A software company that develops its own product in the Netherlands is the archetypal candidate for the innovatiebox, and the reason is written into the statute rather than inferred from it. Software appears by name in the list of qualifying titles, which means a large taxpayer whose intangible asset takes the form of software has one of the two entry conditions satisfied by the nature of the asset itself.
The regime is nevertheless misdescribed almost everywhere, in four specific ways. Its headline figure is not a rate. Its entry conditions are not the same for everyone. Its calculation runs in an order most summaries reverse. And the step that founders most often assume comes first, an agreement with the tax authority, is not what the statute conditions the relief on at all.
What the box does to the number
Article 12b does not set a rate. It provides that qualifying benefits from a qualifying intangible asset produced by the taxpayer itself are taken into account, if the taxpayer so elects in the return for a year, for a fraction of 9 divided by H, where H is the percentage of the highest rate of article 22 applicable in the year the benefit was enjoyed. The relief is a partial inclusion in the base, not a separate tariff.
The arithmetic then produces the familiar figure. With the top rate at 25.8 per cent for 2026, including a euro of qualifying profit at 9 divided by 25.8 of its amount and taxing that fraction at 25.8 per cent leaves an effective burden of 9 per cent, which is how the Belastingdienst states it on its own pages.
The design is self correcting, and this is the part worth understanding rather than memorizing. Because the numerator is fixed at 9 and the denominator tracks the general rate, a change in the general rate changes the fraction and leaves the effective outcome where it was. Saying that the innovation box becomes more expensive if corporate income tax goes up is simply wrong.
Two tickets at the door, and which one you get
The entry condition depends on the size of the taxpayer. Both routes require an intangible asset arising from research and development work for which an S&O-verklaring has been issued to the taxpayer under the wage tax remission act. For a smaller taxpayer, that is the whole requirement.
A larger taxpayer needs the S&O-verklaring and, in addition, at least one of eight further titles. The list runs through patents and plant breeders’ rights granted or applied for, an asset that takes the form of software, certain authorizations under the plant protection and medicines regimes, a supplementary protection certificate from the Octrooicentrum Nederland, a registered utility model protecting the innovation, and an asset linked to any of those. Software sits third on that list, which is why the second ticket is rarely the obstacle for a software group.
Whether a taxpayer is smaller is a defined question with two conditions that must both be met, each measured over the year and the four preceding years, so over a window of five financial years. The gross benefits from the intangible assets covered by the S&O-verklaring, increased by the costs of obtaining those benefits, must together be lower than 37.5 million euro. And the net turnover of the taxpayer, or of the group of which it forms part, taken from the annual or consolidated accounts, must together be lower than 250 million euro. Where the taxpayer belongs to more than one group in the year, the group with the highest net turnover is taken. Both are written as lower than, not as up to, and at the boundary that distinction decides the case.
What has to be yours
The asset has to have been produced by the taxpayer itself. Developing someone else’s asset further is not enough on its own: there is a self produced asset only to the extent that the research and development work has given rise to a new asset arising from that work. Buying intellectual property and improving it does not convert the acquisition into a qualifying asset.
Exclusive licences are treated as equivalent to the underlying title, so a company holding an exclusive licence over one of the eight titles is not outside the regime for that reason. The equivalence has a limit, though: in that case only the benefits derived from the use of the exclusive licence qualify.
The statute also excludes a category outright. Trademarks, logos and comparable assets are not qualifying intangible assets. For a software company this matters at the margin between the product and the brand, because the brand may be the more valuable asset commercially and it is outside the box entirely.
The threshold comes before the fraction
The qualifying benefits are the benefits from the asset, meaning profit rather than gross revenue, and only where the balance of those benefits is positive. Before anything else happens to them, they are reduced by the threshold.
That threshold is not a minimum turnover, which is how it is usually described. It is the balance of production costs of the asset that have not yet been recovered. Until the cumulative benefit from the asset exceeds what it cost to develop, nothing enters the box. For a software company that has spent several years and a substantial payroll building a product before it earned anything, the threshold is the reason the relief starts later than the model assumed.
The order is explicit in the statute and it is the point most summaries get backwards. The nexus fraction is applied to the benefits after they have been reduced by the threshold, not before. And the threshold is not something the taxpayer tracks privately: the inspector fixes, by decision open to objection issued at the same time as the assessment for the year, the balance of unrecovered production costs of each intangible asset at the end of the year.
The nexus fraction and what it punishes
The qualifying benefits are then set at the fraction K divided by T, but not more than the whole. T is the total of the expenses of the year and of all preceding years incurred to produce that asset. K is the qualifying part of those expenses, multiplied by 1.3.
K covers the taxpayer’s own expenditure on research and development performed by it for the asset, and expenditure on outsourcing research and development, directly or indirectly, to an unrelated party, together with expenditure incurred under a collaboration agreement to produce the asset for joint account. What K does not cover is research and development outsourced to a related party, or the cost of acquiring intangible assets from others. T covers all of it.
That asymmetry is the whole mechanism. A group that develops its product in a related development company, or that has bought in the core technology, sees the fraction fall, because the denominator grows while the numerator does not. The 1.3 uplift softens the effect but is capped at the whole, so it cannot rescue a structure in which related party or acquired development carries most of the weight. Expenditure on debt, on premises and anything else without a direct connection to the research and development is excluded on both sides. Where the figures cannot be determined precisely, the statute requires them to be determined in the most appropriate manner, having regard to the nature of the business and of the research and development, and allows linked assets to be dealt with as a group.
The flat alternative
There is a simplified route for taxpayers who do not want to build that calculation. A taxpayer that has produced a qualifying intangible asset in the year, or in one of the two preceding years, may elect to set the balance of qualifying benefits at 25 per cent of the profit before applying article 12b, with a maximum of 25,000 euro, and only where that profit is positive.
The maximum is the point. Twenty five thousand euro of qualifying benefit is a modest relief, and the flat route is designed for the small company for which the documentation of a full nexus calculation would cost more than the difference. Under a collaboration agreement the percentage is apportioned according to the share in the profits, subject to the same overall ceiling. Where the flat route is used, one component of the threshold calculation is set at zero.
For a software group of any scale the flat amount is quickly irrelevant, but it is a reasonable landing place in the first years after a product is finished, while the unrecovered development costs still stand between the company and the full regime.
What the file has to contain
The documentation duty sits in article 12bg and is additional to the administration of hours and costs that the S&O-verklaring already requires under the wage tax remission act. A taxpayer electing into the regime must include in its accounts data showing three things: that it holds one or more qualifying intangible assets produced by itself, up to what amount it obtains qualifying benefits from those assets, and in what manner those benefits have been determined in the most appropriate way.
The third item is the one that takes work. It is a duty to evidence a method, not a figure, which means the reasoning by which revenue was attributed to the asset has to exist as a document rather than as a conclusion. A taxpayer relying on the smaller taxpayer route must additionally hold data showing that it meets the conditions of that route, which in practice means the five year measurements have to be maintained rather than reconstructed.
The election is made in the return
The last surprise is the most useful one. The statute makes the regime an election exercised in the corporate income tax return for the first year in which it is to apply, and the election can be made while the assessment for that year is not yet final. Nothing in article 12b conditions the relief on an agreement reached with the Belastingdienst in advance, and the authority’s own page on how to elect describes the election through the return.
That is worth stating precisely, because advance certainty is discussed so routinely in this area that founders assume it is a precondition. It is not. Whether a company seeks discussion with the inspector before filing is a practical decision about risk, cost and the complexity of its own calculation, not the gateway to the regime.
None of the articles governing the regime carries a notice of future amendment in the consolidated text in force for 2026, which makes it one of the more stable parts of Dutch corporate income tax. That stability is a reason to build the file properly. The regime rewards a company that can show what it built, what it spent to build it, and where the money it now earns actually comes from, and that evidence has to be created while the work is being done.