Front-of-meter vs behind-the-meter

Both arrangements can deliver the same megawatts to the same building. What differs is where the meter sits relative to the generation, and therefore what the facility is actually buying and from whom.

The revenue meter looks like a piece of equipment and functions as a legal perimeter. Everything on the network side of it is inside a regulated relationship. Everything on the customer side of it is inside a commercial one.

That single fact produces most of the table above.

On the network side, the facility is a purchaser. It takes delivered electricity under a tariff or a supply contract, and the party delivering it carries an obligation to serve that is defined and supervised outside the transaction. Reliability is largely invisible to the facility, because somebody else is contractually and structurally responsible for it. What the facility owns is a right to buy, and rights to buy are ordinary contractual assets: they can be assigned, they can be secured, and a lender can read them in an afternoon.

On the customer side, the facility is an operator. The generation's only customer is the load in front of it. There is no market to sell into, no tariff standing behind performance, and no counterparty whose obligation to serve absorbs a failure. Availability is now something the project produces rather than something it buys, and it is produced out of fuel that has to be contracted, equipment that has to be maintained, permits that constrain how it may be run, and people who have to run it. Each of those is a credit exposure, and the list of what moves is set out in behind-the-meter.

The interconnection question splits along the same line. A front-of-meter generator is connecting to sell; the facility separately connects to buy, and the two positions sit with different parties who may have no relationship at all. Behind the meter, the facility's load position is the only one that matters for normal operation — but a retained backup or standby connection, or the ability to export surplus, is a second position in its own right, and it is the one most often discovered late. Whose name each of these sits in is a structuring decision with consequences worked through in who holds the interconnection position.

And the collateral is a different object. A supply contract is a receivable-like right that a security package handles conventionally. A generation asset is plant, land rights, fuel arrangements and an output agreement, secured as a bundle and financeable on its own terms by capital that will not touch the layers around it. That is the separation the data-center capital stack is built on, and the meter is where it physically shows up.

Front-of-meter is the default for good reasons, and the conditions under which it stays the right answer are testable rather than aesthetic.

The timeline actually works. Where a load position is genuinely established, the upgrade scope is known, and the date it produces sits inside the commercial window, there is little to be gained by building an operating business to solve a problem that is already solved. The whole case for moving the meter is schedule, and where schedule is not the binding constraint the case is weak.

The sponsor is not an operator and does not intend to become one. Running generation is a capability, not a purchase. A developer with no operating team, no fuel desk and no maintenance regime that takes on behind-the-meter generation has acquired a business it cannot staff, and the usual response — contracting the capability from a third party — reintroduces a counterparty and much of the complexity the move was meant to avoid.

The tenant wants a regulated counterparty behind reliability. Some offtakers are explicit about this, and where they are, it is close to dispositive. A supply arrangement standing on an obligation to serve is a different credit story from one standing on a project's own operating record, and a tenant entitled to insist on the first will insist on it.

Fuel is the exposure the project does not want. Behind-the-meter generation converts a power procurement into a fuel position, and a fuel position has to be contracted, transported and matched in tenor to the revenue it supports. A sponsor unwilling to hold that exposure should not create it, and taking delivered power is the clean way not to.

And the site's value is the position, not the plant. Where the exit is a sale of powered land or a powered shell, an established grid position is the asset a buyer is paying for. Building generation instead produces a different asset with a different buyer, and it should be a deliberate choice rather than a consequence of a schedule problem.

Behind-the-meter wins where the grid answer cannot hit the date, and where the project can genuinely carry what it is taking on.

The commercial window closes before the queue does. This is the case that drives nearly all of it. Capacity that arrives after a tenant's window is worth materially less than capacity inside it, and where the interconnection timeline runs past the window, on-site generation is not a preference but the only route to the date. The wider version of that argument — how the date is set commercially and the supply strategy chosen backwards from it — is speed-to-power.

The generation has a defined role with a defined end. Bridging generation that carries the facility until grid supply arrives is a coherent structure, provided the documents say it is a bridge. What breaks is financing a bridge as though it were permanent, or as though its end date were known when it is not. The equipment's redeployability is what decides whether that risk is bearable, and it is the same test applied to any asset whose role can end early.

The project can hold the fuel and the operations. Firm supply and firm transport, an operating regime, spares, and permits read for what they allow rather than what they mention. Where those exist, availability is a manageable covenant. Where they are assumed, the structure has an unpriced exposure at its base, and it surfaces at the worst moment.

Generation is a layer the sponsor wants to fund separately. Plant with a twenty-year life is financeable by equipment and infrastructure capital that has no interest in the compute above it or the queue beneath it. Putting it on the customer side of the meter makes it a distinct asset with its own tenor and its own residual, which is frequently the point rather than a side effect. The ownership models, and what each does to the security package, are in financing on-site generation.

Or the connection the site can get is not the one the tenant needs. Where an available position is interruptible and the tenancy is not, on-site capacity standing behind the interruption right is one of the few ways to reconcile them — the repricing mechanism set out in curtailment.

The fork is worth settling early because it constrains several decisions that are usually taken later by people who assume it was already made.

It decides who the utility's customer is, and therefore who can lose the position. A load position held by the facility owner behaves differently on a sale, on a change of control, and on a default from one held by a landowner or a sponsor holding company. That is the entity question, and it is answered once rather than discovered per transaction.

It decides what tenor each layer can carry. Delivered power under a supply contract has the tenor of that contract. On-site generation has the tenor of the equipment and of the arrangement selling its output, which is rarely the same number. A structure that funds a twenty-year plant against a shorter revenue commitment has created the mismatch the whole capital-stack argument exists to prevent.

It decides what a completion test can even measure. Where the facility generates its own power, availability at commercial operation is a demonstration of an operating business rather than a confirmation that a connection is live. Acceptance criteria written for a grid-fed building do not test the thing that can actually fail.

And it decides what happens to the plant when its role ends. Bridging generation reaching the end of its purpose is either redeployable, resaleable, re-contractable, or stranded. That question has one answer at the moment the equipment is specified and a different one afterwards.

The practical discipline is the same as on every other fork on this vertical: decide it against evidence rather than preference. Establish what date the grid route actually produces and what it is conditioned on, price what covering the gap costs including fuel and operations, and confirm what the tenant will accept behind reliability. Those three answers settle it — and the common outcome is not one side of the meter but a deliberate mix, with each layer funded by capital that wants the risk that layer carries.