What does limited recourse actually mean?
That the lenders' claim stops at the project. A limited recourse financing lends to a company formed to own one asset, secured on that company's shares, contracts, accounts and revenues, with no general claim against the sponsor's other businesses if the project fails. The lenders' recovery is the asset and its cash flow, and nothing behind it.
The word doing the work is limited rather than non. True non-recourse is close to a fiction: almost every structure preserves some claim against the sponsor, and the negotiation is about which claims and for how long. A sponsor that hears non-recourse and stops reading has skipped the only part of the term sheet where its own balance sheet is still exposed.
The corollary is what makes the instrument distinctive. Because the lenders cannot reach the parent, they cannot rely on the parent's credit, so they underwrite the contracts instead. The construction contract, the revenue contract or concession, the operating agreement and the insurance package stop being commercial arrangements and become the credit itself. That is the trade in one sentence: the sponsor withholds its covenant, and in exchange the lenders take a view on every counterparty in the chain.
What does a sponsor actually buy with it?
Separation, capacity and partners. Separation is the obvious one: a failed project does not become a claim on the rest of the group, and the sponsor's existing lenders do not see their covenant headroom consumed by an asset that has nothing to do with them.
Capacity is the less obvious and often the larger benefit. An asset with contracted, long-dated cash flow supports far more debt on its own economics than the same asset would attract as an increment of corporate borrowing, because the corporate test is a leverage multiple against consolidated earnings and the project test is a coverage ratio against a defined stream. The sponsor is not borrowing more aggressively. It is being measured on the right number.
Partners come third and matter more than sponsors expect. A limited recourse structure is the only shape in which institutional infrastructure lenders, export credit agencies and development finance institutions can participate at all, because their mandates are written against project risk rather than corporate risk. Choosing the structure is therefore also choosing the lender universe, and for some assets it is the difference between a syndicate and a bilateral.
“The question a sponsor should ask is not whether it could guarantee the debt. It is whether guaranteeing it would buy anything, because on a well-contracted asset the guarantee often buys very little and costs the whole balance sheet.”
What does it cost?
Time, structure and control, in that order. Time first: a limited recourse financing runs a diligence process across every contract in the chain, with lenders' technical, insurance, market and legal advisers each producing a report the credit committee reads. That work has to happen before commitment, not after, because there is no guarantee to fall back on if it turns out badly.
Structure second. The documentation package builds the substitutes for recourse: security over everything, a waterfall that dictates the order in which cash may be applied, reserve accounts funded ahead of distributions, direct agreements letting lenders step into the project's key contracts, and covenants restricting what the project company may do with its own money. Each of those is a separate negotiation, and together they are why the closing set is measured in volumes.
Control third, and it is the cost sponsors underestimate. Distributions are permitted, not assumed: cash leaves the project when the coverage test passes and the reserves are full, and not otherwise. A sponsor used to sweeping surplus cash out of an operating subsidiary is agreeing to stop doing that for the life of the loan. Pricing is the fourth cost and usually the least interesting one, because on a well-structured asset the margin is a smaller number than the sponsors expect and the transaction costs are a larger one.
What recourse survives the ring fence?
More than the label suggests. Three categories routinely survive, and they should be read as the real perimeter of the sponsor's exposure. The first is completion support: a sponsor undertaking, sometimes a full guarantee, that falls away when the project achieves defined completion tests. Until that date the financing is frequently not limited recourse at all in substance, whatever the cover page says.
The second is the equity commitment. Whatever the sponsor has agreed to contribute, and whenever it has agreed to contribute it, is an enforceable obligation supported by a letter of credit or a parent undertaking. Deferred equity is still equity owed, and lenders treat the commitment as a receivable of the project.
The third is the customary carve-out set: fraud, misrepresentation, environmental liability, breach of specified undertakings, and misapplication of project funds. These are usually described as bad-acts recourse, and they are unlimited in amount even where the rest of the structure is not. A sponsor negotiating a limited recourse package should read those three categories first and the margin last, because they are where the balance sheet is genuinely at risk.
| The job | Under a corporate guarantee | Under limited recourse |
|---|---|---|
| Who the lender underwrites | The sponsor group's consolidated credit | The project's contracts and its cash flow |
| What sizes the debt | A leverage multiple against group earnings | A coverage ratio against the project's own stream |
| Where completion risk sits | With the sponsor, by definition | With the contractor, backed by sponsor support until completion tests pass |
| What controls cash | Group treasury policy | The waterfall, the reserve accounts and the distribution test |
| What survives failure | A claim against the whole group | The asset, plus equity commitments and bad-acts carve-outs |
| What the sponsor pays | Balance sheet exposure | Time, adviser cost, and permanent operating constraints |
When is it the wrong instrument?
When the asset cannot carry its own story. Limited recourse works when there is a definable stream, a contract or concession that allocates demand risk, a counterparty a lender will accept, and a construction package with a contractor that can absorb completion risk. Remove any one of those and the structure starts asking the sponsor to fill the gap, which is a guarantee arriving by a longer route and at a higher cost.
It is also wrong where the process economics do not work. The advisory, technical, legal and insurance workstreams cost broadly what they cost regardless of the size of the financing, so the same package that is efficient on a large asset is disproportionate on a small one. A sponsor with an existing corporate facility, spare capacity under it, and an asset that would take months to structure separately is usually better served drawing the facility, and the honest answer at the outset is sometimes that there is no financing to run.
And it is wrong where the sponsor wants the flexibility more than the separation. A project company operates under covenants written by people who will never meet the operating team, and changing course inside that structure requires consent. Sponsors who expect to repurpose, expand or sell parts of an asset opportunistically should price that friction honestly before choosing the instrument, because it is the part that cannot be renegotiated afterward without paying for it.
As of September 2026
Sources: no market figures are quoted in this article. The mechanics described, the security and waterfall package, the completion support that falls away at completion tests, the equity commitment, and the customary bad-acts carve-outs, are standard limited recourse market convention. Structural definitions of concession and public-private arrangements follow the European PPP Expertise Centre, Guide to Public-Private Partnerships (European Investment Bank, 2021), and the APMG International PPP Certification Guide, read 6 September 2026. No gearing ratio, margin, fee or transaction-cost figure appears because no named institution publishes one as a general series, and the article states that rather than supplying a range of its own. Nothing here describes a transaction or engagement.
This position sits within our project finance practice.