War Enters the Risk Register for Data Centre Operators
A Howden report counts 14 large data centre lawsuits in the first half of 2026 against four in all of 2021, as planning fights, noise complaints and war risk reshape how AI infrastructure is underwritten.
Large lawsuits involving data centres more than tripled from four in all of 2021 to 14 in the first half of 2026, according to a Howden report, with planning and environmental disputes the main drivers and war now treated as a core risk category for operators.
The insurance file on a data centre used to be short. Fire, flood, equipment breakdown, a business-interruption clause tied to uptime. That file is getting thicker, and two of its newest chapters — litigation from neighbours and regulators, and physical exposure to armed conflict — are the ones underwriters are now spending the most time on.
A report from broker Howden, summarised by OilPrice, counts large lawsuits involving data centres rising from four in all of 2021 to 14 in the first half of 2026 alone. Planning and environmental disputes drive most of the recent cases. Noise and nuisance complaints — diesel generator emissions, water consumption — have become increasingly common. And war, once a line item excluded and forgotten, is now treated as a core risk category for operators.
A litigation curve that is steepening, not drifting
Four cases in a full year to 14 in six months is not a gradual escalation. It is a change in the character of the asset class. Compared against the 2021 baseline, the first half of 2026 alone already produced more than three times as many large suits as the whole of that earlier year. If the second half of 2026 merely matched the first, the full year would land at 28 — seven times the 2021 figure. That annualisation is an illustration, not a forecast, and second-half filings could run higher or lower.
What matters more than the count is the composition. Planning and environmental disputes are contests over whether a facility gets built, how fast, and on what terms. They land before the concrete is poured or while the site is ramping. That makes them a schedule risk as much as a legal one, and schedule risk in data centres is expensive: capacity is typically pre-sold on contracted delivery dates, and construction financing is priced against them.
Noise and nuisance claims sit at the other end of the life cycle. A generator test that rattles windows, a cooling system that draws heavily on municipal water in a dry summer — these are operational facts that become legal facts once a neighbourhood organises. They rarely threaten the asset, but they impose a running cost: mitigation capex, community settlements, legal spend, and the management attention that goes with all three.
Why war stopped being an exclusion and became a category
Treating armed conflict as a core risk category is the more striking shift. Conventional property policies have carried war exclusions for a century precisely so that insurers did not have to price the unpriceable. Reclassifying war as a category to be assessed — rather than simply carved out — reflects where the infrastructure now sits and what it now does.
Data centres have become strategically legible targets. They concentrate compute, they concentrate power draw, and they are increasingly identified in public planning documents and grid interconnection queues. They also depend on long, fragile chains: subsea cable landings, transformer supply, fuel deliveries for backup generation, and regional transmission. A conflict does not need to reach a building to take it offline. It only needs to reach one of those chains.
The practical consequence for operators is that political violence, terrorism and sabotage cover moves from an afterthought to a negotiated line in the programme — and that siting decisions start carrying a geopolitical score alongside the familiar tests of power availability, fibre routes, water and tax treatment.
Where the cost actually lands in the model
Nobody has published a clean per-megawatt number for what this adds, and inventing one would be worse than useless. But the transmission mechanism is easy to trace. Data centre economics rest on a spread between contracted revenue per megawatt and the all-in cost of delivering it — construction, power, staffing, maintenance, insurance and now legal defence. Insurance and litigation costs sit in the operating stack, which means they compress the yield on stabilised assets and raise the hurdle on new ones.
Nobody has published a clean per-megawatt number for what this adds, and inventing one would be worse than useless.
Three groups feel it differently:
- Colocation and wholesale operators carry the exposure most directly. They own the buildings, sign the community agreements, and answer the lawsuits. Rising premiums hit their net operating income unless they can pass costs through in lease escalators.
- Hyperscalers can absorb the cost more easily but face the schedule risk acutely, because AI capacity commitments are dated and public. A planning dispute that delays a campus by quarters is a competitive problem, not just a financial one.
- Lenders and infrastructure funds underwrite these assets on long horizons. A litigation trend that is compounding rather than flat changes the assumptions behind debt sizing and terminal value.
The market backdrop the risk is arriving into
None of this is slowing the capital. Equity markets closed the session on Monday, 21 September 2026 with the AI-weighted end of the tape leading: the Nasdaq 100 tracker (NASDAQ: QQQ) finished at $741.47, up 2.77% on the day from a previous close of $721.45, with a day range of $727.81 to $743.22. The S&P 500 proxy (NYSEARCA: SPY) closed at $773.50, up 1.55%, and the Dow tracker (NYSEARCA: DIA) at $519.78, up 0.76%. The gap between the three — roughly a point and a fifth between the Nasdaq and the Dow moves on the day — is the familiar shape of a market still paying up for compute.
That is the tension. Capital is being allocated to AI infrastructure at a pace that assumes sites get permitted, built and energised roughly on time. The Howden data points the other way: the friction around those sites is increasing, and the parties generating it — residents, municipalities, environmental regulators — are getting better at it.
What to watch from here
The second-half 2026 lawsuit count is the immediate number. If filings continue at the first-half pace, insurers will have a full-year trend line rather than a spike, and renewal pricing in 2027 will reflect it.
Watch also for how the risk migrates into contracts. Force majeure language, delay-in-start-up cover, and who bears permitting risk between developer and tenant are the clauses where this trend shows up first. Water usage disclosure is another tell: once consumption figures become routinely public, nuisance claims get easier to bring and harder to defend.
Finally, watch siting. If war risk is genuinely being scored, the effect should appear in geography — more capacity in jurisdictions judged politically stable and away from contested corridors, subsea chokepoints and borders. That reallocation, if it happens, will be the clearest evidence that the risk register has actually changed behaviour rather than just paperwork.
Key facts
- Large data centre lawsuits, H1 2026: 14, per Howden report
- Large data centre lawsuits, full-year 2021: 4
- Main litigation drivers: Planning and environmental disputes; noise, generator emissions, water use
- Nasdaq 100 tracker (QQQ): $741.47, +2.77%, close of 21 Sep 2026
Frequently asked questions
How much has data centre litigation increased?
According to a Howden report, large lawsuits involving data centres rose from four in all of 2021 to 14 in the first half of 2026 alone. That is more than a tripling, and because the 14 figure covers only six months, the underlying pace of filings is running well above the 2021 baseline rather than simply catching up to it.
What are these lawsuits actually about?
Planning and environmental disputes are the main drivers of recent legal action against data centres. Alongside those, noise and nuisance complaints have become increasingly common, including objections to emissions from diesel backup generators and to the volume of water consumed by cooling systems at large campuses.
Why is war now treated as a core risk for data centre operators?
War has moved from a standard policy exclusion to a category operators are asked to assess. Data centres concentrate compute and power demand, are publicly identifiable through planning and grid documents, and depend on subsea cables, transformers, fuel deliveries and transmission. Conflict can disrupt any of those chains without physically reaching the building itself.
How do insurance and litigation costs affect data centre economics?
Both sit in the operating cost stack, between contracted revenue per megawatt and the all-in cost of delivering it. Higher premiums and legal spend compress net operating income on stabilised sites and raise the return hurdle on new ones. Planning disputes add schedule risk, which matters because capacity is usually pre-sold on dated delivery commitments.
Who bears the cost — operators or hyperscalers?
Colocation and wholesale operators carry the most direct exposure because they own the buildings and face the claims, and they must recover costs through lease escalators. Hyperscalers can absorb premium increases more easily but are more sensitive to permitting delays, since AI capacity commitments are dated and visible to competitors and investors.
What did equity markets do on 21 September 2026?
At the close on Monday, 21 September 2026, the Nasdaq 100 tracker QQQ finished at $741.47, up 2.77% from a prior close of $721.45. The S&P 500 proxy SPY closed at $773.50, up 1.55%, and the Dow tracker DIA at $519.78, up 0.76%. The AI-weighted index led the session.
