Forschungszulage 2026: R&D Tax Credit for German Startups

You are building software, running experiments, developing a device. The Finanzamt writes you a cheque for 25% to 35% of the personnel and contract costs, even if your GmbH has not paid a single euro in Körperschaftsteuer yet. That is the Forschungszulage (R&D tax credit). The 2026 update raises the ceiling from EUR 3.5M to EUR 4.2M per year for SMEs. Here is what is eligible, how to apply, and where most startup applications actually fail.

What the FZulG actually funds

Eligible expenses are listed in § 3 FZulG and fall into four buckets:

The allowance is the sum of eligible expenses, capped at the annual Bemessungsgrundlage, multiplied by the rate:

• 25% base rate under § 4 Abs. 1 FZulG
• plus 10 percentage points for qualifying SMEs (up to 35%)

The maximum state aid per R&D project per company is EUR 15M (§ 4 Abs. 2 FZulG). That ceiling only bites for very large programs. Most startups hit the annual Bemessungsgrundlage first.

The 2026 uplift: what actually changed

The Forschungszulagengesetz was amended most recently on 22 December 2025 (BGBl. 2025 I Nr. 363). From 1 January 2026:

• Bemessungsgrundlage rises from EUR 10M to EUR 12M per year (§ 3 Abs. 5 FZulG)
• Maximum SME allowance rises from EUR 3.5M to EUR 4.2M per year
• New: 20% overhead flat rate on direct project costs
• Self-employed and partner hourly rate: EUR 100 per R&D hour (up from EUR 70)

For a seed-stage team spending EUR 2M per year on direct R&D personnel (two developers and a technical founder working fulltime on the project), the new overhead pauschale alone adds EUR 400,000 to the Bemessungsgrundlage. At the SME rate that is EUR 140,000 in additional allowance with zero additional spend. Note: the pauschale applies to direct project costs only, not to total company burn.

Two-stage application: BSFZ first, Finanzamt second

The process runs across two authorities. Most founders underestimate the first step.

1. BSFZ Bescheinigung (Bescheinigungsstelle Forschungszulage). You describe the project, its novelty, its technical risk, and the method. The BSFZ decides whether the project qualifies as Grundlagenforschung, industrielle Forschung, or experimentelle Entwicklung under § 2 FZulG. Submission is through the BSFZ web portal, with no fee.
2. Finanzamt application via ELSTER, filed after the end of the Wirtschaftsjahr. You attach the BSFZ Bescheinigung, list the eligible expenses for the year, and the Finanzamt offsets the allowance against your Körperschaftsteuer. Any excess is paid out in cash under § 10 FZulG.

The two steps can run in parallel. You can file the BSFZ application as soon as the project is defined (even mid-year) and then submit the Finanzamt application once the year closes. Completed years remain claimable for the four most recent Wirtschaftsjahre under the general Festsetzungsfrist of § 169 AO. The Anlaufhemmung in § 170 AO can extend that window if the Körperschaftsteuererklärung for the relevant year was filed late.

Where BSFZ applications actually fail

The BSFZ rejection rate is rising. Table.Media's review of BMFTR figures shows 24% rejected in 2020, roughly 26% in 2022, and around 29% in 2024. That is worth internalising before submitting.

In the startup applications I see, rejection clusters around language, not science. The BSFZ reviewer is looking for one thing: systematic, uncertain, novel technical work. If the project description reads as "we built feature X," it dies. If it reads as "we tested hypothesis Y about method Z and did not know whether A would work," it usually goes through. This is true regardless of the actual difficulty of the underlying work. I routinely see teams doing genuinely hard engineering get rejected because their application reads like a product roadmap, while teams doing modest iterations on existing tech get approved because they framed the work as controlled experimentation.

Three concrete rejection triggers:

• "Optimization" language. Routine refinement of existing products does not meet the § 2 FZulG definition of experimental development and is routinely rejected. Replace "optimize" with "investigate."
• No clear Forschungsfrage. Every project needs a single technical question the team could not answer in advance.
• Vague method. "Using AI" is not a method. "Comparing transformer architecture A against B on dataset C" is. The BSFZ Prüfleitfaden (Stand 10/2025) makes clear that reviewers look for falsifiable method, not outcome.

Startup-specific: the cash-refund timing

The payout mechanism that matters for founders is § 10 FZulG. The allowance reduces the Körperschaftsteuer liability first. Any excess is paid out as Körperschaftsteuererstattung, subject to offset against other open tax liabilities of the GmbH under § 226 AO (for example, unpaid Lohnsteuer or Umsatzsteuer). For startups with no KSt liability and no other open tax arrears, the whole excess arrives as cash.

The total cycle from R&D spend to cash in the bank is slower than founders expect:

• BSFZ decision: 4-6 months typical
• Finanzamt processing after year-end filing: 3-6 months
• Combined: 9-12 months from close of Wirtschaftsjahr to payout

The Forschungszulage is therefore not an answer to immediate runway questions. It is a structural reimbursement that lowers the true cost of R&D over the project's life. Used well, it converts a meaningful portion of R&D spend into refundable expenditure. Employee participation schemes like VSOP do not reduce the Bemessungsgrundlage, since § 3 FZulG counts the full Lohnsteuer-pflichtiger Bruttolohn regardless of non-cash compensation. Public grants that fund the same personnel costs, however, are excluded under the Kumulierungsverbot in § 6 FZulG.

Bottom line

For an R&D-heavy startup the Forschungszulage is the most reliable non-dilutive funding source in Germany. The 2026 uplift moves the SME ceiling from EUR 3.5M to EUR 4.2M per year and adds a 20% overhead pauschale that did not exist before. The application risk sits almost entirely at the BSFZ stage, and rejections there are driven by description quality, not project merit. File the BSFZ application early, write it as a controlled experiment, and plan for a 9-12 month lag from spend to cash. The corporate tax framework lowers the effective rate on retained profits over the 2028-2032 trajectory; the Forschungszulage lowers the effective cost of the R&D that produces those profits in the first place.