Trillions are flowing into renewable energy and digital infrastructure assets, based on long-dated cash-flow projections that quietly assume the climate behaves. It doesn’t, and a new wave of regulation is about to make asset owners and lenders price the difference.

Bets on climate volatility

The two largest destinations for institutional real-asset capital this decade are, underneath the financial wrapper, bets on the climate.

Data-centre capital expenditure reached roughly $770 billion in 2025, overtaking upstream oil and gas and level with the entire renewable-generation industry; data centres alone took about a quarter of all infrastructure capital raised, digital infrastructure overtook renewables by value for the first time, and renewables still ran about a third of deal volume (Source: Rystad Energy; CBRE Investment Management, 2025).

A pension fund or insurer allocating to “real assets” today is, more likely than not, buying a data centre that must be kept cool or a renewable energy asset that must receive a minimum solar irradiation. Both run on a climate that no longer reads the investment case.

The cash flows and  the valuations of these assets are, in all likelihood, both mispriced against emerging climate volatility. Climate change is already revaluing the cash flows and asset values of real assets, whether or not the market chooses to price it, and the proof is in the data.

Regulation is now catching up on a mispricing emerging in the real world, not generating the mispricing through policy: the drivers for climate-related asset valuation and cashflow implications existed before the rules and will continue without them.

What follows explains why the mispricing has lasted, what is now forcing it into the open, and, through one worked example in renewable asset portfolios, how the revaluation bites.

Climate change is already revaluing these cash flows and asset values, whether or not the market chooses to price it.

The risk that moved undetected

A physical asset is, in part, worth the discounted value of its cashflows. It is often financed, pledged as collateral, levered and securitised against this projection.

Every one of those steps embeds the same assumption: that the asset performs as modelled. That the data centre stays cool enough, and far enough from wildfires and floods, to honour its uptime guarantees. That the solar farm receives the irradiance needed to produce the megawatt-hours the bank case assumed.

Each is a climate assumption, and for years no one priced it. A risk that neither buyer nor seller prices does not cease to exist; it simply moves undetected, distorting the cash flow and the valuation alike, until something forces it into the open.

The seller of a solar asset and the buyer of its revenue stream discount the same optimistic generation profile, so the price clears and the risk vanishes from view: not because it was small, but because it was symmetric. Naming this risk, and measuring it, is the opportunity.

The regulator catches up

The clearest sign the market knows this is already under way comes from the most conservative institution in the system. From 15 June 2026 the European Central Bank applies a “climate factor” to the collateral banks pledge for liquidity: a discount, asset by asset and on top of the usual haircut, that lowers the value of debt issued by climate-exposed firms (Source: ECB Guideline ECB/2026/1; ECB press release, 29 July 2025).

It is the first time a major central bank has put a number on climate risk in collateral. Tellingly, it scores each asset’s vulnerability by the square root of its remaining maturity. The longer the financing, the larger the discount (Source: ECB Guideline ECB/2026/1, Annex XIIb). Duration magnifies climate risk, and the ECB has written it into a formula.

The factor covers transition risk , the repricing as the economy decarbonises, and not, yet, the physical risk; the ECB says the scope will widen as data and models allow.

Climate risk has two halves, transition and physical; the central bank has drawn the first into the system and left a placeholder for the second. Nor is it only the central bank. Development lenders are moving the same way: the EBRD now screens all its new direct financing for climate alignment and physical climate risk and increasingly expects borrowers to quantify it as a condition of the loan (Source: EBRD, Physical Climate and Carbon Transition Risk Assessment procedures).

For an owner with a thirty-year horizon, the prudent assumption is that physical risk is next to be priced in. What has changed most is the conversation. The institutions grappling with this are visibly relieved that it is no longer only about sustainability but about cash flows, asset values and collateral, the things they have always priced. Once a regulator discounts an asset for climate, every holder needs a way to measure the discount and defend it to an auditor.

To see what is being measured, take one asset class in detail.

The case study: what the repricing looks like on the ground

Solar is the cleanest illustration, because its output is visibly a climate variable and because Spain, the continent’s solar showcase, has the data to prove it.

Spain drew 55.5% of its electricity from renewables in 2025 (Source: Red Eléctrica de España); yet that year the Iberian Peninsula ran below its long-run average for solar radiation while grey, maritime Britain had its sunniest year on record (Source: WMO, State of the European Climate 2025).

The country everyone underwrites for sunshine underdelivered the same year the country no one bets on overperformed. That gap, between the investment case modelled output and the realised megawatt-hours, is the financial risk in one line.

Each point in Figure 1 is one of the sixteen largest Spanish solar plants individually identified and located, from European public ENTSO-E per-unit register, geolocated against the Global Energy Monitor tracker. Location points are colour graded by the share of nameplate MWh actually delivered over a sample week.

The two circled are the case study: Mula, in the Mediterranean south-east, and Garnacha, in the Atlantic-influenced north-west, opposite corners of the country, which is the experiment.

The number everyone quotes is the wrong one

The mainstream scorecard counts installed capacity (the megawatts on the nameplate). But a nameplate is a promise about peak output under ideal conditions, not what an asset delivers on a given afternoon, still less the cash it banks over a year.

The slack that used to absorb the gap is vanishing, as AI and its data centres add load that is large, continuous and unforgiving.

In a system with no slack, value is set by realised rather than installed megawatt hours, and realised output is a weather variable, not an engineering constant.

Across Spain’s reporting solar fleet, the same nameplate translated into very different energy output: realised capacity factors (actual generation as a share of theoretical nameplate generation) ranged from about 16% to 38% in a single week (Source: ENTSO-E Transparency Platform, generation 16.1.A ÷ installed capacity 14.1.B).

A megawatt is a promise; a megawatt-hour is a delivery; a balance sheet is paid in deliveries.

Solar doesn’t fail quietly

Solar’s marginal cost is essentially zero. The fuel is free, the capital sunk. When the sun is there, solar is the cheapest power on the system and sets a low price. When it is not, demand is met by the next plant in the merit order (the queue in which a grid calls on generators cheapest first) and on most stressed European evenings that plant burns gas, priced off a volatile, imported fuel.

The substitution is large and observable. Over a sample week, hourly wind-plus-solar and gas output correlated at −0.71. As renewables fall, gas rises to fill the gap (Source: ENTSO-E, Actual Generation per Production Type 16.1.B&C). And price tracks the shortfall: the lowest-renewable day printed a daily baseload (the day’s average wholesale price) near €100/MWh, the highest near €22 (Source: ENTSO-E day-ahead prices), a more-than-fourfold swing, largely driven by weather dependent renewable output.

Figure 2 makes the substitution visible: the shaded area is Spain’s combined wind-and-solar output over one week, the line its gas output, hour by hour. They move in opposition, every dip in renewables met, almost in lockstep, by a rise in gas.

That mirror image turns a cloudy spell into a price spike, and for the owner of the solar asset, into revenue at its lowest, just when power is dearest: the megawatt-hour the cloud erased is one never generated and never sold, on the day the market paid best.

Where the reliable, independent megawatt-hours are

If realised output is what matters, then where you build is not a detail, it is part of the asset. Two solar farms with identical nameplate capacity in different places represent different cashflows: different yield, different volatility, and importantly for portfolios of similar assets, different correlation with each other. Pricing such portfolios depends also on acknowledging and incorporating this correlation.

Here the market makes a revealing mistake. Correlate the sunshine two Spanish sites actually receive day to day and you get about 84%, high enough for a renewable energy fund to conclude that geographic spread buys it almost nothing (Source: NASA POWER, 2021–2025). But the 84% is mostly illusion.

The clear-sky potential of the two sites (the sunshine they would get with never a cloud, set purely by sun angle and season) correlates at 99%. Pure astronomy, identical at both, impossible to diversify because it is not a risk at all. Strip it out and divide each day’s actual sunshine by its clear-sky potential, leaving only the cloud effect, and the correlation collapses to 25% (Source: NASA POWER, 2021–2025).

Picture washing hanging on a line at two houses far apart. The hours of daylight each gets (99%) follow the calendar identically. How much washing actually dries (84%) mostly tracks the daylight, so the two look alike. But whether it rained on your line today (25%) is local and random. One house soaked while the other stays dry. The 84% misleads; the 25% ( the cloud, the part that actually erases megawatt-hours and revenue) is nearly independent between distant sites. That is the diversifiable risk, and the market is pricing off the wrong number.

Figure 3 is the heart of this discussion. On the left, each bar is the year-by-year correlation of the two sites’ clear-sky index (the cloud component alone) against the dashed 84% of raw output; the bars sit far beneath it, every year. On the right, each dot is a day: if the clouds moved together the dots would hug the diagonal; instead, they fill the square, signalling independence.

Figure 4 unpacks those bars into the daily data behind them: each site’s clear-sky index, day by day (faint), with a thirty-day average in bold (1.0 is cloudless). The bold lines largely evolve independently. When one site sinks into a cloudy spell, the other often holds up, and that lack of persistent co-movement is what the low correlations measure.

Why it is not just distance, and why that needs a model

The instinct is that decorrelation is just distance: spread the assets north and south and the weather impact diversifies. The data disagree. The two most separated assets in our Spanish national portfolio: 393 MW at Mula (Mediterranean south-east) and 114 MW at Garnacha (Atlantic-continental north-west), 517 km apart (Source: ENTSO-E 14.1.B; Global Energy Monitor) had a cloud correlation of 17% – 31% every year.

Yet a farther pair, at Garnacha and Puerto Real, 557 km away, correlated more, at 36% (Source: NASA POWER, 2023–2024). More distance, more correlation. Decorrelation is a function not of kilometres but of climate regime: the two sit in genuinely different weather systems, one Mediterranean, one Atlantic-continental.

Spread on a map is not spread across climate, and which regimes are truly independent is not something you read off a ruler.

Spread on a map is not the same as spread across climate.

Nor is it a backward-looking exercise. Solar assets are multi decade commitments; and must be sited for the climate towards 2040, not the climate up to 2020. Historical irradiance is only half the question. The forward-looking half, how cloud regimes and their correlations shift under a changing climate, is the half most siting decisions skip, and the one a valuation with a thirty-year tail cannot afford to ignore.

Figure 5 shows where this comes from: the solar energy reaching each site each day, with a thirty-day mean in bold. The seasonal waves (high in summer, low in winter) lie almost on top of each other: the 99% shared shape the sites cannot diversify away. The faint scatter below the waves is cloud cutting output, and it is the scatter, not the waves, the two sites experience independently.

Figure 6 averages that resource by season, and a subtler reward appears. The two sites get almost identical sun in summer (about 7.2 kWh/m²/day each) but the northern site is markedly weaker in winter (2.2 versus 2.8), so its annual resource is lower (Source: NASA POWER, 2021–2025).

Latitude trades winter for summer. Location sets not just the level of output but its shape across the year. For an owner matching generation to a forward price curve, the difference between a flat earner and a seasonal one can change the value of the asset.

Siting reduces the risk, leverage magnifies what is left

Good siting buys a great deal, but this is not enough. Diversifying across climate regimes lowers the variance of output but cannot drive it to zero. Some days a single weather system dims half a country; every winter brings a trough; cloud still strips roughly 16% of clear-sky potential off the southern site and 20% off the northern one, year in, year out (Source: NASA POWER, 2021–2025). After the best siting the geography allows, an irreducible residual of climate risk remains.

And these assets are rarely owned outright. They are financed with long-dated leverage. Project debt measured in decades, underwritten against a cash-flow projection. Leverage does to a revenue shortfall what it does to everything: it magnifies it.

A small, persistent, climate-driven adjustment to expected cashflows an unlevered owner could absorb becomes, in a two-decade leveraged structure, a breach of the projection the debt was sized against. This is not our principle alone; it is exactly what the ECB is encoding by scaling its climate factor with the square root of maturity. The longer the financing, the larger the adjustment, whether the regulator is pricing transition risk or the market is pricing the climate risk.

The reckoning: where this lands

Lift back to the institution and the residual stops being a meteorological curiosity: it becomes a number on someone’s books. Climate risk reaches those books by two routes: climate volatility impairs the cashflows an asset produces, and severe hazards can further impair the value of the asset itself.

Every hazard, on every asset, for every institution, resolves into some mix of the two. For a solar farm: cloud cuts the year’s generation, and the discounted value of that generation shortfall, revalued forward, marks it down. For an industrial site in a flood zone: water disrupts logistics, halts production, impacts revenue and destroys plant and inventory outright. Two distinct losses from one hazard. The routes are universal; only their mixture changes.

For the owners of solar generation, a yield booked against a clear-sky bank case is a return overstated, and when the market reprices that assumption, the carrying value follows the cashflows down. On a solar book, climate variability is a haircut to both halves of the return: the cashflows, and the asset value.

Securitise those cash flows and the risk changes name, not substance.

A solar portfolio’s variable generation is the raw material of an asset-backed bond, and an ABS turns on one blunt question: will the cashflows be there? Variability eats straight into the excess spread: the thin cushion of cash above what is owed to noteholders that shields the senior tranche from first losses. An ABS underwritten on an unclouded profile carries too little of it and is mispriced the day it prints.

For the bank the same risk wears another name. To free up capital, banks pass loan-portfolio risk to investors through significant risk transfer: synthetic deals that cut risk-weighted assets, the regulatory yardstick for how much capital a balance sheet must hold against what it owns. An SRT trade rests entirely on assumptions about collateral and asset value. Exactly what the climate factor, and the climate risk behind it, now impact. Capital relief built on a climate-blind asset value is relief on a number about to change.

And with securitisation, risk concentrates. Bundle a hundred solar farms or data centres, carve the debt into tranches, and losses do not fall evenly. They fall first, and in full, on the junior, first-loss tranche.

Even a marginal, climate-driven shortfall in the pool is absorbed there before it touches anyone senior. The structure assumes the underlying assets are diversified but, recall the 25% solar correlation: a single heat wave, or one cloudy regime, does not respect a portfolio boundary. When a single hazard impacts a significant portion of a portfolio, the diversification the tranching was priced on collapses, and the correlated losses arrive together, on the category of holders least able to see them coming.

It is an uncomfortable echo of 2007, mispriced first-loss paper distributed to investors who do not fully grasp the exposure they hold. One hot summer can wipe a junior coupon whether or not regulatory policy like the climate factor ever widens to physical risk.

The tenors make this even more acute. Operators pricing thirty-year power deals for data centres note that two basis points, compounded and discounted over three decades, is an enormous sum (Source: Data Centre Dynamics). On structures this long, a small, persistent error in the cash-flow projection does not stay small. Leverage, duration and discounting stack three multipliers on the one input nobody measured: climate.

Every ABS cash flow, every collateral haircut, every SRT trade on a physical asset now carries a climate exposure someone, finally, has to price.

The institutions now obliged to price this need more than awareness; they need a model they can put in front of a regulator, auditable, asset by asset, in a similar approach to credit and liquidity adjustments post 2008. Measuring the physical climate risk implied against a real asset, is what turns a cloud-cover series into an excess-spread haircut, a collateral mark, a capital number.

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