Fuel supply and tolling for on-site generation

A generation entity behind a data center's meter is usually assessed on the strength of what it sells: a power services agreement with a defined term, an availability obligation and a counterparty whose credit can be examined. That is the analysis in [financing on-site generation](financing-onsite-generation), and it is the right place to start.

It is only half of the entity's credit. The other half is what the plant has to buy in order to perform, every hour it runs, for the whole life of the contract.

The reason this is a financing question rather than a procurement question is the structure of the obligation. A power services agreement commits the generation entity to make capacity available. That obligation is firm, it carries damages, and it does not abate because fuel was unavailable or because fuel became expensive. So an entity with a firm output obligation and a non-firm input has written an uncovered position into the middle of its own contract, and it has done so in a document that reads, on its face, like a strong revenue contract.

Three consequences follow, and they are what a credit committee is actually looking for.

Fuel arrangements have to run at least as long as the power obligation. A supply contract that expires before the power contract leaves a recontracting event inside the structure's term — at a price nobody can know and, in a constrained market, on availability nobody can guarantee. The entity carries that unless the power contract passes it through.

Availability of fuel is a condition of performing at all. Where the plant cannot obtain fuel, it cannot make capacity available, and the availability regime that was supposed to protect the offtaker turns into damages against the entity the financing depends on.

Fuel cost is the largest variable in the delivered price of power. It determines whether the on-site generation is economic against the alternative the customer would otherwise use, and therefore whether the offtaker keeps wanting the arrangement it signed. A contract that is uneconomic for the buyer for years is a contract under renegotiation pressure, whatever its termination provisions say.

The practical test is simple and is frequently not applied: read the supply-side documents against the power contract, clause for clause, and identify every obligation the entity has promised on the output side that it has not secured on the input side.

The word doing the work here is firm. A firm arrangement obliges the counterparty to deliver a stated quantity and gives the buyer a remedy if it does not. An interruptible arrangement gives the buyer access to whatever is available after firm holders have been served, at the supplier's or the transporter's discretion, generally at a lower cost.

The distinction is not a matter of degree. It is the same distinction [take-or-pay](/learn/take-or-pay) draws on the revenue side, applied to the input side: an obligation is credit, an expectation is a forecast. Interruptible supply is a commercial arrangement that usually works and is unavailable in precisely the conditions that make it matter — extreme weather, system stress, and the correlated events under which the data center most needs its power.

What each arrangement can support in a financing differs accordingly.

Securing fuel and securing its delivery are two different contracts with two different counterparties, and a structure can be perfectly sound on one and empty on the other.

This is the most commonly compressed distinction in the whole analysis. A supply agreement establishes that a seller will make a quantity available at a delivery point. Whether that quantity reaches the plant depends on transport capacity between that point and the site, which is generally contracted separately, allocated by its own rules, and constrained by physical infrastructure the generation entity does not control.

What capital examines:

• Whether transport capacity is held firm, and by whom. Capacity held by the supplier for its own portfolio is not the same as capacity held for this plant. Where the entity holds it directly, it usually carries a demand charge whether or not it flows — a fixed cost that behaves like debt service and should be modelled as one.
• The connection to the site itself. A lateral, an interconnect or a delivery facility has to be built, consented and paid for, and it has its own schedule. Where that schedule runs past the plant's commissioning date, the fuel arrangement and the [bridge to energization](bridge-to-energization) are the same problem.
• Whether the delivery arrangement outlasts the power contract, on the same test applied to supply. Transport contracts and supply contracts frequently have different terms, and the shorter one governs.
• Curtailment and priority rules. In constrained conditions, deliveries are allocated by rules the entity did not write. Where higher-priority classes are served first, an arrangement that is contractually firm can still be operationally limited, and the documents should say what happens then.
• Backup capability. On-site storage or an alternative fuel gives the plant a defined period of independence from delivery. Its value is bounded by storage volume, by replenishment, and — importantly — by whether the operating permit allows the plant to run that way at all. A backup capability the permit does not authorise is not a mitigant.

The question that resolves this section: on the worst day the market has, does the plant have fuel at the gate, and which document says so.

Fuel prices move, they move a great deal, and they move most in the conditions under which the plant is most needed. No structure removes that; every structure decides who holds it. There are four ordinary answers and they are not mutually exclusive.

The entity absorbs it. The generation entity buys fuel at market and sells power at a fixed price. This is the simplest arrangement to document and the hardest to finance, because the entity's margin — the thing the capital is repaid from — is the difference between a fixed number and a volatile one. Sized against an assumed fuel cost, the structure is sized against a forecast.

The contract passes it through. The power price has a fuel component that tracks an index or the entity's actual delivered cost, so the entity's margin is preserved regardless of price. This is the arrangement that finances most easily, and it works only to the extent it is complete: pass-through mechanisms that cover the commodity but not transport, or that cap the passed-through amount, or that reset annually against a market that moves daily, leave a residual position with the entity that nobody has named. The pass-through is also the point at which the data-center offtaker discovers that it has taken commodity exposure, which is a conversation better held during negotiation.

It is hedged. Financial or physical hedges fix the cost of a defined volume over a defined period. Hedging is a real answer with two structural limits: hedge tenors are typically shorter than the power contracts they support, leaving an unhedged tail; and a hedge is itself a contract with a counterparty, collateral requirements and margin calls that consume liquidity at exactly the moments the market is moving.

The offtaker supplies the fuel. That is tolling, and it is the subject of the next section.

Behind all four sits the economic question the structure exists to answer. Fuel is the dominant variable input, so it sets the delivered cost of on-site power, and that cost is what the arrangement is judged against — not in the abstract, but against whatever the customer's alternative is: grid supply where an interconnection is available, or no project where one is not. A [behind-the-meter](/learn/behind-the-meter) arrangement that is competitive because fuel was cheap when it was signed is a different asset from one that is competitive because its structure holds under a range of fuel costs. Capital underwrites the second and is regularly shown the first.

In a tolling arrangement, the party that buys the power supplies the fuel to the plant and pays the generation entity a fee to convert it. The entity never owns the fuel, never buys it and never sells a commodity. It sells a service: taking delivery of fuel provided by the toller and making power available from it.

The structure is familiar in merchant power and, more recently, ubiquitous in battery storage. Its application to gas-fired generation whose whole load is a single data center is much less well described, and that is exactly the configuration this page is about — because in that configuration the toller and the power offtaker are the same party, which changes what the structure is actually doing.

What the fee is composed of. A tolling fee generally has a fixed capacity component, payable for making the plant available whether or not the toller calls on it, and a variable component covering the cost of running when it does. The agreement also fixes a conversion rate — the quantity of fuel the entity may consume for a given quantity of power — and allocates the consequences of missing it. That is a commercial parameter, not a technical one: it is how the parties decide who carries the difference between contracted and actual conversion, and it is negotiated, guaranteed and tested like any other performance term.

What it does to the entity's credit. This is the point of the structure, and it is a genuine transformation rather than a reallocation of labels.

• Commodity price risk leaves entirely. The entity has no fuel cost, so its revenue does not depend on a price it cannot control or forecast. Its cash flow becomes a fee for availability, which is a fundamentally more financeable shape.
• Volume risk largely leaves with it. Where the capacity component is payable regardless of dispatch, the entity is paid for being available, and the analysis becomes the one set out in [take-or-pay](/learn/take-or-pay): an obligation to pay is credit; a right to call is a forecast.
• Fuel supply risk moves to the toller. Procuring fuel, contracting transport, and getting it to the plant become the toller's problem. The entity's obligation is to convert what arrives.
• Counterparty risk becomes the whole position. Everything the structure removed has been replaced by a single exposure: whether one party pays a fee for years. On a behind-the-meter plant, that party is also the only customer, so the concentration that already defined the layer is now total. The credit of the offtaker is not one input to the analysis; it is the analysis.
• Availability becomes the entity's dominant obligation. With price and volume gone, what remains on the entity's side is performing when called. That makes the availability regime, the maintenance arrangements and the outage allowances the operative credit terms, and links this structure directly to the equipment discipline in [equipment finance for turbines and gensets](equipment-finance-for-turbines-and-gensets).

Why a data-center offtaker might want it. A hyperscale or enterprise buyer that already procures energy at scale may be better placed to source and transport fuel than a single-site generation entity, and may prefer to hold the commodity position itself rather than pay someone else's risk premium for it through a pass-through mechanism. It also gives the buyer direct control over the input to its own power supply, which is a different proposition from buying delivered power under a [power purchase agreement](/learn/power-purchase-agreement).

What tolling does not solve. It does not create fuel. The obligation to supply moves, and where the toller cannot deliver, the entity is generally excused from availability and still paid the capacity component — so the risk has been transferred, not eliminated, and the transfer is only as good as the toller's own arrangements. Nor does it change the plant's economics: if the delivered cost of power is uncompetitive, tolling determines who bears that, not whether it is true.

How fuel positions behind a financeable-looking generation entity fail:

• The power obligation is firm and the fuel arrangement is not. The most common error, and the most consequential: an availability commitment with damages attached, standing on interruptible supply or interruptible transport. It works until the conditions in which it was supposed to matter.
• Supply is firm and transport is the gap. The entity holds a solid supply agreement and no firm capacity to move the fuel to the plant, or capacity held by another party for its own purposes. The weakest link in the delivery chain sets the position, and it is rarely the one that was diligenced.
• Tenor mismatch between the fuel and the power contracts. Supply or transport arrangements expiring inside the power contract's term put a repricing and re-availability event in the middle of the structure. The entity carries it unless the power contract passes it through, and pass-through mechanisms are usually written for price movements, not for the supply disappearing.
• The pass-through has holes in it. A mechanism that covers commodity cost but not transport charges, or that is capped, or that resets on a cadence slower than the market moves. The entity retains a fuel position it believes it has passed on, and nobody has priced it.
• Hedges shorter than the obligation. A hedged position for the early years and an open one thereafter, presented as a hedged structure. The unhedged tail is where the financing's later years live.
• Tolling concentrates everything into one counterparty. The structure removes commodity and volume risk and leaves a single fee-payer standing behind the entire cash flow — a party that is also the sole power offtaker and, frequently, the only load the plant can physically serve. There is no diversification anywhere in that position, and it should be underwritten as what it is.
• Permits limit the run profile the fuel arrangement assumed. Operating consents can restrict hours, conditions, or the use of a backup fuel. Where the fuel strategy assumes flexibility the permit does not grant, the permit governs and the contractual obligation is the one with damages attached.
• Fuel was treated as an operating detail. The security package lists the equipment, the site rights and the power contract, and the fuel arrangements appear as a line nobody read against the obligations they are meant to support. Where a successor could take the plant and not the fuel contracts, the collateral is a set of machines that cannot run — the same step-in test applied throughout [the data-center capital stack](/learn/the-data-center-capital-stack).

Continuum structures and arranges around generation and its contracts and coordinates the parties in them. It does not supply, market, transport, trade or hedge fuel, does not own or operate generation, does not sell power, and is not a bank, a broker-dealer or a direct lender; it does not hold client funds. Fuel systems, equipment and generation engineering are covered by Pantheon, not here.