A Present Value Mispricing

Data centres occupy a peculiar position in modern capital markets. They are financed as long-duration infrastructure assets but operate like energy-intensive industrial facilities. The investment case is built on two pillars: future contracted revenues support the upside of expected returns, while strong asset valuations anchor the downside. Global data centre capacity is expected to triple over the coming decade, and both equity and debt capital has followed at scale, with confidence in the macro tailwinds.

But beneath the contracted revenue visibility sits a cost base that is not static. And it is in the divergence between the stability assumed at the top line and the drift accumulating at the bottom line, that risk accumulates.

A key driver of that divergence is the drift in climate conditions. Not just the acute, catastrophic version of climate hazards that headlines favour and investors dismiss as rare. The chronic version matters equally: a gradual shift in the operating conditions on which these assets were valued and financed.

If those shifts are absent from asset valuation and discounted cashflow models, both the present value of the asset and the present value of its cashflows are systematically mispriced. The asset continues to operate, revenues remain contracted, but the financial model is anchored to a risk profile that no longer exists.

From Invisible to Measurable

For a long time, the link between physical climate hazards and financial outcomes remained weak. Climate data was coarse and slow to update, and translating it into asset-level cashflows required assumptions that were difficult to defend. The result was acknowledgement without integration: climate risk appeared in qualitative reporting, but not in pricing and valuation.

That constraint is breaking down. Advances in Earth observation (from organisations such as the European and UK Space Agencies, and institutions such as the European Centre for Medium-Range Weather Forecasts) are changing the resolution at which physical risk can be assessed.

Satellite data is beginning to fill the observation gap that left us blind to regional and location specific climate risk. With observations at individual asset latitude-longitudes, and temporal granularity now measured in hours rather than months, the limiting factor is no longer data. It is the ability to translate it into something markets can price: climate risk adjusted cashflows and basis points of asset re-valuation.

In the analysis that follows, we bridge this gap through the analysis of two data centre case studies: one in Spain, where heat stress is increasingly eroding cashflows; one in the UK, where shifting flood risk is altering projected asset value. Different hazards, different entry points into the financial model, same conclusion.

Climate Risk is a Present Value Problem

Climate change is habitually framed as a long-term problem that will materialise gradually, in some sufficiently distant future that it can safely be excluded from today’s investment mandate. Valuation does not work on that timeline.

Long duration assets like data centres are not priced on today’s conditions. They are priced on the present value of 20-30 years of future expected performance. In credit markets, risk is repriced continuously as default probabilities evolve. In equity markets, expected cashflows adjust as operating performance shifts.

Why should climate risk behave any differently?

A flood should not need to occur for an asset to be repriced. A shift in the probability structure of flood occurrence, and in the severity (in metres) given occurrence, should be reflected in value today. Rising summer temperatures do not need to cause system failure to matter financially. A gradual increase in data centre energy demand is enough to erode margins and compress the present value of future cashflows.

The framework is straightforward:

When climate risk is systematically underestimated, both drivers of that equation are affected. Expected losses to asset value are underpriced as hazard probabilities evolve. Expected cashflows are overstated as operating costs rise. The result is a mispricing. Not a future risk, but a present one, embedded in today’s valuations and transactions.

Two Channels, One Conclusion

Physical climate risk enters asset valuation through two distinct channels. Understanding both is necessary to see how mispricing emerges and how it accumulates in the capital structure.

Channel 1: Cashflow Erosion

Climate conditions directly influence the operating cost base. For data centres, one of the primary mechanisms is heat stress and its effect on cooling demand.

As ambient temperatures rise, cooling systems must work harder to maintain thermal conditions, and the relationship between temperature and energy demand is non-linear. Each additional degree does not add a constant cost. It increases the rate at which costs grow.

Where revenues are contracted and utilisation assumptions hold, this is easy to overlook. The top line looks stable. But the cost base is drifting, and a gradual shift in operating conditions is sufficient to reduce free cashflow over the years.

This dynamic is already observable in real-world operations. Empirical studies of industrial manufacturing show that even small increases in temperature reduce production efficiency at the factory level, while modelling of data centres indicates that rising temperatures can increase energy consumption by close to 10% through higher cooling demand.

Channel 2: Asset Value Drift

Where heat stress erodes an assets cashflows, flood risk alters the value profile of the asset itself. Flooding is widely modelled as a probabilistic function of return period: the magnitude of an event expected to occur once in every X years.

As climate change shifts glacial melting, precipitation patterns and flood basin behaviour, that function is evolving. At locations across the world, the expected probabilities of occurrence, and the severity (in metres) given occurrence are increasing.

The financial framework mirrors credit risk:

As both flood risk components (probability and severity) shift upward, expected loss increases, even where no flood event has occurred, yet.

However, unlike credit risk, where terabytes of market data support calibration of default models, physical climate hazard data is thinner and tail distributions are less well characterised.

The analogy is directional rather than calibrated to the same statistical precision. What it captures correctly, though, is the structure: both the frequency and severity of loss are evolving, and valuation should reflect that evolution.

In many cases, models have already indicated a material shift in climate risk probability and severity distributions ahead of the event. During major European floods (2024), climate models accurately predicted the scale of rainfall (300–400mm) before the event.

Case Study 1: Heat Stress and Cashflow Erosion, Spain

Northeastern Spain is attracting multi-billion-euro investments for hyperscale data centre construction. It is also characterised by high summer temperatures, growing water stress, and some of the most pronounced temperature drift on the European continent.

Data centres in Europe have historically been underwritten using temperature distributions that reflect relatively stable climate conditions. Cooling systems are designed around those distributions, with energy consumption modelled accordingly. That distribution is now shifting , not just through extreme spikes, but across the entire annual temperature distribution.

The temperature distributions shown below are derived from models combining satellite observation datasets (ERA5) with forward climate projections (CMIP6), allowing both historical observation and forward-looking drift to be measured consistently at asset level.

The key point is not the increase in occurrences of peak temperature days. It is the shift in the full distribution: what was previously an upper tail event is increasingly part of the baseline operating environment.

To translate this into financial terms, we apply a simplified cooling burden model grounded in empirical research on data centre system behaviour, directly onto the observed and projected temperature distributions:

•        Sub 30°C: Cooling systems in stable operating regime.

•        30°C - 35°C: Linear additive increase in cooling demand.

•        35°C - 40°C: Accelerated increase, thermally stressed regime.

To understand the temperature-linked cost impact for a data centre in the selected region of Villanueva del Gallego, Spain, we model a representative data centre a total IT load of 250 MW (Power Usage Effectiveness, PUE, normalised to 1.0 for analytical clarity), implying an annual baseline electricity consumption of approximately 2.2 TWh, or 6,000 MWh per day.

The result is a persistent upward drift in additional energy demand, not driven by a single extreme event, but by an increasing proportion of operating time spent in high-cost temperature regimes.

The three figures below trace this chain from volume to value. Figure 6 shows the incremental energy burden (GWh), the additional electricity consumed purely as a result of rising temperatures in the geographical region.

* A PUE of 1.0 is used as a modelling simplification. In practice, typical values range from 1.2 to 1.6. The relative climate-driven uplift applies regardless of the baseline PUE assumption, but real asset figures should be scaled accordingly.

Figure 7 translates that volume into cost, applying a representative Spanish baseload electricity price (€75/MWh). Figure 8 then illustrates the resulting cashflow compression under these projected climate conditions over a 20-year asset lifetime.

The first two inputs to any data centre valuation, utilisation and contracted revenue, receive detailed scrutiny. The third, the cost base, is typically anchored to historical operating conditions. If those conditions are shifting, the cost base is being systematically understated.

Market risk compounds this climate risk factor. The same climatic conditions that increase electricity consumption tend to coincide with system-wide stress in energy markets, where demand pressures can rapidly elevate prices. In addition, geopolitical dynamics creating disruptions to global energy flows (such as tensions around key transit routes like the Strait of Hormuz) can shift input costs for power generation.

The result is a layered exposure: climate drives higher electricity demand, while stressed market conditions can simultaneously increase the price of each MWh. The gap between base and stress scenarios reflects the convex  cost profile of such climate effects: operators are forced to procure more power, and to do so at higher prices.

Incremental energy demand driven by heat stress is amplified by market volatility, resulting in a materially larger impact on total energy costs than either factor would imply in isolation.

For equity investors this translates into lower cashflow accumulation, compressed internal rates of return. For lenders, gradual erosion of debt and service coverage ratios over the life of the loan.

The system continues to operate. But the cost of keeping it running increases, against assumptions set under a different climate.

Case Study 2: Flood Risk and Asset Value Drift, United Kingdom

Where heat stress implies cashflow impacts, severe flooding has the capability to significantly damage the value of the asset on the balance sheet. Flood maps calibrated on historical data are increasingly inadequate as precipitation, riverine and coastal dynamics embedded in them are anchored to a climate that is no longer the operating baseline.

As the occurrence of extreme events intensifies, the relevant geography is no longer just the site itself, but the entire upstream catchment. A storm occurring tens of kilometres away can generate river discharges that propagate downstream, affecting assets well outside standard mapped floodplains.

The analysis below uses a representative UK location with a data centre asset value of £2.5 billion, with a credit line of £1.25 billion (and therefore an initial loan-to-value of 50%). Flood depth is represented as a probabilistic function of return period under historical observations and forward-looking climate scenarios.

The critical point: for a given return period, a 1-in-25-year event, a 1-in-50-year event, the expected magnitude is increasing. The same probability now carries higher physical impact. The distribution has shifted not merely at the tail but across the body.

Using a representative damage function from the Joint Research Centre (European Commission), we map flood magnitude to proportional asset loss.

This introduces the second dimension: not only are events becoming more severe, but the damage associated with those events is increasing across the distribution. Both the probability and the severity of loss are evolving simultaneously.

As both components of expected loss shift upward, asset value declines, not because anything has failed, but because the risk profile has evolved. Climate models and satellite observations now allow both components to be continuously updated at asset level.

For our representative £2.5 billion asset, this produces a measurable erosion of value. Not due to a realised flood, but due to the evolution of the underlying risk profile. More specifically, value adjusts to reflect the increasing probability that flooding-linked cashflows will occur within the asset’s operating lifetime.

Propagation Through the Capital Structure

The implications do not stop at the asset. They propagate through the capital structure in ways that are structurally important and often invisible from either direction.

For equity investors, the impact is felt at exit. As the expected future value of the asset declines, the residual value, and therefore the multiple on invested capital, compresses. Total return to equity is lower, not because revenues underperformed, but because the asset was priced on a risk profile that had already shifted.

For lenders, the dynamic is more insidious. As climate-adjusted asset value declines, the collateral value underpinning the loan erodes and loan-to-value ratios drift upward.

Infrastructure debt facilities are typically structured with LTV covenants in the 60–75% range. The danger of gradual drift in climate risk exposure is that it may not breach LTV and collateral quality thresholds while risk remains embedded in expectation rather than reflected in price.

Valuation models continue to show full collateral value, even as the probability and severity of climate-related loss increase in the background.

Unlike market risk, which reprices continuously, this form of collateral erosion develops gradually within the underlying risk profile but often remains unrecognised in valuation. Expected flood risk increases incrementally (through shifting probabilities and severities) but is not fully embedded in asset values or lending metrics.

When a climate event, such as a flood, eventually occurs, that embedded risk is realised in physical damage and the gap between assumed and actual collateral value closes in one step. What should have been a gradual adjustment instead becomes a step-change in LTV and collateral quality. 

By that point, remediation options are materially worse than they would have been at origination: additional equity may not be available, asset sales may be forced under stressed valuations and refinancing may be more expensive or unavailable. A sudden repricing event reflecting a gap that had been building over time.

The Repricing Gap

Across both impact drivers, the mechanisms differ. The implication is the same.

Assets continue to be valued on stable operating conditions, static cost assumptions, and historical risk distributions. In reality, cost bases are shifting, hazard distributions are evolving, and expected losses are increasing. The result is not a future adjustment. It is a present mispricing, embedded in today’s models, compounding across the duration of the investment.

This should be familiar territory. In credit markets, a company does not need to default for its bonds to reprice. A shift in the probability of default is sufficient. The same principle applies here. The asset does not need to flood for its value to change. A shift in the probability distribution of flooding, and in the damage at each point on that distribution, is a sufficient marker of elevated financial risk.

As climate risk becomes more measurable, the gap between what can be known and what is priced becomes more visible. That gap will not persist indefinitely. The repricing, when it comes, is unlikely to be driven by a single catastrophic event. It will be driven by a gradual realisation, by enough participants, simultaneously, that the assumptions embedded in existing models no longer reflect the operating environment.

In practice, this means financial markets should:

• Reprice cost assumptions using forward climate-adjusted conditions

• Integrate expected loss into asset valuation, not just insurance overlays

• Stress-test IRR and LTV under shifting climate distributions

Those who price climate-adjusted cashflows, expected losses, and collateral now will make decisions on the basis of accurate valuations. Those who do not will discover, at exit or at repricing, that the gap had been accumulating for years.

WieldMore Investment Management is an FCA-authorised firm specialising in climate aware investment management. This article draws on Pomelo, WieldMore’s climate risk analytics platform.