Technology

First published

The Data Centre Owner May Not Be the One Who Gains Most From Flood Protection

Our interpretation

A model of climate risks to data centres suggests that the person paying for protection may not receive most of its financial benefit. Site-specific risk and demand for work remain unverified.

The Signal in brief

Why it matters

A landlord and tenant may need to agree who pays for a protective upgrade and who benefits before a useful project can go ahead.

The model is not proof of loss at a particular site. Full report methods are gated and no new paid order or margin was verified.

Explore this signal

Emerging data

A data centre can be new and still sit in the path of a flood, wildfire or damaging wind.

A report from climate-risk company Jupiter Intelligence, listed on its resources page on 13 November 2026, models expected risks across 14,073 operating sites in 187 countries. Its public summary estimates US$4.4 billion in annual expected losses from flood, wind and wildfire. Those are modelled losses, not recorded bills or proof that any particular building is unsafe.

The report identifies a potential problem in who pays for protection. Equipment can belong to a tenant while building changes are paid for by a landlord. In Jupiter's model, raising the lowest floor to the 100-year flood level cuts total expected losses by 23%; at third-party facilities, the landlord retains an estimated 22% of the benefit. These figures depend on the model and are not a quote for any individual property.

We think the opportunity may lie in getting owners and tenants to agree on a practical remedy. A qualified engineer could compare the actual site risk, the lease responsibilities and the cost of a specific protection measure. The parties could then decide whether and how to share the cost. No new paid repair order or service margin has been established here.

For someone seeking work, the first step is to speak with an existing qualified engineering or facilities firm about a narrowly defined assessment. Obtain the required permissions, a real customer scope and a price covering labour, insurance and site access before taking on work. Do not claim expertise you do not have.

The idea weakens if the property is not materially exposed, a tenant or landlord already handles the assessment, the lease allocates the cost differently, or the proposed remedy costs more than the expected benefit. The full report is gated, so its complete methods were not independently reviewed; its summary says the model excludes convective storms, hail and grid failures.

Sources: Jupiter Intelligence report summary, listed 13 November 2026, https://info.jupiterintel.com/data-center-report-risk-in-the-racks and https://www.jupiterintel.com/resources ; The Business Times, 13 November 2026, 18:12 SGT, https://www.businesstimes.com.sg/international/data-centre-investors-underestimate-climate-risks-see-these-not-priority-report .

Report this passage

What could happen next

  1. First-order consequence

    A site owner and tenant can compare exposure and identify who owns vulnerable equipment.

  2. Second-order consequence

    A qualified engineer may identify a protection measure and estimate its cost.

  3. Third-order consequence

    The parties could negotiate how to share costs and expected benefits.

  4. Fourth-order consequence

    A viable agreement may lead to a contracted upgrade; that is a forecast, not a verified order.

  5. Fifth-order consequence

    Low site risk, high upgrade costs or lease disagreements may prevent the work or make it uneconomic.

What would change our view?

What would disprove it

A site has low exposure, protective work is already in place, or the cost exceeds the expected benefit.

Back to Signals