Budget 2027: Govern the Commitment Before You Sign It
Budget season is underway, and this one has a structural feature the last few did not.
A large share of what is being budgeted for 2027 is not a cost you can turn off. Multi-year compute contracts with take-or-pay clauses. Long-term power purchase agreements. Priority-allocation premiums to secure capacity that is genuinely scarce. Enterprise AI licences sold on annual commitments rather than seats consumed.
Each of these converts a variable cost into a fixed one. That is the whole story, and it is worth being blunt about what it means: an assumption you sign in Q4 stops being an assumption. It becomes a floor under several years of P&L, and no amount of reforecasting will move it.
Finance has spent two decades getting good at explaining variances after the fact. The commitments now on the table need a different competence, exercised earlier.
The Practice Almost Nobody Has
Five things are being recommended for the 2027 cycle, and four of them are familiar. Three-scenario budgets with a named owner per scenario. Energy and compute as their own budget lines rather than buried in IT. Rolling reforecast, monthly or quarterly, instead of an annual set-and-forget. Documented business logic behind every assumption.
Good practices, widely discussed, and most teams are somewhere on the path.
The fifth is the one that is genuinely rare: governing the commitment before signature. Not reviewing it, not approving the spend, but running the thing through the model while it is still a draft — with its term, its floor, its escalators and its exit terms — and looking at what it does to the three scenarios you are carrying.
Every other practice on that list operates downstream of a decision. This one operates on the decision itself, which is the only point at which the answer can still change anything.
Why the Model Has to Exist First
Here is where the sequencing bites, and where most teams will fail this quarter without noticing.
A commitment review needs four things at once: the term structure of the contract, the three scenarios you are already carrying, a way to push the contract through each of them, and a comparison against the world in which you do not sign. Assemble that in a fresh spreadsheet under deadline and you will produce something. You will not produce something anyone can audit in February, and you will not produce it fast enough for the decision to wait on it.
The uncomfortable part is that the deadline is not yours. Vendors selling scarce capacity know it is scarce and price the option to wait accordingly. If the analysis takes four days, the commitment gets signed on judgement and the model catches up afterwards. That is not a governance failure anyone will write down, and it is how it actually goes.
So the practical question for the next few weeks is not whether to adopt commitment governance. It is whether the structure that would make it possible exists before the first contract lands on the desk.
What That Structure Needs to Hold
Not much, but specifically these:
Assumptions that are named and owned, separately from the numbers. The take-or-pay floor is not a cost line, it is an assumption with a signatory. When someone asks in June why the compute line cannot flex, the answer should be a named person and a date, not an archaeology exercise.
Scenarios that coexist over one structure. Three living versions of the same logic, comparable line by line. Three workbooks called base, upside and downside are not scenarios; they are three things that will disagree by March.
A dependency chain that propagates. Change the contract start from January to April and the payment schedule, the cash profile, the covenant headroom and the EBITDA bridge should all move, in order, without anyone re-deriving the chain by hand.
A history you can point at. Which version of the assumption was live when the commitment was approved. This is the part PE-backed and audited teams learn the hard way, usually once.
The Reforecast Consequence
There is a second-order effect of all this that is worth naming, because it changes what a reforecast is for.
When most costs were variable, a reforecast was a tracking exercise: actuals came in, you updated, the gap narrowed. When a large share is contractually fixed, the reforecast stops being about accuracy on those lines. You already know the compute cost for 2027. What the reforecast has to tell you is something else: how much of the plan is still steerable, and where.
That is a different output. It is a shrinking pool of discretionary spend, reported against a fixed base, month after month. Teams that keep running the reforecast as a variance report will get very precise about numbers that cannot move, and lose track of the ones that can.
A Reasonable Sequence for the Next Six Weeks
- Separate the committed from the discretionary in the 2027 draft. If you cannot produce that split in an afternoon, that is your finding.
- Put energy and compute on their own lines, whatever their size today. A line item that exists can be governed; one buried in IT overhead cannot.
- Build the three scenarios on one structure, not three files. If comparing them is a project, they are not on the same structure.
- Name an owner per scenario and per material assumption. Not a department, a person.
- Write the commitment test before the first contract arrives. A standing question set: term, floor, escalator, exit cost, and what it does to each scenario. Applied in an hour, not four days.
None of this requires new software. It requires the structure to exist before the pressure does, which is a calendar problem rather than a tooling one.
The Takeaway
The 2027 budget is being built under two forces that do not reconcile: heavy infrastructure spend on one side, investors asking for discipline on the other. The collision shows up as multi-year commitments converting variable costs into fixed ones, and once signed, they are outside the reach of every governance process finance normally runs.
The practice that matters is the one that happens before signature, and it only happens if the model is already standing. Built under deadline, it arrives after the decision and documents it rather than informing it.
Layerz keeps a model as structure separate from its data, so scenarios live over one logic, assumptions carry an owner, and a change propagates instead of being re-derived. Whatever you use, get it standing before the first contract lands.