With effect from 1 January 2026, the legislature has appreciably widened the research allowance for the third time since it was introduced. The Act on an immediate tax investment programme to strengthen Germany as a business location, approved by the Bundesrat on 11 July 2025 and promulgated in the Federal Law Gazette of 18 July 2025, raises the ceiling on eligible expenditure from €10m to €12m per financial year and once again brings the treatment of own work, contract research and overheads closer to the reality of research in German industry[3][4]. For SMEs under Annex I of the General Block Exemption Regulation, that means, in the maximum case, an allowance of €4.2m a year — per group of companies, of course. In parallel, the number of projects assessed by the Certification Body for the Research Allowance (BSFZ) has passed 59,000, and with it the density of material for a view the Mittelstand could not have had when the scheme began: which documentation errors sink applications in the first BSFZ wave, and what does an application look like that also survives a later tax audit?
What changes in 2026
The research allowance has been in force since 1 January 2020 and has, in six years, become the largest direct R&D support scheme for companies in Germany. ZEW policy brief 09/2025 puts the expected annual volume after the 2025 reform at around €4bn and describes the cumulative increase in the funding volume since introduction as 124 per cent against the original framework[11]. The main changes taking effect for financial years from 2026 can be grouped into five blocks.
First, the assessment base. Section 3(5) FZulG now grades the maximum eligible expenditure in four steps: €2m (January to June 2020), €4m (July 2020 to 27 March 2024), €10m (28 March 2024 to 31 December 2025) and €12m from 1 January 2026[1]. Second, the hourly rate for own work: sole traders and active partners may claim €100 per hour, up to 40 hours a week, for hours worked after 31 December 2025 — previously the rate was €70[5]. Third, a flat-rate recognition of overheads: for financial years beginning after 31 December 2025, a further 20 per cent of the other eligible expenditure may be claimed as part of the assessment base, with no itemised evidence[1]. Fourth, contract research: for contracts awarded after 27 March 2024, 70 per cent of the fee counts as eligible expenditure, up from 60 per cent[1]. Fifth, the ten-point SME bonus under section 4 FZulG remains, so SMEs continue to apply a rate of 35 rather than 25 per cent[2].
The aid ceiling per project under section 4(2) FZulG remains €15m, but the effect on total funding in groups of companies is worth noting: linked enterprises share a joint annual assessment base under section 3(6) FZulG, which is the critical parameter for Mittelstand holdings with several operating units — more on that in section 3[1].
The assessment base in detail
Taking the €4.2m figure as a headline leads into two arithmetic traps. First, the assessment base is not a cash but an expenditure and payroll calculation. Second, it applies per linked group of enterprises, not per legal entity. A sober look at section 3 FZulG is therefore obligatory for mid-sized companies.
Staff expenditure under section 3(1) FZulG consists of the wages subject to wage tax deduction plus the employer's contributions to social security — but only to the extent the employee concerned worked on the eligible R&D project. In practice, gross salaries plus employer social security contributions count, but not one-off bonuses without a link to the work, or tax-free travel expenses[1]. Own work by shareholders or sole traders is capped under section 3(3) at the €100 hourly rate mentioned, up to 40 hours a week. Active partners in a partnership can reach the same rate too, if a service or employment contract expressly regulates remuneration and the record of hours is robust.
New for the 2026 financial year is the overhead flat rate under section 3(3b) FZulG. Twenty per cent flat on the other eligible expenditure sounds like little, but it appreciably raises the assessment base on typical Mittelstand projects with staff-intensive development — with no need to evidence rent, IT, energy or quality costs individually[5]. For companies that have so far shied away from attempting overheads at all, this is the easiest lever in 2026.
Depreciation on movable assets is eligible under section 3(3a) FZulG if the asset was acquired after 27 March 2024 and is used exclusively in the R&D project[1]. Exclusivity is the hurdle at which claims fail in practice: a test rig running research Monday to Thursday and series production on Friday is not, in the eyes of the BSFZ and the tax office, exclusively dedicated to R&D — part-time use requires clean, robust time allocation, or the depreciation share drops out of the assessment base entirely.
For many mid-sized companies, the treatment of linked enterprises is the decisive point. Section 3(6) FZulG provides that enterprises belonging to a group have to share the maximum assessment base[1]. Anyone running an operating GmbH, a sales company and a holding company within a group and locating development in all three units has to divide the €12m limit between them in advance — or face a dispute with the tax office at the assessment stage. The yardstick for defining the group is Article 3 of Annex I to the GBER, which defines linked and partner enterprises[9].
SME vs non-SME — the ten percentage points
Section 4 FZulG grants a research allowance of 25 per cent of the assessment base; SMEs under Annex I of the GBER can claim, on application, an increase of ten percentage points to 35 per cent[2]. The difference sounds modest but decides €1.2m of annual funding when the assessment base is fully used — for a mid-sized high-tech developer, the difference between an additional development team and financing the work at market rates.
The conditions for SME status are clearly defined in Annex I of the GBER: fewer than 250 employees, annual turnover up to €50m or an annual balance sheet total up to €43m; the independence criteria under Article 3 apply on top[9]. The catch — as ever — lies in the criteria that rest not on individual accounts but on consolidated figures. Three constellations regularly cost Mittelstand companies the SME bonus.
First: a holding of more than 25 per cent by a strategic investor or a parent company. Article 3(2) of Annex I to the GBER classifies such an enterprise as a partner enterprise; where the holding exceeds 50 per cent or there is a controlling relationship, the enterprise becomes linked and headcount, turnover and balance sheet figures are consolidated[9]. Many mid-sized companies that are clearly SMEs on their own figures lose the status through the parent holding — and yet the application is often still made at 35 per cent. That surfaces at the tax audit at the latest.
Second: growth past the thresholds. The reference date rule in Article 4 of Annex I to the GBER allows a buffer in time: the enterprise loses its SME status only once the thresholds are exceeded in two consecutive financial years. For the financial year in which a threshold is first breached, the 35 per cent rate remains permissible in principle — documenting the closing figures then becomes relevant to any review[9].
Third: public shareholdings. If a public body holds more than 25 per cent, the enterprise is not an SME regardless of its size — with narrow exceptions for business angels, universities, and autonomous local authorities with fewer than 5,000 inhabitants and an annual budget below €10m[9]. In North Rhine-Westphalia, Bavaria and Baden-Württemberg there are individual Mittelstand structures with municipal owners in which exactly this exception has to be checked.
The ZEW policy brief calculates that SMEs can claim around 48 per cent of the potential funding volume under the research allowance — helped by the fact that the cap on eligible expenditure limits payouts to large companies[11]. A Mittelstand company that applies as a non-SME is right to use the fallback rate of 25 per cent — one that is an SME but applies at 25 per cent gives away around a third of the allowance.
The first BSFZ wave: documentation errors
Six years after introduction, the certification body's files allow a sober look at typical error patterns. Between 16 September 2020 and 31 March 2026 the BSFZ registered 48,863 applications covering 59,537 projects[6]; the number of decisions has been growing since at a rate of several thousand project assessments a quarter. Combined with the review guide 10/2025 and the BSFZ's FAQ, a precise picture of the recurring documentation errors emerges[8][7].
The most common error pattern is confusing a product description with a research description. The BSFZ assesses projects against three overarching criteria: novelty, scientific and technical uncertainty, systematic planning — derived from the OECD Frascati Manual and the requirements of the GBER[7][12]. Anyone who describes a product benefit or a competitive advantage in the application form without placing the technical challenge and the advance in knowledge over the state of the art precisely misses at least two of the three criteria — regardless of how innovative the result is in the market. Rejection then follows not from technical triviality but from wording that fails to distinguish routine engineering from genuine research.
The second most important error pattern is incomplete records of hours. The application to the BSFZ is a pure certification procedure in which no vouchers, contracts or project reports have to be submitted — the BSFZ itself only assesses whether the project is eligible in substance, not the actual costs[7]. At the second stage, though, the tax office demands evidence that the hours claimed can be attributed to the R&D project in question and did not go into ordinary operations. Anyone reconstructing after the year-end which employees worked how many hours on which project will not reliably discharge the burden of proof.
The third error pattern concerns delimiting the project. On the BSFZ's definition, a project for the purposes of the research allowance covers a self-contained scientific and technical objective with a clearly defined work plan; standing themes such as "product maintenance" or "digitalising production" are not projects within the meaning of the Act[8]. Registering a collective project mixing several genuinely separable development and routine elements risks rejection of the whole certification — even where individual sub-projects would be eligible.
Fourth: contract research without a clean separation of contracts and invoices. For contracts awarded after 27 March 2024, 70 per cent of the fee is eligible; the condition is that the contractor is based in the EU or an EEA state and that the subject matter of the contract is contractually assigned to the R&D project[1]. Mixed invoices, showing development and series production on the same document, lead to reductions and even to the whole contract being disallowed at the assessment stage. In the Mittelstand this matters particularly, because regular suppliers often perform hybrid contracts — the clean cut between the development and the production part has to be built into the quotation from the start.
A worked example: a mid-sized company with 45 employees
The following case works through a typical Mittelstand constellation for the 2026 financial year on the basis of the current version of the Act. The starting point: a family-owned mechanical engineering company with 45 employees, annual turnover of €14m, a balance sheet total of €18m — clearly an SME under Annex I of the GBER. Development team: eight engineers, five of them working 80 per cent on R&D projects, three 100 per cent. Average gross salary €78,000, employer social security contributions at a flat 20 per cent. The managing shareholder demonstrably puts 15 working hours a week into leading development on a flagship project, over 46 working weeks a year. A contract with a research provider close to a technical university in Bavaria for €220,000, awarded in January 2026. Plus a CNC test rig acquired in April 2026, a depreciation base of €180,000 over five years, operated exclusively for R&D.
The calculation along the sections runs as follows. Staff expenditure under section 3(1) FZulG comes to: five people at 80 per cent R&D × €78,000 × 1.2 = €374,400; three people at 100 per cent × €78,000 × 1.2 = €280,800. Total staff expenditure: €655,200[1]. The managing shareholder's own work under section 3(3) FZulG: 15 hours × 46 weeks × €100 = €69,000. The research contract with the university-affiliated institute enters the assessment base at 70 per cent under section 3(4) FZulG: €220,000 × 0.7 = €154,000. The depreciation share for the test rig under section 3(3a) FZulG: €180,000 ÷ 5 = €36,000 of annual depreciation for 2026 (nine months from April pro rata: €27,000; to keep the example simple we use the full year).
Subtotal of other eligible expenditure without the overhead flat rate: €655,200 + €69,000 + €154,000 + €36,000 = €914,200. The overhead flat rate under section 3(3b) FZulG, at 20 per cent of that subtotal, adds a further €182,840[5]. That puts the assessment base at €1,097,040 — well below the new ceiling of €12m, but with a recognition of overheads that would not have been possible at all without the 2026 reform.
With SME status, the research allowance is 35 per cent of the assessment base under section 4 FZulG[2]: €1,097,040 × 0.35 = €383,964. Without the overhead flat rate it would have been €319,970 — the pure effect of the 2026 reform comes to around €64,000 for this company and this mix of projects. At an unchanged rate of 25 per cent (no SME status), the allowance would have fallen to €274,260 — the ten percentage points of SME bonus are worth €109,704 a year in this case.
The research allowance is not paid out but offset in the next tax assessment — either against the income tax assessed or, to the extent the allowance exceeds the tax, as a refund claim. For the mid-sized company in this example that means a liquidity effect of nearly €384,000, typically arriving 12 to 20 months after the end of the financial year. Over a multi-year development programme of three to five years it adds up to a volume that in the Mittelstand carries two or three additional development posts — or makes the step to series readiness financeable in the first place.
Compliance and tax audits
The two-stage nature of the procedure — BSFZ certification on eligibility in substance, assessment by the tax office on the amount — has an underestimated consequence for the Mittelstand: the tax audit arrives years after certification and demands evidence that the eligibility certified and the expenditure actually claimed match. The Finance Ministry's circular on the FZulG remains the central application guidance; a new version is going through consultation and will bring together the additions made by the Growth Opportunities Act and the immediate tax investment programme[10].
The audit applies its standard at three points. First, staff expenditure: the audit wants to see timesheets or comparable records of activity that attribute hours unambiguously to a certified project. The orally handed-down "80 per cent R&D" ratio, rarely questioned in the application, does not hold up in an audit without a written basis. Mid-sized companies that have so far recorded time only at project or cost centre level need finer granularity for the research allowance — ideally automated and audit-capable.
Second, delimitation against other funding instruments. Section 7 FZulG imposes a prohibition on cumulation: expenditure already funded under another state aid regime may not additionally enter the assessment base of the research allowance. The constellation that matters in Mittelstand practice is the combination with the Zentrales Innovationsprogramm Mittelstand (ZIM): staff expenditure and contract shares funded by the ZIM agency have to be taken out of the research allowance exactly, or recovery looms. Anyone drawing a ZIM grant and the research allowance on the same project has to keep a clean separation in the cost plan and in the accounts — at the level of the cost category, not the project name.
Third, consolidation at group level. Where several linked enterprises apply in the same financial year, the tax office requires an allocation calculation showing how the €12m ceiling was divided between them[1]. The Mittelstand regularly comes unstuck here on the organisational side: when development sits in one company, sales authority in a second and parts of the projects with a jointly controlled third, coordination is needed across company boundaries — and in the year-end rush it often starts too late.
The BSFZ review guide in its 10/2025 version is public and turns the three assessment criteria — novelty, uncertainty, systematic planning — into individual check points; for mid-sized companies, reading it before applying is not optional but obligatory[8]. To prepare for a tax audit, a documentation line is advisable that represents the chain from BSFZ certification to project description to time records to payroll accounts to fixed asset accounting to contract research invoices as a closed whole, reproducible at any time.
Two observations to close. The federal government's 30th subsidy report of October 2025 for the first time no longer classifies the research allowance as a tax concession within the meaning of the definition of a subsidy — a signal that the instrument is being anchored politically as a permanent element of German innovation policy[13]. And the ZEW policy brief notes that around 19,000 companies had applied by mid-2025 — against an estimated eligible population that is still considerably larger[11]. Taken together, the two observations mean: the research allowance is not going to be withdrawn, but it is used only by those who set the procedure up systematically — sound in substance, in the accounts and beyond the tax audit.
For mid-sized companies with 30 to 250 employees, the 2026 framework is more than an update to a ceiling: the overhead flat rate, the raised hourly rate for own work and the 70 per cent rate for contract research widen the assessment base systematically and, combined with ZIM and EU projects, make the allowance a stable backbone of R&D financing. The barrier to entry stays the same — a clean project architecture built on novelty, uncertainty and systematic planning, one that holds up before the BSFZ assessment just as it does before the tax audit three years later.
- [1]Section 3 FZulG — eligible expenditure and the assessment basegesetze-im-internet.de (BMJ) · 2026Open source
- [2]Section 4 FZulG — the amount of the research allowancegesetze-im-internet.de (BMJ) · 2026Open source
- [3]Act on an immediate tax investment programme to strengthen Germany as a business location (Federal Law Gazette I no. 161, 18 July 2025)Federal Law Gazette · 2025Open source
- [4]Legislative file 281/25 — immediate tax investment programme (approved 11 July 2025)Bundesrat · 2025Open source
- [5]The immediate tax investment programme — the research allowanceCertification Body for the Research Allowance (BSFZ) · 2026Open source
- [6]Current figures from the Certification Body for the Research AllowanceBSFZ (DLR Project Management Agency / VDI/VDE-IT / AiF Projekt GmbH) · 2026Open source
- [7]FAQ — assessment criteria and the application procedure for the research allowanceBSFZ · 2026Open source
- [8]Review guide of the Certification Body for the Research Allowance, as at 10/2025BSFZ · 2025Open source
- [9]Annex I to the General Block Exemption Regulation (Regulation (EU) No 651/2014) — the SME definitionEUR-Lex / European Commission · 2014Open source
- [10]The research allowance — topic page with the Finance Ministry circular on the FZulGFederal Ministry of Finance · 2026Open source
- [11]Rammer, C.: The research allowance — a new funding instrument grows and thrives (ZEW policy brief no. 09/2025)ZEW — Leibniz Centre for European Economic Research · 2025Open source
- [12]Frascati Manual 2015 — Guidelines for Collecting and Reporting Data on Research and Experimental DevelopmentOECD · 2015Open source
- [13]30th subsidy report of the federal government (Finance Ministry monthly report, October 2025)Federal Ministry of Finance · 2025Open source
The research allowance in 2026 for startups: what a €4.2m ceiling really means
A €12m assessment base, a 35% SME rate on application, plus 20% overheads — a worked example for an eight-person engineering team.
The de minimis trap: when one funding programme blocks the next
The €300,000 threshold counts across three tax years — and across group boundaries. How to stop application B devaluing application A.
ZIM 2026: what the BMWK reform changes for Mittelstand applications
New rates, new priorities, new deadlines — and what innovation managers should adjust in their pipeline planning now.
From the analysis into the application.
We show the platform on a real case.