After a much-needed break from my pal Claude, I returned recently to the next and most challenging step in my AI-assisted adaptation and resilience portfolio experiment: mapping company operations exposure to physical climate risk.
As I have pointed out before, physical climate risk is location-specific. So to understand where the investment opportunities are for resilient companies in the context of this portfolio, I needed to integrate location-specific data.
In an ideal world, physical climate risk would be mapped across the supply chain from raw materials to end consumer. However, this is a hugely complex and ambitious task that I plan to work towards gradually.
As a first step, company operations provide a useful grounding for building physical risk understanding.
I knew even going into this step it would be a challenge given this project is leveraging publicly available data and building everything from scratch.
However, the experience of building reinforced the immense processing power of Claude making it possible to achieve this as an independent researcher (even if I did have to shell out an extra £10 for Claude credits on top of my monthly subscription of £20).
The result of this research is an AI-driven operational resilience index for the largest 141 US-listed companies. Let’s dive into the key findings.
TLDR: Asset intensity combined with geographic exposure is a key driver of corporate physical climate risk. Asset-light businesses like Healthcare and IT – and top index performer McDonalds – benefit from low direct balance sheet exposure. However, they often have high concentration risks further back in their supply chains not captured by this index. The low ranking of asset-heavy sector like energy and real estate reflects the importance of geographic operations. Nonetheless, relatively resilient companies can be found in low ranking sectors where geography is in the firm’s favour.
The Set Up
The Objective
To rank large US companies based on their operational physical risk in a world aligned with a 2.0°C world using publicly available data.
A key aim for this process was to use only high-quality, publicly available data that is, in my experience, underutilised by investors.
I initially looked at using S&P500 companies but instead drew heavily on S&P100 and large cap names after Claude advised me that the availability of operational data becomes patchier with smaller firms.
The Operational Data
One of the factors that has always held back my research in this space has been that there is no centralised, open source database of corporate operations. So, any attempt to map operations across multiple companies would require expensive subscriptions (which can still be very patchy and opaque) or manually trawling through endless corporate disclosures.
Enter Claude.
Claude pulled data on facility locations for the 141 companies analysed come from public company disclosures, primarily SEC 10-K filings, annual reports, sustainability reports, and regulatory filings. Facilities are revenue-weighted where revenue data is available.
It is important to say that this is not a complete picture of corporate facilities. The realities of data availability meant using the data that Claude could find.
While less of an issue for asset light businesses, Claude especially struggled mapping industries with distributed operations particularly consumer staples and diversified industrials. Consumer staples and diversified industrials end up with the lowest confidence scores in the index but, as it happens, were not the focus of my analysis anyway.
Nonetheless, readers should view these results as directional rather than a perfect measure of operational resilience. I would like to keep working on this to enhance the coverage and therefore accuracy over time.
For utilities, Claude initially just used the primary location but upon prompting used service territory maps filed with state public utility commissions provide territory-level coordinates.
Here is a summary of the split of geographic data:
| GICS Sector | Companies | Facilities | Critical Facilities | Countries |
| Information Technology | 22 | 65 | 18 | 8 |
| Financials | 18 | 36 | 8 | 6 |
| Health Care | 17 | 60 | 20 | 7 |
| Industrials | 16 | 75 | 24 | 5 |
| Energy | 14 | 64 | 16 | 8 |
| Utilities | 12 | 12 (61 territories) | 12 | 1 |
| Consumer Discretionary | 11 | 47 | 14 | 8 |
| Consumer Staples | 10 | 37 | 10 | 6 |
| Real Estate | 9 | 34 | 8 | 3 |
| Communication Services | 9 | 32 | 8 | 4 |
| Materials | 6 | 31 | 10 | 6 |
| TOTAL | 144 | 493 | 160 | 15 |
The Climate Data
I sourced climate hazard data from Probable Futures, a non-profit climate literacy initiative that I’ve spoken about on here before, and complimented it with the USDA Forest Service Wildfire Hazard Potential (WHP) dataset published by the Fire Modeling Institute.
The Probable Futures data used covers a wide range of countries for a 1.5°C and 2.0°C world, with 2.0°C used as the reference for scoring. The Wildfire Hazard Potential data reflects current vegetation, terrain, climate, and historical fire behaviour in the US.
*A reminder that current warming is estimated at between 1.2°C and 1.4°C.
| Source | Physical Risk Metric | Relevance |
| Probable Futures | Days above 35°C per year | Direct heat stress — workers, cooling infrastructure, equipment reliability |
| Probable Futures | Likelihood of year-plus drought (%) | Water stress — manufacturing, energy cooling, agriculture, mining |
| Probable Futures | Change in 1-in-100-year storm frequency (x) | Acute event risk — hurricane, extreme precipitation, infrastructure damage |
| Probable Futures | Change in total annual precipitation (mm) | Trend signal — reinforces drought and flooding dimensions |
| Probable Futures | Change in wildfire danger days | Forward-looking fire risk growth — combined with USFS absolute scores |
| USDA Forest Service Wildfire Hazard Potential | WHP scores are assigned on a 1–5 scale (Very Low to Very High). | Existing wildfire risk is an important floor for current and future risk |
I assigned a weighting to each physical risk factor that created a final resilience score per company, calculated by Claude. If you would like access to the detailed methodology, ranking and scores for the 141 companies in this analysis, please email [email protected].
The Operational Climate Resilience Index
In summary, the Index draws on:
| 144 | 493 | 15 | 6 | 2 |
| Companies | Facilities mapped | Countries Identified | Hazard dimensions measured | Warming scenarios used |
In terms of output, every company is given an operational resilience score of between 0 and 10, with 10 being completely resilient and 0 being highly vulnerable to physical climate risks. They are then ranked from highest resilience to lowest resilience.
It is important to note that no company in this index scores full marks for resilience, highlighting that even ‘top performers’ are exposed to physical climate risk. The highest single company resilience score in the index is 7.91 for McDonald’s. The lowest is 3.42 for Public Storage. To access the full company rankings, please email [email protected].
At the sector average level, there is less variation – the highest sector Healthcare scores 6.62 versus the lowest sector – Real Estate – at 5.36. This reflects the fact that there is a lot of within-sector variation. This underscores that there are potential investment opportunities for resilience across all sectors – even lower scoring ones.
Results: The Least Resilient Sectors among America’s Largest Companies
Key Takeaways from the Most Resilient Sectors: Asset-light businesses like Healthcare and IT – and top index performer McDonalds – benefit from low direct balance sheet exposure. However, Pharma’s reliance on Indian API, Tech’s reliance on Taiwanese chips and McDonald’s reliance on agricultural commodities highlight that significant risk lurks further back in the supply chain.
The Most Resilient Sector: Healthcare
In first place, healthcare tops the resilience rankings for two distinct reasons depending on sub-sector:
- Managed care insurers lead because they are essentially administrative businesses — no manufacturing, no physical infrastructure, just software and human capital. Even their geographic locations are replaceable. Climate can also drive revenue directly through increasing heat illness, respiratory claims, and vector-borne disease.
- Pharma scores well for a different reason — manufacturing geography. Eli Lilly, Merck, and AbbVie – included in the list – have historically concentrated high value production in the lower risk northeastern US and Europe.
However, the pharmaceutical sector highlights a very important limitation of direct operations mapping. Most large pharmaceutical companies source a significant portion of their APIs from Indian contract manufacturers, primarily clustered around Hyderabad, Mumbai, and Ahmedabad.
India is highly exposed to extreme weather already and supply chain modelling highlights that it tops global hazard distribution. Mapping direct operations does not reflect that exposure.
If it were, India supply chain exposure would be an important drag on pharma rankings.
Second Most Resilient Sector: Financials
In second place, financials score well because the highest-weighted names are asset-light platforms — Visa, Mastercard, S&P Global, Moody’s, BlackRock, CME Group. Meanwhile, banking pulls the average down due to physical exposure of major locations for large banks.
However, a very strong caveat is needed here: operational exposure is not a particularly meaningful measure of physical risk exposure for banks. So, this result should not be over-interpreted. Book exposure, mortgage portfolio concentration and investment portfolio exposures are very important, meaningful drivers of physical climate risk for many of these firms and cannot be captured in a direct facilities mapping like this.
For a more meaningful steer, keep an eye out in my upcoming sector risk series for detail on a financial sector-specific approach.
Third Most Resilient Sector: IT
In third place, IT scores well primarily because of the prevalence of fabless semiconductor companies and asset-light software platforms, but with meaningful dispersion within the sector and important limitations worth bearing in mind.
Fabless chipmakers — NVIDIA, Qualcomm, Broadcom, AMD — own no manufacturing. Their physical risk is limited to office campuses rather than industrial assets. The actual fabrication happens at TSMC in Taiwan and Samsung in Korea, which face their own risks but don’t sit on these companies’ balance sheets.
This is another key example of the limitation of mapping operational exposures.
The reality is that Taiwan sits in one of the world’s most active typhoon corridors. The island also faces water stress and TSMC’s fabs are extraordinarily water-intensive. When Taiwan experienced a severe drought in 2021, it forced water rationing to agricultural users to protect fab operations.
At 2°C warming, both the intensity of typhoons reaching Taiwan and the frequency of drought periods are projected to increase. The concentration risk is extreme. TSMC manufactures roughly 90% of the world’s most advanced semiconductors.
NVIDIA, AMD, Qualcomm, Broadcom, and Apple are all critically dependent on TSMC. But, just as with India for pharmaceuticals, this does not show up in their operational mapping because it exists further back in the supply chain.
On the flip-side of this story, the standout resilient name is Analog Devices, the only major semiconductor company with owned fabrication in low-risk locations — Wilmington, Massachusetts and Limerick, Ireland. Zero wildfires, no hurricanes, low heat. This is a great example of input geographic resilience versus peers.
Finally, California concentration drags the average scores for the sector down in the rankings. KLA Corporation, PayPal, and several Silicon Valley names face Bay Area wildfire and heat risk. Semiconductor equipment manufacturers Applied Materials and Lam Research have primary manufacturing in California, which faces compounding physical risks.
Further, the rise in data centre ownership among otherwise asset-light tech firms raises their direct operational exposure to physical risk.
The Most Resilient Company: McDonald’s
McDonald’s scores highest in the list of the 141 largest US companies because of its franchise model. The company owns very few of the buildings its restaurants operate in — approximately 95% of McDonald’s locations are franchised, meaning the franchisee owns or leases the physical space and bears the direct cost of flood damage, storm shutdowns, and heat-related operational disruption. McDonald’s itself primarily owns the brand, the supply chain relationships, and the real estate it leases back to franchisees.
The franchise network is geographically dispersed across the entire US, meaning no single climate event can concentrate losses at the corporate level. When a hurricane closes Florida McDonald’s restaurants, the franchisees absorb the immediate loss. McDonald’s collects royalties as a percentage of sales so it is exposed to some royalty income loss.
However, the wide geographic dispersion of locations means the risk is not overly concentrated in a single location and its own balance sheet is insulated from the direct physical damage.
As with pharma and tech, McDonald’s physical risk is lurking further back in the supply chain. McDonald’s supply chain is heavily concentrated in agricultural commodities — beef, potatoes, wheat, dairy, chicken — all of which face direct and growing physical climate risk. Furthermore, its coffee sourcing for McCafé touches some of the most climate-vulnerable agricultural geographies in the world.
This underscores the importance of complementing operational physical risk exposure with supply chain analysis to produce a more complete picture of corporate resilience to extreme weather.
Results: The Least Resilient Sectors among America’s Largest Companies
Key Takeaways from the Least Resilient Sectors: Energy and Real Estate have the highest physical asset intensity in the universe, the least geographic flexibility, and the highest concentration in demonstrably climate-exposed locations. However, resilience winners can be found in lower risk regions, e.g. operators in Appalachian Gas.
The Second Least Resilient Sector: Energy
Energy scores low in these rankings because the sector’s core assets — refineries, platforms, pipelines — cannot be relocated. They sit where the resource is, which in the US means the Gulf Coast and the Permian Basin.
Gulf Coast refiners face 2-3x storm frequency at 2°C. Hurricanes Harvey 2017 and Ida 2021 demonstrated these aren’t theoretical risks. Meanwhile, Permian Basin operators face a different problem — extreme water stress that is basin-wide and independent of individual operator behaviour.
The sector is saved from scoring even lower by its internal bifurcation. Appalachian gas companies like EQT sit above 7.0 — the basin is confirmed low-risk across all hazard dimensions. But they are outnumbered by Gulf Coast and Permian names that drag the average down.
The Least Resilient Sector: Real Estate
Real estate scores lowest of all sectors because it is the greatest prisoner to geography: it is location-dependent. A managed care insurer can theoretically move its headquarters. A REIT cannot move its buildings. Physical risk is structural for real estate in a way it isn’t for most other sectors.
The S&P 100 REIT universe also happens to be concentrated in three of the highest physical risk US geographies.
There is also a structural asymmetry that makes real estate uniquely exposed. Climate events that destroy or damage other sectors’ assets create demand for reconstruction — Home Depot, Waste Management, and engineering services firms benefit from hurricanes. REITs don’t benefit from their own assets being damaged. The demand tailwind that partially offsets physical risk for some sectors doesn’t exist here.
The Least Resilient Company: Public Storage
The bottom performer in the whole index reflects what happens when asset intensity meets high risk geography.
Public Storage scores lowest because it combines California geographic concentration with a business model that has no offset mechanism. Claude found that a quarter of its portfolio by revenue is in California — facing wildfire, drought, and extreme heat simultaneously across all three hazard dimensions that matter most.
Conclusion and Next Steps
The consistent theme across this analysis is that asset intensity drives physical climate vulnerability. Companies whose primary assets are data, intellectual property, networks, or contractual relationships score higher than companies whose primary assets are factories, mines, or physical infrastructure. That structural difference explains most of the ranking differential between sectors.
But clearly operational mapping alone tells only part of the story and has a multitude of limitations. The gaps — India API supply chains, TSMC Taiwan fabrication, agricultural commodity inputs — are where the next layer of analysis is needed. If you’d like to follow along as I continue building, subscribe to the Resilient Investor using the button below.
The full methodology, individual company scores, and detailed findings for all 141 companies are available on request — email [email protected].