One euro spent, two ways to book it. What changes is when the cost hits profit, and for a regulated grid operator, whether that euro earns a return.
Spending becomes capex only if it creates an asset: a resource the company controls, that is expected to bring future economic benefits, and whose cost can be measured reliably. For intangibles like software, IAS 38 adds that it must be identifiable (separable, or arising from a contract or legal right).
Control is the word people skip. It means the power to get the benefits and to stop others getting them, usually through a legal right. You do not need to own the IP: a licence you control is an asset. A subscription you lose the day you stop paying is a service, so it is opex.
Importance, size or "strategic" labels do not make something capex. Only the asset test does.
Same mechanism, two names. Depreciation spreads the cost of physical assets (cables, transformers, servers). Amortisation does the same for intangibles (software, licences, patents). Together: D&A.
Straight-line is the default:
Figures in €k, 3-year life. Years 2 and 3 repeat only the amortisation lines. Cash leaves on the same day in both cases.
The P&L runs top to bottom: revenue, minus operating costs, gives EBITDA; minus D&A gives EBIT; minus interest and tax gives net profit.
Opex sits above EBITDA. D&A sits below it. So capitalising moves cost out of EBITDA entirely, and investors, lenders and bonus schemes often watch EBITDA. It also moves the cash out of operating cash flow into investing cash flow, so "operating cash flow" looks better too.
Total profit over the asset's life is identical either way. Only the timing and the line it sits on change.
A regulated distribution operator cannot set its own prices. The regulator sets allowed revenue, roughly:
That asymmetry is the capex bias: capitalised spend makes money, expensed spend does not. Regulators fight it with TOTEX (treat all spending alike) or, in Italy, a fixed share of all spending capitalised regardless of its label.
The WACC is not pure margin. It compensates lenders and shareholders. The operator's real gain is the gap between the allowed rate and what its capital actually costs.
Maintaining an asset you already have is opex. Only spending that adds new capability can be added to the asset, and IAS 38 says that happens "only rarely".
Once a cost has been expensed it can never be capitalised retroactively (IAS 38 para. 71).
Paying in advance for a service creates a short-term asset that unwinds as the service arrives. It never enters D&A.
Staff time spent building an asset can go on the balance sheet, which is one reason companies build in-house instead of buying SaaS.
If an asset loses its value early, the remaining book value is written off in one hit. The risk of capitalising.
Statutory accounts, tax and the regulator can each use different lives. In Germany the regulator fixes software at 3 to 5 years.
A signed but unperformed contract is not a liability, unless it turns loss-making (onerous). Then a provision is booked.
IFRS does not define EBITDA. Each company computes its own, so compare definitions before comparing numbers.
EBITDA: no change, because the licence is capitalised and amortisation sits below EBITDA. Net profit: down €300k (1.2m ÷ 4), before tax. Cash: down €1.2m in investing cash flow.
Opex: it buys a service consumed over time, not an asset. Outside the base year it is not in the allowance, so the operator carries it until the next base-year reset. Bought in the base year itself, it is built into the following period's allowance.
Because it pays for the capital tied up: interest to lenders and a return to shareholders. The operator only gains where the allowed rate exceeds its real cost of capital. It is also paid on a shrinking balance, so over a 3-year life it adds up to about 13% of the original cost, not 6.58% a year on the full amount.