What is a special-purpose vehicle (SPV)?

A special-purpose vehicle — also called a special-purpose entity, a project company, or a propco — is an entity formed to hold a defined asset or project and to do nothing else. It has one business, one set of contracts, and one set of obligations.

The purpose is ring-fencing, and it works in both directions:

• Outward. The parent's other liabilities, creditors and failures cannot reach the asset or the cash flow it produces. Whoever finances the vehicle is exposed to one project, not to everything else its sponsor is doing.
• Inward. If the project fails, the loss is contained within the vehicle rather than travelling back into the sponsor's balance sheet — subject to the specific recourse the sponsor has agreed to retain.

That containment is what makes [project finance](project-finance) possible at all. A lender cannot look through to a single project's cash flow unless something prevents that cash flow from being consumed by obligations that have nothing to do with the project.

The vehicle is not a clever device. It is ordinary infrastructure, used in essentially every real-asset financing, and the interesting questions are entirely about whether the separation it claims is genuine.

An entity does not become ring-fenced by being described as one. The separation is created by a set of undertakings — usually called separateness covenants — and by conduct that matches them.

• A limited purpose, written down. The constitutional documents and the financing agreements confine the vehicle to owning and operating the specified asset. Anything else requires consent.
• A prohibition on other debt and other obligations. The vehicle may not borrow elsewhere, grant security elsewhere, or guarantee anyone else's obligations. A vehicle that can guarantee its parent's debt is not ring-fenced in any meaningful sense.
• Its own books, accounts and records. Separate bank accounts, separate financial statements, no commingling of funds. This is the covenant most often broken in practice, and it is broken casually, by treasury convenience rather than intent.
• Controlled cash. Revenue flows into accounts governed by the financing documents, with a defined order of payments and conditions that must be met before anything is distributed upward. The controlled-accounts structure is what converts "the asset generates cash" into "the cash reaches the obligation first".
• Independent governance. In stronger structures, a director whose consent is required for insolvency filings or fundamental changes, so the vehicle's fate is not purely a parent decision.
• Non-petition and limited-recourse undertakings. Counterparties agree not to place the vehicle into insolvency and to look only to its assets. This prevents one unrelated creditor collapsing a structure everyone else depends on.
• A genuine transfer of the asset in. The asset and its contracts have to actually be held by the vehicle, cleanly, with no retained claims behind them. An asset still sitting at the parent while the vehicle holds a contractual right to its cash flow is a different and weaker structure.

The test to apply is not whether the documents exist but whether conduct matches them. Separateness that is documented at closing and abandoned in operations is a defence nobody will get to rely on.

The reason SPVs matter so much in this sector is not corporate hygiene. It is that [an AI data center is four assets at one address](the-data-center-capital-stack) with economic lives that differ by an order of magnitude, and the vehicle is the mechanism that lets them be held — and therefore financed — separately.

Separating the layers into distinct vehicles allows land and interconnection to be held by long-dated infrastructure capital, the shell to be financed as a facility, generation equipment to be treated as an energy asset, and compute to be held under equipment or lease structures with a life that matches it. Each layer is then financed by capital that understands it, on a [tenor](tenor) that matches it.

The cost of that separation is that the seams have to be documented. Common ownership was doing that work silently; once the layers sit in different vehicles, every relationship between them becomes a contract that has to stand on its own:

• A lease or easement from the land vehicle to the facility vehicle, long enough to outlast the facility's own financing.
• A power purchase or tolling arrangement between the generation vehicle and the facility, on terms that survive a change of ownership on either side.
• A hosting or colocation agreement between the facility and the compute owner, with access rights that survive a dispute — the point made at length in [collateralizing compute](/compute/collateralizing-compute).

Those intercompany contracts have to be real. Whoever finances one layer will read the contracts that bind it to the layers above and below, and will treat a related-party document written on non-commercial terms as evidence that the separation is presentational. The [financeability gates](/sites/what-makes-a-site-financeable) are, in large part, a test of exactly these seams.

The vehicle is a container. It changes who bears what and what can reach what. It does not improve the asset inside it, and four misconceptions are worth naming.

It does not create credit. A weak cash flow in a well-drafted vehicle is a weak cash flow. Ring-fencing determines who is exposed to it; it does not make it larger or more certain.

It does not survive its own paperwork being ignored. Commingled accounts, shared staff without a services agreement, informal upstreaming of cash, and undocumented intercompany dealing all weaken the separation the structure depends on — and they weaken it precisely when someone is looking for a reason to attack it.

It is not free. Each vehicle carries formation, administration, audit, tax and reporting costs, and every intercompany relationship it creates is a document that has to be negotiated. Separation is worth its cost where it unlocks capital matched to a layer, and not otherwise. Creating vehicles because the structure chart looks more sophisticated with them is a real and common waste.

It does not answer the securities questions. Who may hold interests in a vehicle, how those interests may be offered, and to whom, is a legal question governed by securities law and belongs to counsel. Continuum's role in these structures is to structure the deal and coordinate the parties — establishing what sits in which vehicle, what binds the vehicles to each other, and what each layer needs to be financeable on its own terms.