Most asset managers already have physical risk data. Few have data that holds up under regulatory scrutiny, client reporting, or real decisions.
That was the premise of a recent Clarity AI webinar, The Physical Risk Gap: What Today’s Datasets Are Missing, featuring Jakob Thomä, Founder and CEO of Theia Finance Labs, and Alice Borgonovo, Climate Senior Product Manager at Clarity AI. The conversation moved the discussion from ambition to implementation.
Investors’ forecast on physical risk
Drawing from a recent survey with investors and financial sector stakeholders about the future of physical risk, Thomä framed the rest of the conversation with one statement: governments are not expected to move toward aggressive adaptation investment over the next five years.”Governments are expected to react, not prepare,” he noted. That means the world is heading down a costlier, reactive path rather than a cheaper, preventive one.
“Governments are expected to react, not prepare.”
Two findings from that survey stood out as connected but underappreciated. First, insurance is not expected to ride to the rescue. Private insurers are increasingly unwilling to absorb the insurability challenges tied to some of the most exposed regions. At the same time, respondents expect firms to lean more heavily on insurance and adaptation technology to manage risk, one of the more unresolved dynamics in the market today.
Second, and perhaps more overlooked, is migration. Physical risk models tend to assume a factory is hit, production stops, and everyone simply stays put. In reality, social dislocation is likely to arrive not as a slow trickle but in sharper waves after major events, reshaping exposure in ways that pure geospatial hazard mapping won’t capture on its own.
The survey also pointed to a near-consensus that climate tipping points now need to be part of physical risk analysis, alongside inflation, public debt, and corporate balance sheet effects. Yet even with that complexity mounting, the single dominant risk factor investors still expect to matter most this decade is extreme acute weather, precisely the mechanism that geospatial, asset-level analysis is built to capture.
Beyond the headquarters: why asset-level data is non-negotiable
A structural weakness shows up repeatedly in the market: a persistent focus on company headquarters rather than the assets that actually drive a business. “Maybe we are assessing the risk of a financial district in downtown New York, or in London, rather than actually analyzing where the beating heart of the company is,” said Alice Borgonovo from Clarity AI.
“Maybe we are assessing the risk of a financial district in downtown New York, or in London, rather than actually analyzing where the beating heart of the company is.”
Manufacturing sites are often reliant on natural resources like fresh water, and disproportionately located in emerging markets with higher exposure to acute hazards. The fix isn’t covering more assets, it’s covering the right ones. “It’s not just about how many assets are covered, but which assets are covered. That is really the key question to understand the real exposure of a company,” she added.
“It’s not just about how many assets are covered, but which assets are covered. That is really the key question to understand the real exposure of a company.”
Moving past that means combining geolocation, building height, and proximity to infrastructure like the electricity grid. “An address is not sufficient,” Borgonovo said.
Clarity AI’s approach layers AI-assisted geolocation, built with climate risk partner RiskThinking.ai, with quality assurance and direct client feedback.
From checkbox exercise to financial reality
A recurring theme during the webinar was whether physical risk is still a disclosure formality or genuinely factored into decisions. “Historically, we’ve been used to the concept of this being more of a checkbox exercise,” said Borgonovo. That’s changing, for three reasons: regulators applying more scrutiny under TCFD and IFRS S2, investors wanting physical risk reflected directly in financial assessments, and simply lived experience. “We just witnessed a heat wave… It’s no longer a 2050 discussion. It’s a tomorrow, it’s a today discussion,” she said, pointing to the BP AGM pushback on scaling back climate disclosure.
“Historically, we’ve been used to the concept of this being more of a checkbox exercise.”
That shift shows up in time horizons too, from long-term 2050 scenarios to two, three, five years, refreshed after major events like an El Niño cycle. “You don’t take something serious unless you have a strategy. Otherwise it lands back in the cupboard until the next heat wave rolls around,” said Thomä.
“You don’t take something serious unless you have a strategy. Otherwise it lands back in the cupboard until the next heat wave rolls around.
Why hazards and scenarios shouldn’t be assessed one at a time
The conversation then turned to how hazards and scenarios are chosen, and why that choice determines whether an analysis holds up to scrutiny. Take “global weirding”: not just that the planet is warming, but that weather is more erratic and hazards increasingly compound. Death Valley’s record heat was immediately followed by Hurricane Hilary’s record rainfall in the same region. “We shouldn’t just say the heat wave is the hazard I want to focus on. We should make sure that all the different perils interact with one another,” said Borgonovo.
The same logic applies to scenarios. Rather than committing to a single pathway like 1.5°C, Clarity AI favors a stochastic view across multiple scenarios at once. “Do not commit to a single scenario. Be open, take into consideration more than one,” she said, an approach that’s both more defensible to regulators and closer to how hazards actually unfold.
“Do not commit to a single scenario. Be open, take into consideration more than one.”
A harder question remains: even where data has improved, do investors know what to do with it? “The data gap is in many cases smaller than people perceive. The bigger gap now is what investors want to do with the data,” said Thomä, whether passive strategies should shift exposures ahead of a forecast event, and when active managers should move.
“The bigger gap now is what investors want to do with the data,” said Thomä, whether passive strategies should shift exposures ahead of a forecast event, and when active managers should move.
How to actually move from ambition to implementation
The “physical risk data gap” is less about access; most institutions can now get asset-level, AI-assisted geospatial data. It is more about whether that data is granular, defensible, and current enough to survive scrutiny.
Losses appear to be the main driver of urgency. “I think you need to see losses… I give it one or two more years,” said Thomä. On the regulatory side, disclosure requirements have expanded significantly, but the harder question now is whether firms can “walk the talk.”
“I think you need to see losses… I give it one or two more years.”
Nature-related risk and physical risk are increasingly seen as converging rather than separate topics, with water-hungry data centers as a clear bridge between transition and physical risk.
Two blind spots stand out. Housing and insurability can shift “in major trouble” territory as fast as any risk out there, yet remain absent from most scenarios, including the NGFS’s own, according to Thomä. And supply chains beyond Tier 1 suppliers remain, in Borgonovo’s words, “a bit of the Wild West.”
Closing the gap means moving past headquarters-level shortcuts, treating hazards and scenarios as interacting systems, and giving social and supply-chain risk the same rigor already applied to weather events. The datasets are improving fast. The real question now is whether investors, insurers, and regulators are ready to act on them.
Author Information
