Capitalisation rules decide when software spending stops being an expense and becomes an asset. That is a reporting question. It also creates a standing incentive acting on architecture, and nobody in the room describes it as a technical decision.
The mechanism
Under IFRS, cost accumulation for an internally generated asset begins when the recognition criteria are met, and expenditure already expensed is not reinstated.
The behavioural consequence is direct. Every month a project remains in the research phase is a month permanently expensed. Meeting the criteria earlier moves more of the total onto the balance sheet.
The criteria include demonstrable technical feasibility, intention to complete, ability to use the result, probable future economic benefits, adequate resources, and reliable measurement of attributable cost. Several of those are judgements about how settled the design is.
So the rule rewards a project that can say early: we know this works, we intend to finish it this way, and we can measure what it costs. That is a description of early commitment to a design, which in genuinely exploratory work is the thing good engineering avoids.
Where the pressure lands
Against deliberate disposability. Building something you intend to discard is exploration by construction. It is also expenditure with no recognition, and a team that structures work as throwaway prototyping is choosing the treatment that expenses everything.
Against parallel alternatives. Carrying three candidate designs keeps the feasibility question open by design. Narrowing to one closes it, and the narrowing is what the criteria reward.
Against reopening decisions. Once feasibility has been asserted and costs are accumulating, revisiting the architecture reopens the assertion. The cheapest path is to keep building the thing already declared feasible.
None of this requires anyone to act in bad faith, which is what makes it durable. Each individual judgement is defensible. The direction they collectively point is not one anyone chose.
The US change relocates it
The internal-use software model most teams know distinguishes preliminary, application-development, and post-implementation activities, with eligibility attaching to the middle stage.
ASU 2025-06 removes those prescriptive stages, keying recognition instead to funding authorisation together with probable completion and use as intended, taking significant development uncertainty into account. The amendments are effective for annual periods beginning after 15 December 2027, with early adoption permitted.
That moves the question from what activity is occurring to what has been authorised and decided. The pressure does not disappear; it attaches to the authorisation event and to the assessment of development uncertainty rather than to a stage label.
For the next two reporting cycles it also means two comparable entities can be on different models, one having early-adopted and one not — so a single undated checklist explains neither.
Inference costs have the same shape
A token-priced invoice is a pricing unit, not an accounting conclusion, and separating the components is required before any classification: calls consumed during operation, calls used in development and testing, prepaid access with its expiry and refund terms, separately controlled code, and equipment.
The relevant anchor for hosted services is that access under a service arrangement is a service rather than a customer software asset, that separately controlled code needs its own assessment, and that an advance payment can create a prepayment. A non-refundable payment does not establish intangible-asset recognition.
The incentive here acts on description rather than design: how usage is characterised determines its treatment, and the characterisation is made by people who know which answer is preferred.
The adjacent pressure on procurement
The same dynamic appears at approval thresholds. Whole-life proposal cost is the basis for assessing against a delegated limit, and federal policy prohibits breaking down an aggregated requirement merely to use simplified procedures or avoid requirements above a threshold.
The prohibition exists because the incentive exists. Four orders of 15,000 for an agreed inseparable deliverable total 60,000, and recording each alone conceals the number that matters — while a genuinely optional investigation creating no further commitment is a different case whose follow-on should still be visible.
Near-threshold amounts and year-end timing are prompts for inquiry, not findings. Distinguishing staged learning with a real stop decision from concealment requires the scope, the dependencies, and the chronology.
The rule
What stays fixed is that a reporting boundary applies force wherever a judgement determines which side of it work falls on. What changes is where the boundary sits, and moving it moves the pressure rather than removing it.
The countermeasure is separation. The technical decision about how settled a design is, and the accounting judgement about whether the criteria are met, are made against their own evidence and recorded separately — with the dated activity record preserved so a later reviewer can see which came first.
Not to be confused with
An allegation. No prevalence of budget-driven relabelling is established, and genuine technical progress and a documented policy change are the competing explanations for any observed shift. Claimed milestones get compared against observed work before anything is inferred.
Tax treatment. Financial reporting, budget approval, and research tax relief are three separate analyses on the same spending, with their own definitions and their own incentives.