Energy infrastructure returns are often eroded long before the losses show up in the accounts. Ørsted’s Sunrise Wind project shows how. In January 2025, the company announced a $1.7 billion impairment for the fourth quarter of 2024. About $600 million of it was tied to delays and higher costs at Sunrise Wind, and the rest to higher U.S. interest rates and lower seabed lease values. [1] By then, the project’s economics had already shifted.

This is the hidden cost of late-stage risk. When a delay finally surfaces as an impairment, cost overrun, or return compression, the underlying risks have often been accumulating for months.

This is not an unusual story. A 2022 McKinsey analysis of more than 500 large capital projects found that cost overruns averaged at least 79% relative to initial budget estimates, while schedule delays averaged 52%. McKinsey reads the gap as pointing to poor cost and schedule estimates at project approval, not only to poor delivery. [2] Oxford professor emeritus Bent Flyvbjerg analyzed more than 16,000 projects and found that 91.5% ran over budget, over schedule, or both. [3] In energy infrastructure specifically, a 2025 study looked at 662 energy projects in 83 countries, representing $1.36 trillion in investment. [4] More than 60% experienced cost overruns, and construction ran about 40% longer than planned on average. [4] Nuclear projects averaged 102.5% above budget. [4]

These outcomes expose a broader weakness in how energy infrastructure risk is assessed before capital is committed. Investment decisions are often made during diligence, but the analysis behind them tends to rely on backward-looking signals: comparable projects, feasibility studies, and models built around existing assumptions.

The risks that ultimately reshape a project often emerge gradually and outside that initial framework. Interconnection timelines can shift, equipment deliveries can be delayed, and permitting milestones can move, with each change adding pressure to the schedule and budget. Any one of them may look manageable on its own. Together, they can raise costs and cut into the returns the project was meant to deliver. By the time the damage reaches the budget or the schedule, the investment thesis may already rest on assumptions that no longer hold.

How Risk Is Priced at Diligence Today

The standard diligence process for large energy infrastructure investments draws on three categories of information:

1. Past project performance from comparable assets

This means looking at similar projects that have already been built and asking: How long did they take, how much did they overrun, what delays happened, and what risks showed up? For example, a team may look at previous solar, LNG, or transmission projects in the same region to estimate likely cost and schedule risk.

2. Static feasibility models built at a point in time

This is the financial and technical model created during diligence. It may include expected construction cost, interconnection wait time, transformer lead time, permitting timeline, financing cost, internal rate of return (IRR), and contingency. The problem is that it is usually a snapshot. It reflects what the project looked like when the model was built, but may not update as queue positions, equipment lead times, or financing conditions change.

3. Desktop permitting and regulatory reviews

This means reviewing permits, approvals, land use rules, environmental requirements, and regulatory obligations from documents, rather than from live project movement. It can tell you what approvals are needed and whether they appear achievable. It may not show whether one permit depends on another, whether agencies are delayed, or whether the sequence of approvals could compound schedule risk.

All three are backward-looking by design. They tell you what similar projects did. They do not tell you what this project's queue position will look like at financial close, or where its transformer lead time will sit in 18 months, or whether the regulatory sequence has a dependency that no one has mapped.

The issue is not in the diligence effort itself, as investment teams are typically thorough. Rather, the signals most relevant to project finance risk often fall outside the information sources these teams routinely use. Queue positions shift, lead times change, and permitting sequence risks emerge after the analysis is finished, while the process remains largely anchored to fixed documents, historical comparisons, and point-in-time reviews. A desktop permitting review, for example, can identify the approvals a project needs. What it may miss is how those approvals depend on one another, and how a delay in one agency’s review can slow the next step in the sequence.

A static feasibility model has a similar limitation. It captures the project’s assumptions at the time the model is built: the queue position, expected interconnection wait, equipment lead times, financing costs, and contingency. But those inputs can change. If the queue shifts three months later, or transformer lead times extend, the model will not reflect that change unless the assumption is updated.

The result is a systematic mispricing of project finance risk. It arises at the point in the lifecycle when the risk is most consequential and least expensive to address.

Where the Cost Actually Accumulates

The research points the same way. McKinsey notes that decisions get locked in as a project advances, reducing the freedom to make changes. [2] The firm also reports that strong preconstruction work repeatedly lifts a project’s net present value by an amount equal to 20% or more of its total installed cost. [2] In energy infrastructure, cost also does not rise in step with delay. The 2025 study of 662 energy projects found that the rate of cost escalation increased significantly once construction ran more than 87.5% over its planned schedule. [4] The mechanism is well understood, even if diligence models underuse the data. The conditions that generate a capital project cost overrun are often present and measurable months before they produce a schedule event. The schedule event is the moment of visibility, not the moment of creation.

Consider how this plays out in the U.S. interconnection queue, which held more than 2,060 GW of generation and storage capacity at the end of 2025. [5] In the regions with available data, the typical project built in 2025 took more than five years from interconnection request to commercial operation, and the median wait just to sign an interconnection agreement was well over three years. [5] Even a signed interconnection agreement does not fix the timeline. Of the capacity that signed an agreement between 2000 and 2022, 41% had withdrawn by the end of 2025, and late withdrawals can force other projects to be restudied. [5] Now take an illustrative case. A project enters diligence with a confirmed queue position and a projected 30-month wait, and reaches financial close 12 months later. In that interval, withdrawals ahead of it force restudies, and the wait stretches to 44 months. Yet the project’s schedule, financing structure, and IRR model were all built on 30 months. None of them updates automatically, so the model stays wrong until someone notices.

Acting on a risk signal when it first appears is usually cheaper than waiting until it becomes a schedule event. Capital is not yet fully committed, and options are still open.

At the project finance level, this translates directly. In the case above, a 30-month wait priced into the model at financial close turns out to be 44 months at construction start. That does not produce a proportional IRR adjustment, but a non-linear one. This is because debt service continues, equity is exposed for longer, and the construction window has compressed against a fixed set of contractor commitments. The IRR impact of a 14-month schedule extension on a $2 billion project is not a rounding error. It can be a material capital loss that was often visible in the queue data before financial close.

What Changes When Risk Intelligence Enters the Investment Stack Earlier

The value of earlier risk intelligence lies in giving capital allocators better visibility into emerging risks and the ability to act before capital is committed. With six months of visibility into interconnection delays, procurement lead times, and permitting sequence risk, the investment case changes. Contingency can be sized against current market data rather than historical norms. Transformer lead times, now beyond 24 months in most major markets, can be reflected in the schedule before they become a construction constraint. [6] Equity returns can be priced against a range of schedule outcomes, not a single point estimate. Covenants can be tied to queue position confirmation, rather than permit issuance alone.

Earlier visibility can change the investment decision itself. A development finance institution (DFI) investment director might spot a deteriorating interconnection position six months before financial close. At that point, there is still room to adjust the strategy, renegotiate terms, or reconsider the investment. Discovering the same issue six months after financial close turns a manageable risk into a capital problem.

The information needed to make earlier decisions already exists. The underlying signals sit across the project lifecycle in interconnection developments, supply chain constraints, and permitting progress. The challenge is connecting them early enough to guide investment decisions before problems emerge.

Lessons from Ørsted’s Sunrise Wind Project

The Ørsted example shows how late-stage risks reshape a project before they reach the accounts. The company's announcement named three drivers. One was a 75-basis-point rise in its cost of capital during the fourth quarter of 2024. Another was a revaluation of seabed leases. The third was Sunrise Wind itself. Drawing on lessons from turbine installation at Revolution Wind, Ørsted reassessed its schedule and costs, with monopile foundations a major source of higher costs. [1]

A rising cost of capital, delays installing turbines on a sister project, and higher costs for a long-lead component may each look manageable on their own. Together, however, they can alter the construction timeline, extend the financing period, and weaken expected returns. In a leveraged infrastructure investment, small shifts can materially change the return profile on which the investment decision was based.

The warning signs behind late-stage losses are rarely hidden. They emerge across the project lifecycle in financing markets, procurement and fabrication, execution on comparable projects, interconnection activity, permitting records, and contractor commitments. The challenge is that these signals are often reviewed separately and only after the impact is visible.

The opportunity for AI-driven risk intelligence is to change that timing: to connect fragmented signals early enough to influence investment decisions, rather than explain losses after they occur. That is where better risk visibility creates value.