Sale-leaseback vs refinancing

The asset exists, it works, and capital is tied up in it. Both routes convert some of that back into cash. What they do to the sponsor's position afterwards could hardly be more different.

The routes diverge on one question: does the sponsor still own the asset when the transaction closes?

Under a sale-leaseback the answer is no. The buyer takes title, so it takes what the asset is worth in the future — the whole of it, downside and upside. That is why proceeds can approach full value: the buyer is paying for an asset, not lending against one, and it is compensated through the lease rather than through a margin. It is also why the sponsor's own credit matters less. What is being underwritten is the asset and the strength of the covenant to pay rent on it.

Under a refinancing the answer is yes. The sponsor keeps title, keeps the residual, keeps the upside — and therefore cannot expect to be advanced the whole value, because the lender needs headroom between what it advances and what the asset would fetch. That headroom is the cost of keeping ownership, expressed as proceeds foregone.

The control rows are the ones sponsors underestimate. Both routes impose constraints; they are simply written in different documents. A lease governs how the asset may be used, maintained, insured, modified and moved, and it governs it for the whole term. A loan governs distributions, further borrowing, disposals and covenant tests. Neither is obviously lighter. Which one binds harder depends entirely on what the sponsor intends to do next, and that should be tested against the actual drafts rather than assumed.

The transferability row is where sale-leasebacks most often die, and it dies late. A sale requires that the asset can actually be sold: no blocking security interest, no hosting or facility agreement restricting a change in the equipment's owner, no site arrangement that fails to survive it. A sale-leaseback over an asset a facility agreement will not permit to change hands is not available, however attractive the terms. That check belongs at the start of the process, not in documentation.

Sale-leaseback wins where the sponsor wants out of the ownership position, not just out of the capital commitment.

The residual is a risk the sponsor does not want. This is the strongest case and it is the common one on compute. Equipment bought outright during a supply squeeze frequently sits on a sponsor's balance sheet carrying an obsolescence exposure the sponsor never intended to take. Selling it moves that exposure to a party whose business is holding it.

The sponsor's credit is the constraint. Where corporate capacity is used up or the sponsor is small relative to the asset, borrowing against the asset may not be available on useful terms while a sale of it still is. The buyer is underwriting the equipment and the covenant, not the sponsor's whole business.

Maximum proceeds matter more than upside. A sponsor redeploying capital into a second site is trading a residual it may never realise for capital it can use now. That is a rational trade where the next project earns more than the retained upside is worth.

The layer needs to be separated after the fact. Where compute was bought outright and bundled into a structure built for infrastructure tenors, a sale-leaseback is how the layer is retrofitted into something matched to its life. It is the corrective for a tenor mismatch already created.

It is the wrong answer where the sponsor believes the asset will be worth materially more than the buyer does, where the operating restrictions in the lease would interfere with how the facility actually runs, or — decisively — where the asset cannot be cleanly transferred. On that last point the diligence is documentary and it is quick: identify every existing security interest, read the hosting and site agreements for transfer restrictions and change-of-owner consents, and confirm that access rights survive. The considerations are the same ones that make a security position hard to enforce, catalogued on collateralizing compute.

Refinancing wins where the sponsor wants the capital but not at the price of the asset.

The asset is a long-term hold. Land, interconnection and shell are held for decades by parties who intend to own them. Selling a durable asset to release capital that could have been borrowed against it is an expensive way to solve a timing problem.

The sponsor expects to be right about value. Refinancing keeps the upside. A sponsor with a well-founded view that the asset will appreciate — a site whose power position is becoming scarcer, a facility in a market that is tightening — is selling that view cheaply if it transfers ownership.

Operating freedom matters. A lease term that dictates use, maintenance standards, modification rights and end-of-term condition can be materially more restrictive than a covenant package, particularly on an asset the sponsor intends to reconfigure. Where the plan involves changing how the asset is used, ownership is worth keeping.

The asset has improved since it was funded. A completed and contracted facility is a different credit from the construction project it was, and refinancing is how that improvement is realised. It is the normal end state of a build financed on a construction facility — the point taken up in construction loan vs forward funding.

The sponsor has credit capacity and a reason to use it. Where the corporate balance sheet can carry the borrowing and the project is the sponsor's core business, using it is the simpler transaction and usually the faster one.

Refinancing is the wrong answer where the sponsor has no capacity left, where the asset's value is falling fast enough that a lender will advance little against it, or where the real problem is that the sponsor is holding a residual it cannot price. That last case is worth stating plainly: refinancing releases capital but changes no risk. A sponsor whose difficulty is exposure rather than liquidity has not solved anything by borrowing against the exposure.

Whether the borrowing sits at corporate or project level is the prior question, and it is on corporate credit vs project finance.

Both routes behave differently by layer, and a sponsor holding a full stack usually applies different answers to different parts of it.

Land and interconnection. Long-lived, appreciating in tight markets, and the foundation of everything above. Refinancing is the default; a sale here is a decision to exit the position rather than to release capital from it.

Shell and fit-out. The classic sale-leaseback asset, and the one where the structure has decades of precedent. A completed building with a creditworthy occupancy covenant is exactly what long-dated real-asset capital wants to own. The financeability tests it must pass first are on what makes a site financeable.

Generation and power equipment. Established sale-leaseback territory. Long-lived equipment with predictable behaviour and a functioning secondary market supports either route; the choice is usually driven by whether the sponsor wants to own the generation position or merely to use it.

Compute. Where the residual argument is sharpest, because values fall in steps rather than smoothly. Sale-leaseback is the corrective for equipment bought outright that should never have been held; refinancing against accelerators is possible but the headroom a lender needs is wide, so proceeds are correspondingly thin.

Two sequencing points are worth stating. First, a sale-leaseback is easier before an asset is encumbered than after. Security granted early over an asset intended for later sale is the single most common obstacle, and it is created by sponsors who had not yet decided what they would do. Second, the two are not exclusive over time. Refinance the facility, sale-leaseback the compute inside it, and each layer ends up funded by capital that understands it — which is the separation argument applied to an asset that already exists rather than one being structured from scratch.

The prior question on the compute layer, for a sponsor who has not yet bought anything, is on lease vs own compute. Where the perimeter of the release is also open — one asset or several — the trade-offs are on single-asset vs portfolio financing.