In 2024, I visited the Norwegian Embassy in Copenhagen with a banker, a senior vice president in DNB’s energy division, in conjunction with a cross-border carbon transport and storage initiative. With his background in petroleum engineering, rather than purely finance, he has an ability to connect details that few others notice, and that is probably why he ended up responsible for evaluating project financing for the group’s new international energy investments.

On our trip, he mentioned something to me that I didn’t really think about until the sudden push to reignite European industrial capacity in light of the current critical minerals embargoes. He described a sort of hidden pathology in the way large energy infrastructure gets financed: a risk framework written for one purpose, applied to another it was never calibrated for, making the infrastructure Europe has decided it needs more expensive to build. His example was the Mijnbouwwet, the Dutch mining act.

The Mijnbouwwet was written to govern onshore mining; subsidence, seismicity, groundwater in a densely populated river delta. Since a 2012 amendment it also governs the permanent storage of CO₂ in depleted offshore gas fields, where the physical risk is lower by orders of magnitude. The liability standards did not change to match. So a banker pricing a loan against an offshore storage project is pricing a category of risk that the law still treats as if it were onshore mining, and so the cost of that mismatch ends up being reflected as a higher interest rate.

That was his observation, and our conversation is the heart of this essay. What struck me afterwards is that the same project carries a second version of the same problem from a different angle; in the banking capital rules themselves. Those rules are called Basel IV, and they were designed in part to prevent another 2008 by standardising risk weights across banks. The question I want to ask is whether a framework built to make banks safer can, applied to strategic infrastructure, quietly raise the cost of the very projects Europe is trying to finance. The two frameworks were written by different people for different reasons, neither with the other in mind. Yet, both frameworks affect project balance sheets in much the same way.

Mispriced risk

Legally, banks have to maintain enough reserve capital to absorb losses when loans default, even large ones. Exactly how much capital is required gets determined internationally in Basel, Switzerland, by the Basel Committee on Banking Supervision (BCBS), ¹ which has operated under the Bank for International Settlements (BIS) ² since 1973. The standards it sets are not legally binding in themselves, but national regulators implement them in their binding form; in the EU through the Capital Requirements Regulation. ³

Basel IV and the output floor

The current regulatory standards come from dealing with very different global economic situations over time. In 1988, Basel I introduced a simple risk floor where banks had to hold capital equal to a minimum of 8% of their risk-weighted assets (including interest bearing loans like your mortgage). Then in 2004, Basel II allowed banks with sufficient in-house modelling skills to use their own internal models to evaluate credit risk and calculate how much capital any given loan required on an individual basis to be rated within safety limits. It seems logical that a local bank with detailed knowledge about its borrowers can price risk more accurately than a regulator working from a broad categorical perspective. At least, it allowed for lower capital requirements, especially for large infrastructure projects.

When the 2008 financial crisis hit, banks had been using their internal models for about four years to justify holding relatively small amounts of capital against various financial instruments (like the infamous mortgage-backed securities and collateralised debt obligations) that turned out to be far riskier than modelled. Part of the problem was that models were not designed to anticipate how correlated failures might be, so when the housing market failed, everything began failing at once. ⁴ The diversification benefits the models assumed did not hold, and Basel III followed immediately after the crisis, raising reserve capital requirements across the board. Beyond capital levels being too low, there was a second problem with the internal models producing risk scores that were too variable to supervise or compare across institutions.

Now, Basel IV introduces an output floor that still lets banks use their own internal models, but bars those models from producing a capital requirement below 72.5% of what the sector’s standardised models would require (European Parliament and Council of the European Union_{2024, art. 122a}). The floor came partially into force at 50% at the start of 2025 and rises to 72.5% by 2030. For larger corporations with revenues over €500 million, the option to use internal models is removed entirely; those exposures must use the standardised approach regardless of how well the bank knows the borrower. So any unrated company or project of that scale attracts a 100% risk weight, meaning the bank must hold capital proportional to the full loan value, irrespective of the borrower’s actual credit history. For project finance covering infrastructure, energy storage, and large-scale industrial investment, the standardised weights run as high as 130% before a project is operational, falling to 100% once it is running (European Parliament and Council of the European Union_{2024, art. 122a}).

Standardising risk scores across borders makes a lot of sense, but better comparability comes at a cost. To compare risk across institutions you need a common unit, and the unit the standardised approach uses is categorical, not historical. So, in short, a project with stable cash flows, a long operational life, and a government-backed revenue floor will get assessed in exactly the same way as a speculative project of the same size. This is workable in most cases, but it becomes overly expensive when a project carries a certain kind of liability the category was not designed for; which is precisely what my new banker friend was on about.

Open-ended liability of the Dutch mining act

The Dutch Mining Act, the Mijnbouwwet, was passed in 2003, more than a decade before the BCBS began finalising Basel IV. It was also written by someone thinking about geology, seismicity, and environmental liabilities rather than capital ratios. Its job is to regulate subsurface activities in the Netherlands like oil and gas extraction, salt mining, geothermal energy, and since a 2012 amendment, the permanent storage of carbon dioxide (Rijksoverheid₂₀₀₂).

Storing carbon dioxide in offshore wells is, in physical terms, much safer than onshore mining. For instance, depleted North Sea gas fields are very stable geological formations with proven containment records; industry analysis expects over 99.99% of injected carbon dioxide to remain in the subsurface, putting the probability of a significant leak below 0.01% or less than one project in ten thousand (Zero Emissions Platform_{2019, pp. 51}). But the mining act does not apply different standards to offshore CO₂ storage. Instead, it applies the same conservative framework developed for onshore mining (stricter scrutiny on soil movement, seismicity, groundwater effects) to a context within the subsurface category where the risks are materially different.

Getting an offshore CO₂ storage licence is hard, but lenders only care about insuring against the liability they are financing. Furthermore, the borrower’s obligations do not end when injection stops and the well is closed, and that is a price that cannot be pinned to a number when the loan is signed, and it can be expanded by the regulator at any point if the agreed upon subsurface science changes. That combination of open-ended, unbounded, and mutability is structurally different from almost anything else a bank finances, and it makes it very hard to price risk given the following Mijnbouwwet provisions:

The minister must withdraw or amend a licence if there’s leakage, non-compliance, or “new scientific findings and technological progression”. There’s no time limit on it. A site can be performing well and still have its licence changed because the science moved on. That’s sensible environmental policy. It’s very hard to price as a loan (Rijksoverheid_{2002, art. 31h}).

The licence can’t pass to the state until at least 20 years after injection stops, and only once the operator has pre-funded the next 30 years of monitoring (Rijksoverheid_{2002, art. 31j}).

Even after handover, the previous licence holder can still be liable for costs the state runs into later. So the liability tail runs to 50 years or more, and nobody can put a number on it when the loan is signed (Rijksoverheid_{2002, art. 31k}).

The act also maintains what is known as an ALARP ⁵ standard which means that risk must be reduced to a level where further reduction would be considerably disproportionate relative to the benefit. Note, this is not a fixed standard because as technology and science advance, what was considered acceptable in 2025 may no longer be acceptable in 2035.

You might be surprised when I say this is good law. I’m not going to argue it’s bad because for onshore mining in one of Europe’s most densely populated countries, conservative long-tail liability provisions are entirely defensible. The problem isn’t the mining act on its own terms. It’s that the same miscalibration appears from the other direction in Basel IV where a sound risk model is doing a job in a context it was not designed for.

Regulation out of step

Regulation is almost never in total unison with the thing it tries to govern. Sometimes it lags, like when a law written for one activity gets stretched over a newer one it was never drafted for, the way the mining act was stretched from onshore mining to offshore storage. Sometimes it runs the other way, like when a framework arrives with a fixed, forward looking template and imposes it regardless of the specific case, just like the way Basel IV’s output floor does by design. One is regulation catching up, and the other is regulation deciding in advance. These are different types of systemic failures, but they both lead to risk scores being disconnected from the material risks a project actually carries. A CCS hub and storage well in the Netherlands is unlucky enough to sit where both apply at once, and neither framework is built to notice the other. This is also partly why CO₂ storage in the Netherlands is more expensive than in other European countries, and will likely remain so for the foreseeable future.

To a bank, a CCS hub usually looks like an unrated SPV (a special-purpose vehicle, meaning a company set up for the project with no credit history of its own). Under the standardised approach that puts it in the 100% risk-weight bucket, or 130% before it starts operating. So Basel IV sees an unrated project company and assigns maximum capital, whatever its cash flows or revenue support look like. Meanwhile, the mining act has already made the operator put up substantial financial securities (insurance, parent-company guarantees, monitoring funds) against the liability tail. You’d think that would count for something with the bank. It doesn’t. The CRR only counts protection that runs to the lender and can be claimed if the borrower defaults, and only from a closed list of eligible collateral (European Parliament and Council of the European Union_{2024, arts. 194, 197}). The mining act security fails on both counts. It’s held for the State, so the bank can’t touch it if the borrower defaults. And it’s there to pay for cleaning up a CO₂ leak, not to repay a loan. Wrong pocket, wrong purpose. As far as Basel IV is concerned, the security might as well not exist, and the bank holds full capital anyway.

To be fair, there’s a decent counterargument. When the European Banking Authority modelled the full impact of these rules, it found the overall effect on EU bank capital to be modest: tier 1 requirements rise by under 8% at full implementation, which banks can raise comfortably over the phase-in. CRR3 even has an infrastructure supporting factor that lowers the risk weight on qualifying infrastructure (European Banking Authority_{2024, pp. 9}). So where’s the problem? It’s in the average. The output floor bites hardest where internal models used to give the lowest weights, which means the safe, long-dated, well-modelled loans this essay is about. A retail bank barely notices. A bank lending to an unrated infrastructure SPV feels it sharply. The supporting factor helps, but it doesn’t reach past the €500 million revenue cliff, where internal models are barred anyway. A small number for the whole system is what you’d expect when a real cost lands on a narrow slice of it.

So the bank holds maximum capital against a project that the law has already forced to provision for its own worst case. One framework reads the project through a law older than it, the other through a category built to ignore it. Because both apply at once, the gap between priced risk and real risk gets wider, and somebody has to pay for it.

Risk migrates

When financing costs more than a project’s economics can carry, the risk doesn’t disappear. It usually just migrates or gets pushed onto someone else.

Porthos and Aramis

In the Netherlands you can see this in how CCS infrastructure actually gets built. Porthos, the first large offshore CCS hub in the EU, only reached a final investment decision with a €102 million EU grant, SDE++ subsidies covering the gap between storage costs and the EU carbon price, and state support through the permitting delays. When the Dutch Court of Audit reviewed the project in 2024, it found the economics depended on continued public funding, with the government committed to grants of up to €2.1 billion (Netherlands Court of Audit_{2024, pp. 5}). The security the mining act demands is part of that bill. Industry analysis puts the probability-weighted risk cost of a North Sea storage project below €1 million. Operators still have to set aside roughly €589 million against the worst case (Zero Emissions Platform_{2019, pp. 35}). The subsidies end up covering the difference.

Aramis, the larger follow-on hub, is a sharper example. Shell and TotalEnergies, both investment-grade with large balance sheets and every commercial reason to decarbonise, scaled back after the first phase. In April 2025 the Dutch government put in roughly €640 million to keep the project going. I should be careful here. Oil majors pull back for lots of reasons, like the carbon price, uncertain demand, or somewhere else they’d rather spend the capital, and I can’t prove the mining act and Basel IV are why. But when two of the largest energy companies in the world don’t like the terms on a project the state calls strategic, I think it tells you something about how the wider lending market is pricing it.

Danish tender failure

Denmark went a different way. The 2025 CCS tenders asked bidders to take on the long-term, whole-chain risk of capture, transport, and storage, with delivery deadlines and penalties if storage fell short. In practice that pushed post-closure liability onto public waste-to-energy companies. Municipal utilities aren’t built to carry long-tail geological risk, and their backstop is the taxpayer. Most wouldn’t take it. Of the ten prequalified bidders for the 28.7 billion kroner pool, eight didn’t submit a final bid. A KPMG review of the withdrawals found that “more than half of the actors were not willing to take on the tender’s contractual obligations given the overall risk picture”, with risk allocation and the penalty regime among the main barriers (Eskerod_{2026, pp. 6}). Normally either the state carries the risk directly, or a private company carries it and gets paid for it. The Danish tender created a third arrangement: quasi-public companies holding liabilities they have neither the mandate nor the capital for. Mostly, they walked away.

So, in the Netherlands the cost landed on subsidies while in Denmark it landed on municipal utilities, although they ultimately declined taking that risk—and that’s really to be expected.

Blind rulesets

Strategic risk, invisible to credit

Basel IV was calibrated for the risks of 2008: correlated failures in a tightly connected banking system. Europe’s risks look different now. The closure of the Strait of Hormuz and the critical minerals embargoes aren’t credit events, so they don’t show up in any risk weight. A bank lending to a Dutch CCS project, one that lets European heavy industry decarbonise without moving abroad, can face higher capital requirements than a bank lending to an unrated importer of Chinese industrial inputs. In credit terms the two loans may genuinely be similar. That’s exactly the issue: the standardised approach only measures the chance of default, and it can’t see strategic dependency at all. The US has partly worked around this. Inflation Reduction Act tax credits, like §45Q for carbon storage, give qualifying projects a revenue floor that banks can underwrite (Internal Revenue Service_{2022, §13801}). European support like the Dutch SDE++ is valuable, but it’s harder to count as collateral, and Basel IV doesn’t tell the two apart.

Development-finance objections

The natural objection is that development finance exists for exactly this. The European Investment Bank, KfW, and national promotional banks have long financed complex infrastructure that commercial banks can’t price efficiently. If Basel IV pushes commercial banks away from CCS and industrial infrastructure, maybe the answer is simply more development finance. This argument has real force, and the division of labour (commercial banks for standard credit, development banks for complex long-dated infrastructure) is broadly sensible.

The difficulty is scale. The Clean Industrial Deal, the Net-Zero Industry Act’s 50-million-tonne CO₂ storage target for 2030, the Critical Raw Materials Act’s processing ambitions (European Commission₂₀₂₅): these need more financing than development bank balance sheets alone can supply. A market for strategic infrastructure that depends entirely on public institutions is itself a concentration risk, with political constraints, sovereign borrowing limits, and none of the discipline that comes from private capital being at stake. Public finance clearly has a role. The question is whether Basel IV is pushing the balance further toward public dependence than either the banking regulators or the industrial policymakers ever intended.

Strategic autonomy

The banker was not making a political argument during our trip to Copenhagen. He was describing a technical friction from his own work where a risk model carefully designed to measure one thing, but applied to something else, produces financing costs that nobody explicitly intended. Basel IV has the same pathology, reaching the same projects from the banking side. Both frameworks were designed by people trying to be responsible. Neither was designed with the other in mind, and neither was calibrated for a moment when Europe’s industrial capacity has become a question of strategic autonomy.

Risk frameworks are a bit like maps. A map only ever shows what it was drawn to show, and leaves off the rest—sometimes because the terrain had not appeared yet when it was made, sometimes because it was drawn at too coarse a scale to register what is actually there. The Mining Act is the first kind of map; Basel IV’s output floor is the second. Either way, what the map omits does not stop existing. The cost of that gap is not abstract. It shows up in state subsidies for projects that should be commercially viable, in public utilities holding liability they were not built to carry, and in CCS hubs that reach construction only after governments step in to make the numbers work. None of it appears in the Basel IV output floor—a risk that has been standardised out of sight, and out of mind, just when it matters most for future European industrial capacity.