Asset-Level Climate Risk: The New Insurance Baseline

Swiss Re’s 2025 estimate puts global insured natural catastrophe losses at $107 billion, with total economic losses at $220 billion. That’s not a “bad year” anymore—it’s a pricing signal. The market is telling anyone who still relies on broad, top-down climate assumptions that they’re late.

Apollo Global Management’s decision to expand from portfolio-level climate screens to bottom-up, asset-level risk reviews is a sharp example of where financial risk is heading. It’s also a preview of what insurance carriers (and the energy and utilities companies they insure) need to operationalize: risk decisions that resolve all the way down to the asset, location, and operating constraint level.

This post sits in our AI in Energy & Utilities series for a reason. Data centers, transmission corridors, substations, wind farms, gas plants, battery sites—these are physical businesses. Extreme weather doesn’t “affect the sector.” It affects specific assets on specific days, and the cost shows up as downtime, repairs, supply chain delays, and—crucially—insurance availability.

Why asset-level climate risk is replacing “top-down” scoring

Answer first: Top-down risk scoring is too coarse to price and underwrite a world where flood, wildfire, heat, and wind losses can reprice a neighborhood—or a collateral pool—overnight.

Apollo started doing top-down analyses in 2023 and is now expanding to more granular reviews to identify company-level risks before closing deals. That move matters to insurers because it reflects a broader truth: capital providers are demanding evidence that valuations and cash flows can survive climate volatility.

When you underwrite or insure large physical portfolios—commercial real estate, mortgages, energy infrastructure—small differences at the asset level compound fast:

• A mortgage portfolio isn’t “coastal” or “inland.” It’s a set of addresses with different elevation, drainage, building materials, and local mitigation.
• A utility isn’t “exposed to storms.” It has feeders, poles, substations, vegetation management cycles, and mutual aid dependencies.
• A data center isn’t “in a hot region.” It has a cooling design, water sourcing constraints, power redundancy, and curtailment risk.

That’s why Apollo’s “loan-level mapping” and evaluations across drought, flood, heat, and wildfire exposure are the direction of travel. For insurers, the implication is blunt: pricing and capacity decisions will increasingly be validated (or rejected) at the same granular level.

The quiet shift: climate is now treated like a standard financial driver

JPMorgan’s climate advisory lead framed it well: investors need to assess climate effects the way they assess inflation, debt coverage ratios, or political risk—because it hits cash flows and costs.

Insurance has always been in the cash-flow business. What’s changing is that climate signals are getting pulled into decisions earlier—deal screening, lender terms, covenants, and insurance placement—rather than showing up as a surprise at renewal.

What Apollo’s approach teaches insurers about underwriting modernization

Answer first: Asset-level risk reviews only work if they’re automated, repeatable, and explainable—this is exactly where AI in insurance moves from “nice to have” to mandatory.

Apollo’s stated approach includes integrating risk assessments into every deal and refining measurement as technology and data improve. Insurers should read that as an operating model requirement: continuous risk measurement rather than annual modeling exercises.

Here’s the practical translation for property and specialty carriers.

1) Underwriting can’t stop at hazard maps

Hazard layers are useful, but they’re not enough to price resilience. Underwriters increasingly need vulnerability and operational context, such as:

• Roof type, cladding, defensible space, and ember exposure for wildfire
• Drainage, first-floor height, and nearby water infrastructure for flood
• Heat days, cooling dependencies, and water constraints for data centers and industrial sites
• Wind vulnerability, line hardening status, and vegetation proximity for utilities

AI supports this shift by combining heterogeneous data—geospatial, imagery, engineering data, maintenance logs, and prior claims—into a consistent feature set underwriting can actually use.

2) “Transition risk” is no longer only about regulation

Apollo explicitly mentions physical and transition risks. For insurers, transition risk shows up in surprisingly operational ways:

• A carbon or water policy change can increase operating costs at energy-intensive sites.
• Community opposition or permitting delays can extend construction periods and inflate builders risk exposure.
• Supply chain substitutions (new vendors, new materials) can change failure modes and warranty risk.

AI can help identify these patterns earlier by learning from project histories: schedule slip predictors, contractor performance signals, and “change order” volatility.

3) Resilience is becoming insurability

Jefferies analysts highlighted the real-world outcomes: higher premiums, lower asset values, and reduced access to insurance.

That last piece—capacity—hits energy and utilities first because many assets are large, location-specific, and hard to move. If a substation becomes difficult to insure at a reasonable deductible structure, the cost doesn’t just land on the utility. It flows through to regulators, ratepayers, lenders, and project investors.

Insurers that build credible asset-level views can do something competitors can’t: offer coverage with confidence where others can only reduce limits or walk away.

AI in energy & utilities: where asset-level climate risk becomes concrete

Answer first: In energy and utilities, asset-level climate risk is basically a reliability problem—and AI is the most scalable way to connect hazard forecasts to operational decisions.

Apollo called out data centers specifically: power is a major operating expense, and water sourcing, use, and recycling affect cost, regulatory exposure, and resilience. That same logic applies across energy infrastructure.

Data centers: power, water, heat, and downtime all intersect

If you insure (or finance) data centers, you’re underwriting a chain of dependencies:

• Heat waves drive cooling load, which drives power demand, which can raise curtailment risk.
• Water restrictions can make certain cooling strategies expensive or non-viable.
• Grid congestion and wildfire-related PSPS events can create correlated downtime risk.

AI contributes in three practical ways:

1. Demand forecasting + weather forecasting fusion to anticipate peak loads and price spikes
2. Predictive maintenance to reduce failure probability during stress periods
3. Scenario simulation to quantify expected downtime and business interruption exposure

For insurers, the underwriting question becomes more specific than “Is it in a risky area?” It becomes: How does this facility perform under the top 10 stress scenarios that matter?

Utilities: climate volatility is a claims volatility problem

Wildfires and storms don’t just create property losses; they create liability, business interruption, and cascading infrastructure damage.

Asset-level analytics (often AI-enabled) help carriers and utilities agree on a shared, testable picture of risk drivers:

• Vegetation encroachment likelihood by segment
• Pole condition and replacement prioritization
• Substation flood hardening benefit vs. cost
• Restoration time estimates based on crew access, mutual aid availability, and spare parts

When carriers can see these drivers, pricing can reward mitigation rather than punish geography.

Renewables and storage: small data gaps become big pricing errors

For wind, solar, and battery storage, underwriting accuracy depends on details that are often messy:

• Microclimate variance across a single site
• Equipment model changes mid-project
• Curtailment patterns and congestion risk
• O&M practices that affect failure rates

AI helps by standardizing disparate project documentation and operational telemetry into comparable risk indicators. The best implementations don’t “replace the engineer.” They make the engineer’s judgment scalable.

A practical blueprint: how insurers can build asset-level climate risk with AI

Answer first: Start with one line of business and one peril, build a repeatable data-to-decision workflow, then expand—because the hard part isn’t the model, it’s the operating system around it.

If you’re a carrier, MGA, or reinsurer trying to modernize climate risk modeling, this sequence works.

Step 1: Pick a decision that matters (and measure it)

Choose one underwriting decision where asset-level clarity changes outcomes, such as:

• Property limit and deductible structure for flood-prone commercial risks
• Wildfire eligibility and defensible space credits for specific ZIP clusters
• Business interruption waiting period adjustments for data centers

Define success metrics upfront: quote-to-bind, loss ratio by tier, referral rate, inspection utilization, renewal retention.

Step 2: Build an asset-level “risk record” that underwriters trust

A useful asset-level record typically includes:

• Location intelligence (coordinates, elevation, proximity to fuels/water)
• Building/asset attributes (construction, roof, protection, redundancy)
• Hazard frequency and severity features (not just a single score)
• Mitigation evidence (photos, inspection notes, maintenance logs)
• Claims signals (cause of loss patterns, near-miss indicators)

AI is most valuable when it reduces manual effort here: extracting attributes from documents, normalizing inspection notes, reading imagery, and flagging inconsistencies.

Step 3: Use AI for prediction—and for triage

Not every account needs a bespoke model run and human review.

A pragmatic approach:

• Straight-through for clearly low-risk assets with strong data completeness
• AI-assisted referral for borderline risks (missing fields, conflicting signals)
• Specialist review for high-value, high-accumulation, or unusual exposures

This is where carriers see real cycle-time improvements without sacrificing control.

Step 4: Make it explainable enough to sell and renew

If you can’t explain a risk decision, you’ll lose it in negotiation.

The bar isn’t academic interpretability; it’s practical clarity:

• Which features drove the decision?
• What mitigation actions would reduce price or increase capacity?
• How will the carrier verify changes at renewal?

Explainability is also how you avoid the “black box” trap in climate pricing, where brokers and insureds assume it’s arbitrary.

People also ask: what does “asset-level climate risk” mean in insurance?

Answer first: Asset-level climate risk means underwriting and pricing based on the exposure and resilience of a specific location and asset, not broad regional averages.

In practice, it’s the difference between:

• “This county has high flood risk” (portfolio-level)

and

• “This substation floods when rainfall exceeds X inches in Y hours because of drainage constraints; hardening reduces expected downtime by Z%” (asset-level)

Insurers that can support the second statement will write better risks—and avoid subsidizing avoidable losses.

Where this is headed in 2026: continuous risk, continuous pricing pressure

Apollo is reacting to a world where extreme weather can knock asset values down quickly and permanently. Insurers are already living in that world—through rate adequacy fights, capacity pullbacks, and reinsurance volatility.

The better path is to treat climate-driven disruptions as operational variables that can be measured, mitigated, and priced with precision. That requires AI in insurance in the unglamorous places: data quality, asset-level records, model monitoring, and underwriting workflow.

If you’re in energy and utilities, this matters even more. Grid reliability, data center uptime, renewable performance, and catastrophe exposure are converging into a single question: Can you prove resilience at the asset level?

If you want to pressure-test your current climate risk modeling and underwriting workflow—especially for energy, utilities, and data center exposures—start by mapping your data-to-decision path for one peril and one product. Where the handoffs break is where the losses hide.

What would change in your portfolio if every risk decision had to be justified down to the asset, not the region?