Corporate credit vs project finance

Both routes finance the same building, the same megawatts and the same equipment. What differs is what the capital is actually looking at, and therefore what can go wrong for whom.

The whole difference follows from one thing: what the lender can reach if the project disappoints.

Under corporate credit the answer is everything the sponsor owns. That is a wide recovery pool, which is why the analysis can be light on the project itself — the project is a use of proceeds, not the credit. It is also why a sponsor with a strong, diversified business can move quickly: the credit work was done at the corporate level, and this transaction is an increment on it.

Under project finance the answer is the project entity and nothing else. The recovery pool is narrow and specific, so every input to it has to be examined: whether the site is controlled for long enough, whether the power position is documented, whether construction can be completed at a fixed cost, and above all whether someone creditworthy is contractually obliged to pay for the output. That is the diligence burden in the table, and it is not bureaucratic drag. It is the direct consequence of removing the sponsor from the recovery analysis.

The control terms follow from the same fact. Where the cash flow is the only security, capital insists on seeing it: the accounts it lands in, the order it is applied in, the conditions under which it may leave the entity. A sponsor experiencing that as interference is misreading it — it is the price of the ring-fence, and the ring-fence is what the sponsor bought.

The last row is the one most often overlooked at signing. Corporate capacity is finite. A sponsor intending to build one facility can spend it freely; a sponsor intending to build four cannot, because the first deal will have priced the second.

Corporate credit wins under conditions that are easy to state and easy to test.

The sponsor is large relative to the asset. Where the project is a modest fraction of an established business, ring-fencing it buys little and costs real time. The sponsor's credit is the better credit, and using it is the cheaper answer.

The project has no offtake yet. Early-stage development, a speculative shell, a site being assembled ahead of demand — none of these has a contracted cash flow to underwrite, so there is nothing for project finance to attach to. Someone has to carry that risk on a balance sheet, and it will be the sponsor's.

Speed is the binding constraint. Where a power position, an equipment allocation or a site option expires on a date, a structure that takes months to negotiate may be the wrong structure regardless of its merits. An existing corporate facility that can be drawn this quarter is worth more than a better-designed one that closes after the option lapses.

The project will be refinanced later anyway. Corporate credit is frequently the bridge, not the destination. Build on balance sheet, reach completion and contracted operation, then move the asset into a project structure once it has the characteristics project finance requires. That sequence is common and deliberate, and it is a different decision from choosing corporate credit permanently.

The conditions under which it is the wrong answer are the mirror image, and the sharpest of them is concentration. A sponsor whose entire business becomes one project has not avoided project risk. It has taken the same risk without the ring-fence — and it has attached the rest of its enterprise to it.

Project finance wins where the project can genuinely stand alone and the sponsor cannot afford to hold it.

The contracted revenue is strong enough to be the credit. This is the gate. A firm, long-dated obligation to pay from a counterparty that survives stress is what makes the structure possible; anything softer is not offtake and will not support it. The tests that separate the two are the same ones applied to any contracted asset, and they are set out on compute offtake as credit.

The asset is large relative to the sponsor. This is the common case in the AI buildout, and it is why so much of the sector is structured this way. A developer whose balance sheet cannot absorb a facility of this size has one route to financing it, and it runs through the project's own cash flow.

The sponsor intends to build more than one. A repeatable per-project structure keeps corporate capacity intact for the next site. The first deal costs more to document; the fourth costs far less, because the template exists and the counterparties are familiar with it.

The equity has partners. Ring-fencing is the natural home for co-investment. A defined entity with defined cash-flow rights is something a partner can underwrite; an undivided interest in a sponsor's balance sheet is not.

It is the wrong answer where the offtake is thin, where the site still has open gates — the five tested on what makes a site financeable — or where the transaction is small enough that documentation cost swamps the benefit of the ring-fence.

One caution specific to this sector: a ring-fence that includes the compute layer inherits its obsolescence profile, which is not what project-finance capital prices. Where the structure is meant to be long-dated, the compute usually should not be inside it.

This is the first decision because it constrains the others, and a sponsor that makes it late usually discovers it has already been made by default.

It sets the perimeter. Whether one asset is financed alone or several together is a live question only once the ring-fence exists — see single-asset vs portfolio financing. Under corporate credit the perimeter is the whole company and there is nothing to choose.

It sets who carries completion risk. A construction facility inside a project entity looks very different from construction funded off a corporate line, and the choice between funding the build and buying it complete is examined in construction loan vs forward funding.

It sets what happens to the compute layer. A project entity built for twenty-year infrastructure is a poor container for a four-year asset. Deciding to ring-fence usually means deciding to hold the compute somewhere else, which is where lease vs own compute starts.

It sets which counterparty is acceptable. Corporate credit can tolerate an offtaker that project finance cannot, because the sponsor stands behind the shortfall. That difference is most of what separates the two cases in neocloud vs hyperscaler offtake.

The practical discipline is to establish the fork before negotiating anything else, and to establish it against evidence rather than preference: read the offtake contract, size the asset against the sponsor, and count how many more projects the sponsor intends to build. Those three answers decide it, and they are all knowable on day one.