t actually know how good that data layer is, or how reliable it is,” says Christiaen.

SHAKEOUT AHEAD?

Geospatial companies have swallowed rivers of capital in recent years. TerraWatch Space estimates total investment in 2024 came to US$1.7bn (excluding government contracts), down slightly from US$1.9bn in 2023. Most cash flowed to the application layer, with providers focused on insurance, utilities, and financial services seeing the greatest interest.

However, Ravichandran explains that financing for this layer dropped 14% year-on-year, caught up by the broader climate tech downturn. Commercial headwinds are likely to persist for those companies that are narrowly focused on climate, nature, and adaptation applications.

Of course, there will be exceptions. Government-backed players like Xoople may enjoy a capital cushion as sovereign governments increasingly view EO as strategic infrastructure, and look to incubate their own capabilities separate from the US — which under the Trump administration appears to be purposely blunting its geospatial edge. Still, institutional backing can only take a company so far without commercial alignment.

“It's my opinion that the field of geospatial data and service providers is overcrowded,” says Cole. “Prior VC investment bubbles are a long way from settling. Anyone playing in this space is merely a testing ground and winners should be looking for M&A opportunities or embedded partnerships at the very least.”

Some enterprises have already taken this advice. In February, Vancouver-based EarthDaily Analytics acquired the niche wildfire and forestry geospatial start-up SkyForest. That same month, geodata giant Fugro snapped up EOMAP, specialists in mapping and monitoring of marine and freshwater environments.

It’s capitalism red in tooth and claw — and the competition (and consolidation) is only likely to heat up. After all, the geospatial market sits at the intersection of multiple megatrends: escalating climate risks, a Trump-induced retreat from public data provision, the AI explosion, and a hunger across the private sector for superfine, asset-level risk insights.

But while the level of interest and demand may be higher than ever, the rapidly evolving commercial space makes it perilous for the unprepared. Sust Global’s Gilbert, though, is optimistic:

“I can see a world where there will be a massive range of vertical-specific, geospatial AI platforms or tools — which can still be very valuable and be very good businesses. That's fucking awesome – think about all the problems that you can solve.”

Thanks for reading!

Louie Woodall
Editor