Verisk Analytics, Inc. · NASDAQ:VRSK

Verisk Analytics — Emerging Issues Webinar: AI Data Centers: An Evolving Risk Landscape (October 2026)

73 slides from Verisk Analytics's conference presentation, Oct 1, 2026

1. Slide 1: AI Data Centers: An Evolving Risk Landscape

   AI Data Centers: An Evolving Risk Landscape An Emerging Issues Webinar October 1, 2026 verisk.com

2. Slide 2: Policy Statement

   Policy Statement The policy of Verisk Analytics and subsidiary companies is to comply in all respects with federal and state antitrust laws. With this in mind, we want to mention that during all seminars held under our auspices, this policy prohibits discussion of certain topics. Because we want to

3. Slide 3: Today’s Presenters

   Today’s Presenters Laura Panesso Eric Price-Glynn Christopher Beston VP, State Relations and Regulatory Reporting Senior Principal, MarketStance Director, Catastrophe and Risk Solutions Kevin Kuntz Sam Haynes Kevin Sullivan VP and ISO Chief Engineer VP of Data Analytics, Verisk Maplecroft Analyst, E

4. Slide 4: Today’s agenda

   Today’s agenda 01 Introduction: The Evolution of Data Center Risk 02 AI Data Centers: Legislative & Regulatory Update 03 Virginia Data Centers and Economic Context 04 NatCat and Geographic Concentration Risks 05 Water Availability and Emergency Response 06 Strikes, Riots, and Civil Commotion Conside

5. Slide 5: A New Era of Risk?

   A New Era of Risk? The evolution of the AI data center 5 ©2026 Verisk Analytics, Inc. All rights reserved .

6. Slide 6: Data Centers: New Era, New Risks?

   Data Centers: New Era, New Risks? The changing face of data center development $84.1 billion The year-to-date total on construction start spending as of end of July ➢ “Hyperscale” Emphasis: Driven largely by AI model computation, requiring massive server deployments, extensive cabling, and redundant

7. Slide 7: Data Center Density

   Data Center Density Seattle, WA Minneapolis, MN Portland, OR 4,000+ data center locations in the U.S. Columbus, OH Boston, MA Labeled cities are home Kansas City, MO to 50+ data centers Ashburn, VA Santa Clara, CA Sterling, VA Manassas, VA Data locations San Jose, CA Phoenix, AZ derived from Verisk

8. Slide 8: Enterprise / Legacy vs. Hyperscale Data Centers

   Enterprise / Legacy vs. Hyperscale Data Centers Enterprise / Legacy Data Centers Hyperscale Data Centers vs Onsite facility or single building Expansive campuses A few hundred (or thousand) servers 5,000+ servers 1–20 MW demand 100+ MW demand 3-8 kW per rack 30–50 kW per rack (100+ kW/rack deployed,

9. Slide 9: Data Center Water Use: Small Nationally, Intense Locally

   Data Center Water Use: Small Nationally, Intense Locally Demand is tripling, with much of that water sourced from municipal systems and growth occurring in water-stressed regions 3x growth in U.S. data center water 97% 2 of 3 consumption in a decade of data center water is purchased new data centers

10. Slide 10: Data Center Electricity Use: Emerging Risk Signals

    Data Center Electricity Use: Emerging Risk Signals • May account for up to 15% of U.S. electricity use by 2030, per DOE report • Power demand could reach 194 GW by 2035, according to one estimate • 10 of 16 NERC power reliability regions in U.S. identified as at risk over next three years Source: No

11. Slide 11: Hyperscale Data Centers: Converging Pressures

    Hyperscale Data Centers: Converging Pressures Rising capacity demands and compressed construction timelines • 70%+ of 112 markets surveyed report contractor capacity tightening or overstretched Growing demand for resources • 2026 study: If 2024 water-use intensity persists, U.S. data centers in tota

12. Slide 12: State of Play

    State of Play An overview of state and federal data center regulatory and legislative activity 12 ©2026 Verisk Analytics, Inc. All rights reserved .

13. Slide 13: Federal Data Center Overview

    Federal Data Center Overview • Executive Order 14141 (Jan 14, 2025; subsequently revoked): ▪ Accelerates permitting and interconnection for large-scale Artificial Intelligence (AI) infrastructure, including hyperscale data centers. • Executive Order 14318 (Jul 2025): ▪ Adds FAST-41 accelerated timel

14. Slide 14: AI Oversight Backdrop

    AI Oversight Backdrop • State Authority Preserved ▪ State regulatory activity continues: The proposed 10-year federal moratorium on state AI regulation was struck by a 99–1 Senate vote, and states continue to develop AI-related requirements • NAIC Framework ▪ The 2023 Model Bulletin sets expectation

15. Slide 15: State Siting, Energy & Water Rules

    State Siting, Energy & Water Rules • Siting and Land Use ▪ Local zoning, setback, noise, and permitting requirements govern proposed data-center development. A growing number of jurisdictions have adopted or considered temporary moratoria while evaluating potential effects • Energy and Large-Load Ta

16. Slide 16: Virginia Data Centers and Economic Context

    Virginia Data Centers and Economic Context CAPEX, Construction & Commercial Lines Forecast for Linked Non-Utility Services 16 ©2026 Verisk Analytics, Inc. All rights reserved .

17. Slide 17: Macro Context: Rapidly Growing Exposures

    Macro Context: Rapidly Growing Exposures Information & Data Processing Services Fixed Assets • Asset values rose >3x the CAGR of the rest of the 600 private economy 2015-2024 500 Billions ($) 400 • Warehousing ~10% CAGR was a distant second place 300 200 • Perspective: the values of structures and e

18. Slide 18: Data Center Construction: Primary Classifications

    Macro Context: Data Center Construction Classes NAICS Description Role in Data Center Construction General Contractor / Building Shell 236220 Commercial & Institutional Building General contractor for data center building shell Construction Data Center Construction: 236210 Industrial Building Constr

19. Slide 19: Data Center Construction: Secondary/Tertiary Classifications

    Macro Context: Data Center Construction Classes NAICS Description Role in Data Center Construction Secondary – Site Work & Foundation 238910 Site Preparation Contractors Site grading, excavation, land clearing 238110 Poured Concrete Foundation & Foundations, structural concrete for server halls Data

20. Slide 20: Are Data Centers Built By Local Contractors?

    Are Data Centers Built By Local Contractors? Evidence from the 2022 Economic Census • No matter where located, data center 24% 17% 17% 15% 11% construction likely depends more heavily Power and Structural Commercial Industrial All Construction on out-of-state expertise than many other Communication

21. Slide 21: Data Center MEP Categories by Occupation

    Data Center Mechanical, Electrical, Plumbing (MEP) Categories by Occupation Electrical Electricians, Fire Alarm Installers, Power-Line Installers, Helpers-Electricians Mechanical HVAC Mechanics, Sheet Metal Workers, Insulation Workers (Mech), Industrial Machinery Mechanics Plumbing Plumbers/Pipefitt

22. Slide 22: Has Data Center CAPEX Produced Wage Pressures in MD, VA?

    Has Data Center CAPEX Produced Wage Pressures in MD, VA? Evidence Mixed 1.2 Normalized Ratios: Maryland and Virginia MEP Wages 1.1 • Average wage growth 2015-25 1.0 Ratio MEP to All Construction Wages 0.9 • Maryland and Virginia Electrical workers 0.8 3.96% (CAGR) 0.7 • All construction workers 3.51

23. Slide 23: Data Center Service Providers: A Significant Market Focus

    Macro Context: Data Center Construction Classes Group 1: Electrical, Mechanical, Cooling, Fire Protection Servicers Data Center Service Providers: • Installation and maintenance of electrical systems, switchgear, generator wiring A Significant Market Focus • Repair/maintenance of generators, UPS sys

24. Slide 24: Data Center Servicers: Group 1

    Data Center Servicers: 2027 All Lines Premium Group 1 Forecast 2027 Core Lines Loss Ratio Forecast 9.0% 0.80 • Electrical, Mechanical, Cooling, Fire 8.0% 0.70 7.0% 0.60 Protection Service Providers located 6.0% 0.50 proximate to VA data centers 5.0% 0.40 4.0% 0.30 • Modest growth in MD and VA 3.0% 2

25. Slide 25: Data Center Servicers: Group 2

    Data Center Servicers: 2027 All Lines Premium Forecast Group 2 8.0% 2027 Core Lines Loss Ratio Forecast 0.70 7.0% • Security, waste, and professional services 0.60 6.0% • Modest growth in MD and VA 0.50 5.0% 0.40 • Loss ratios more favorable, in general 4.0% 0.30 3.0% • Group 2 represents about $1.3

26. Slide 26: NatCat and Geographic Concentration Risks

    NatCat and Geographic Concentration Risks 26 ©2026 Verisk Analytics, Inc. All rights reserved .

27. Slide 27: The Natural Catastrophe Perspective (CRS)

    The Natural Catastrophe Perspective (CRS) Exposure Concentration Aggregation of insured value across portfolios and perils Benchmarking / Growth Existing market information lacks comprehensive details of the U.S. data center market for existing / potential portfolio decision making Improved Location

28. Slide 28: Verisk U.S. Data Center Exposure Database Product

    Verisk U.S. Data Center Exposure Database Product Examples of Location, Physical and Operational attributes captured • Capacity • Power Redundancy • Verisk BINS GeoPins • Building Level • Cooling Redundancy Rooftop Geocoding Footprint and Shapefile • Operator • Type • Area (Sq. Ft.) • Year Built • C

29. Slide 29: Example – 2 Building Colocation Campus

    CO LOCA TI ON Example – 2 Building Colocation Campus Sterling, VA CAMPUS OVE RVIE W 2 Operator QTS Facility Type Colocation Status Operational City Sterling County Loudoun BUILDIN GS ON THIS CAMPUS State VA ZIP Code 20166 Campus ID QTS Ashburn 2 Campus Capacity 75.0 MW 75.0 MW Aerial view — Esri, Ma

30. Slide 30: Concentration of Data Centers in U.S – All Perils / Regions Implicated

    Concentration of Data Centers in U.S – All Perils / Regions Implicated 30 © Verisk Analytics, Inc. All rights reserved .

31. Slide 31: Critical Infrastructure in the Crosshairs of Severe Thunderstorms

    Case Study – Hail Intensity in the U.S. Data Center Footprint Area (ft2) 2 Million 800k 300k 100k 20k Critical Infrastructure in the Crosshairs of Severe Thunderstorms 31 © Verisk Analytics, Inc. All rights reserved .

32. Slide 32: Water Availability and Emergency Response Considerations

    Water Availability and Emergency Response Considerations Hyperscale data centers, community first- response capability, and Public Protection 32 ©2026 Verisk Analytics, Inc. All rights reserved .

33. Slide 33: Community Fire-Protection Considerations for Data Center Development

    Community Fire-Protection Considerations for Data Center Development The exposure extends beyond the data center property line. Continuous demand Cooling-water demand can increase normal system consumption and reduce the supply margin available for firefighting. Concentrated development Multiple cam

34. Slide 34: Two Water Demands Must Be Evaluated Together

    Two Water Demands Must Be Evaluated Together Continuous cooling demand can coincide with an episodic fire-flow requirement. 1 Operational demand 2 Fire demand Seasonal cooling load Sprinklers Cooling technology Hydrants and hose streams Water source Required duration Peak-day conditions Extended ope

35. Slide 35: Measure the Remaining Firefighting Margin

    Measure the Remaining Firefighting Margin Annual consumption alone does not reveal fire-water resilience. Total reliable system capability Stress-test this margin against FLOW Needed fire flow Maximum-day community consumption PRESSURE Residual pressure Data-center operational demand DURATION Storag

36. Slide 36: First-Response Capacity Can Be Affected Before the Water Runs Out

    First-Response Capacity Can Be Affected Before the Water Runs Out Large incidents can consume the community’s operational reserve. UNIT AVAILABILITY Companies committed for an extended period SPECIALIZED RESOURCES Electrical, battery, generator, fuel, or hazmat expertise SYSTEM COVERAGE Move-ups, mu

37. Slide 37: Potential Public Protection Implications

    Potential Public Protection Implications A data center does not automatically change a classification. Measured community capability determines the effect. WATER SUPPLY FIRE DEPARTMENT COMMUNICATIONS + CRR Available supply after First-alarm response Alarm receipt and dispatch consumption Deployment

38. Slide 38: Build Resilience Into Development Approval

    Build Resilience Into Development Approval Consider an integrated, cumulative assessment and assign mitigation to the project. Evaluate Mitigate 1 Maximum-day and seasonal ✓ Dedicated fire-water storage and demand pumps 2 ✓ Reclaimed or non-potable Facility fire flow and duration cooling supply 3 ✓

39. Slide 39: Three Takeaways

    Three Takeaways 1 Evaluate the system, not only the building The objective Site compliance alone does not establish community-wide resilience. Enable development 2 Test peak and concurrent conditions without reducing Flow, pressure, duration, recovery, outages, and a second incident matter. communit

40. Slide 40: Strikes, Riots, and Civil Commotion Considerations

    6 Strikes, Riots, and WARNING If you see this message, the image placeholder Civil Commotion on the section divider has been removed. To restore the image placeholder, reset the layout by going to Home → Slides → Reset Considerations Keep image placeholder within the boundaries of this placeholder—d

41. Slide 41: Data centers as collateral damage during riots

    Data centers as collateral damage during riots 41 © Verisk Analytics Inc. All rights reser ved.

42. Slide 42: Data centers in the path of riot events

    Data centers in the path of riot events Downtown Minneapolis, center of the George Floyd riots in 2020 Source: Verisk, OSM 42 © Verisk Analytics Inc. All rights reser ved.

43. Slide 43: Data centers in the path of riot events

    Data centers in the path of riot events Buildings damaged by the riots; • Vandalism • Looting • Fire Source: Verisk, OSM 43 © Verisk Analytics Inc. All rights reser ved.

44. Slide 44: Data centers in the path of riot events

    Data centers in the path of riot events • Four data centers within two blocks of riot damage • Two data centers within damaged buildings Source: Verisk, OSM 44 © Verisk Analytics Inc. All rights reser ved.

45. Slide 45: Data centers in the path of riot events

    Data centers in the path of riot events • AAL from Verisk SRCC Model for the US • Verisk Data Centers Dataset • Examples of severe SRCC events in Minneapolis Source: Verisk Data Centers Dataset, Verisk SRCC Model for the US 45 © Verisk Analytics Inc. All rights reser ved.

46. Slide 46: Data centers as targets for protests

    Data centers as targets for protests 46 © Verisk Analytics Inc. All rights reser ved.

47. Slide 47: Protests in opposition to data centers by year globally

    300 Protests in opposition to data centers by year 250 globally Count of protests in opposition to data centers 200 150 100 50 0 2020 2021 2022 2023 2024 2025 2026 Source: Verisk, ACLED 47 © Verisk Analytics Inc. All rights reser ved.

48. Slide 48: Geographic distribution of protests in opposition to data centers

    United States - 296 South Korea - 14 Canada - 11 Nine countries – 2 to 5 each Six countries – 1 each Geographic distribution of protests in opposition to data centers since 2020 • 63% of all protests globally since 2020 have occurred in the United States in 2026 • 84% of all protests globally in 202

49. Slide 49: Data center opposition protests by U.S. state since 2026

    Data center opposition protests by U.S. state since 2026 • Texas the most in 2026 • Virginia the most since 2020 • Every state bar two has seen more protests in 2026 already compared to 2025 Source: Verisk, ACLED 49 © Verisk Analytics Inc. All rights reser ved.

50. Slide 50: Location of protests in opposition to data centers

    Location of protests in opposition to data centers 2% 2% 3% 93% Operational data center Under construction Site of planned development Other location, away from data center sites Municipal or government buildings, • Almost all peaceful prominent public places, offices of • Access to locations blocke

51. Slide 51: Protest motivations

    • Water usage • Conversion of land Protest motivations • Environmental impacts • Power usage and emissions • Proximity to other buildings like schools or communities • Concerns about potential health or quality-of-life effects on nearby communities • Development of supporting infrastructure (e.g., p

52. Slide 52: Potential insurance considerations associated with SRCC

    Potential insurance considerations associated with SRCC Data Centers as direct riot targets Data centers as collateral damage from riots • Enterprise or hyperscale data centers; • More likely to affect micro or small data centers given urban locations; • Buildings and contents damage and business in

53. Slide 53: Emerging Issues Custom Research Workshop

    7 Emerging Issues WARNING If you see this message, the image placeholder Custom Research on the section divider has been removed. To restore the image placeholder, reset the layout by going to Home → Slides → Reset Workshop Keep image placeholder within the boundaries of this placeholder—do not crop

54. Slide 54: Introducing the Emerging Issues Custom Research Workshop

    Introducing the Emerging Issues Introducing the Emerging Issues Custom Research Service Custom Research Workshop With the Emerging Issues Custom Research Workshop, you’ll receive actionable insights to prepare Receive your business actionable insights tofor the next prepare yourera of riskfor the bu

55. Slide 55: Engage with Emerging Issues

    Engage with Emerging Issues Sign Up for Our Weekly Newsletter ❖ Original research ❖ News updates ❖ Webinar & event alerts ❖ Curated newsfeeds Want a briefing on a specific emerging risk for your organization or team? Contact us: [email protected] 55 ©2026 Verisk Analytics, Inc. All rights re

56. Slide 56: Sources and Resources

    Sources and Resources 56 ©2026 Verisk Analytics, Inc. All rights reserved .

57. Slide 57: Sources and Resources

    Sources and Resources "Energy Demand from AI," International Energy Agency. https://www.iea.org/reports/energy-and-ai/energy-demand-from-ai "Hyperscale AI Data Centers Energy Usage: 2026 Breakthrough Technology," MIT Technology Review, January 12, 2026. https://www.technologyreview.com/2026/01/12/11

58. Slide 58: Sources and Resources

    Sources and Resources "FedRAMP vs. FISMA Compliance," Comparitech. https://www.comparitech.com/data-privacy-management/fedramp-vs-fisma-compliance/ "FedRAMP Authorization Scope," FedRAMP, U.S. General Services Administration. https://www.fedramp.gov/docs/authority/scope/ "Title 40, Part 63, Subpart

59. Slide 59: Sources and Resources

    Sources and Resources "Communities Are Raising Noise Pollution Concerns About Data Centers," Environmental and Energy Study Institute. https://www.eesi.org/articles/view/communities-are-raising- noise-pollution-concernsabout-data-centers "Data Center Approvals," Milrose Consultants. https://www.milr

60. Slide 60: Sources and Resources

    Sources and Resources "Data Center Expansion in Virginia: Closing Critical Gaps for Informed Water Planning and Permitting," Center for Secure Water, University of Illinois Urbana-Champaign. https://securewater.illinois.edu/data-center-expansion-in-virginia-closing-critical-gaps-for-informed-water-p

61. Slide 61: Sources and Resources

    Sources and Resources "Mitigating Schedule Risks in Data Center Projects," iRecruit. https://www.irecruit.co/insights/mitigating-schedule-risks-data-center-projects "Zurich Shares Firsthand Insights in Data Center Risks Right Now," Zurich North America, 2026. https://www.zurichna.com/media/news-rele

62. Slide 62: Sources and Resources

    Sources and Resources "NERC 10-Year Peak Demand Forecast Jumps 24% on New Data Center Loads," Utility Dive. https://www.utilitydive.com/news/nerc-10-year-peak-demand-forecast-jumps-24-on- new-data-center-loads/810955/ "Data Center UPS Systems," MEP Academy. https://mepacademy.com/data-center-ups-sys

63. Slide 63: Sources and Resources

    Sources and Resources "The AI Revolution: How Fuel Cells Are Solving the Data Center Power Challenge," Bloom Energy. https://www.bloomenergy.com/blog/the-ai-revolution-how-fuel-cells-are-solving- the-data-center-power-challenge/ "Stressed US Grid Forcing Data Centers to Get More Flexible," Reuters,

64. Slide 64: Sources and Resources

    Sources and Resources "Drained by Data: The Cumulative Impact of Data Centers on Regional Water Stress," Ceres. https://www.ceres.org/resources/reports/drained-by-data-the-cumulative-impact-of- data-centers-on-regional-water-stress "2025 Q2 Water Market Insider: Data Centers," WestWater Research, 20

65. Slide 65: Sources and Resources

    Sources and Resources "The Cybersecurity Blind Spot in Data Center Building Systems," Data Center Dynamics. https://www.datacenterdynamics.com/en/opinions/the-cybersecurity-blind-spot-in-data- center-building-systems/ "AI Data Centers: Cyber-Physical Risk," DeNexus. https://www.denexus.io/resources/

66. Slide 66: Sources and Resources

    Sources and Resources "Judge Halts Google Data Center Project in Minnesota," Construction Dive. https://www.constructiondive.com/news/judge-halts-google-data-center-project/822152/ "CoreWeave, Inc. Class Action Lawsuit," Kessler Topaz Meltzer & Check. https://www.ktmc.com/crwv-coreweave-inc-class-ac

67. Slide 67: Sources and Resources

    Sources and Resources "Inference Becomes the Next AI Chip Battleground," Data Center Knowledge. https://www.datacenterknowledge.com/data-center-chips/inference-becomes-the-next-ai-chip- battleground "AI Training vs. Inference: How Do They Impact Cooling?," AIRSYS North America. https://airsysnortham

68. Slide 68: Sources and Resources

    Sources and Resources "Fire, Water, Knock Out Google Cloud in Paris," The Stack. https://www.thestack.technology/google-cloud-paris-water-fire-europe-west9/ "Google Cloud Suffers Outage in Europe Amid Water Leak, Fire," The Register. https://www.theregister.com/off-prem/2023/04/26/google-cloud-suffe

69. Slide 69: Sources and Resources

    Sources and Resources "Americans Oppose AI Data Centers in Their Area," Gallup. https://news.gallup.com/poll/709772/americans-oppose-data-centers-area.aspx "Power-First Development: How Data Center Projects Are Now Designed Backwards," Datacenters.com. https://www.datacenters.com/news/power-first-de

70. Slide 70: Sources and Resources

    Sources and Resources "From Incident to Insight: Fire Risk in Modern Data Centers," ScienceDirect. https://www.sciencedirect.com/science/article/pii/S095042302500350X "NERC Issues Rare Level 3 Alert Over Data Center Load Losses," Utility Dive. https://www.utilitydive.com/news/nerc-issues-rare-level-

71. Slide 71: Sources and Resources

    Sources and Resources "Judge Halts Google Data Center Project in Minnesota," Construction Dive, June 5, 2026. https://www.constructiondive.com/news/judge-halts-google-data-center-project/822152/ "DOE: AI Data Centers Are Transforming America's Transmission Map," Data Center Knowledge, July 24, 2026.

72. Slide 72: Before they Were ISO Forms … They Were Emerging Issues

    Before they Were ISO Forms … They Were Emerging Issues 72 ©2026 Verisk Analytics, Inc. All rights reserved .

73. Slide 73: Thank you

    Thank you About Verisk This material was used exclusively as an exhibit to an oral presentation. It Verisk is a global data analytics provider may not be, nor should it be, relied on as reflecting a complete record of the helping the insurance industry make discussion. No part of this presentation s

Verisk Analytics — Emerging Issues Webinar: AI Data Centers: An Evolving Risk Landscape (October 2026) on 6ix