R&D tax credit in mechanical engineering: what is eligible and what is not

Developing new machinery, control systems or manufacturing processes? Germany’s research allowance can support your development work where novelty, technical risk and a systematic approach can be demonstrated. Explore eligible activities and the boundaries with routine work below.

Check eligibility Dossier

Funding rate up to 35%
Design & testing eligible
Retroactive for open years

Written by Max Nodes, Founder and Managing Director · 25 of 25 supported BSFZ applications certified (as of 09/2026) · Last updated: September 22, 2026 · Law as of: September 2026

Example: thermal correction

Mechanical engineering: extend the process window

New approachThermal correction directly at the CNC cell

Open questionHow does the part stay accurate under heat?

ExperimentCompare sensor readings with the toolpath

In mechanical engineering the scope of the project decides, not the size of the machine.

The certification office does not assess how complex a plant was, but whether a technical question was open beforehand and answered systematically. A special-purpose machine builder rarely loses an application on the technology; usually it is lost because routine design and development sit in the same project.

Definition

Is development in mechanical engineering eligible?

Yes, as soon as the behaviour of a design was open before the first test run. The rate is 25 % of the assessment basis, 35 % for SMEs; for contract research commissioned after 27 March 2024, 17.5 % or 24.5 % of the fee before the additional flat rate (Section 4 (1) FZulG). Eligibility is decided by novelty, technical risk and a systematic approach under Section 2 FZulG, never by the sector.

Novel kinematics, drive and control concepts, manufacturing processes, test rigs
Employer gross pay of designers and test engineers for the hours in the project
Material for the prototype itself is not eligible
Routine design following transferable rules falls out

Rates, costs and limits

The essentials on the research allowance in mechanical engineering

Funding rate: 25% of the assessment basis, 35% for SMEs. For projects started after 31 December 2025, a 20% flat rate increases the assessment basis (Section 3 (3b) FZulG); the EUR 12 million cap still applies (Section 3 (5) FZulG).
Eligible: Novel kinematics, drive and control concepts, manufacturing processes, test carriers and prototypes. What counts is not the industry but whether novelty, technical risk and systematic approach can be demonstrated.
What counts: The employer gross salary of design, test and simulation engineers for the hours spent on the project. Since 2024 also the depreciation of movable assets used exclusively for the project, such as a test rig. Material for the prototype itself is not eligible.
Ceiling: Assessment base of EUR 12 million per financial year, maximum research allowance EUR 4.2 million.
Contract research: 70% of the cost of commissioned development work is eligible (Section 3 (4) FZulG): materials testing, simulation providers, university institutes, external design offices.
Retroactive: Applications for closed financial years from 2022 are possible as long as the assessment period of the respective year is still open; for 2020 and 2021 it has expired (§ 12 FZulG in conjunction with § 169 AO). Projects started before 1 January 2026 do not receive the overhead allowance.

Which activities in mechanical engineering are eligible?

The industry does not decide eligibility. The certification office checks every project against three criteria: novelty, technical risk, systematic approach. If one fails, the application fails. Below are the activities we encounter most often at machine and plant builders, with the verdict we expect.

Boundary case: eligibility depends on the specific project.

Design & kinematics

Novel kinematics or axis arrangement with unknown vibration and stiffness behaviourEligible
Eligible. Novelty: No machine concept available on the market achieves the required dynamics or accuracy within your installation space or load specification. Technical risk: Whether natural frequencies, stiffness and positioning accuracy land in the target range is not known before simulation and test carrier. Systematic approach: Layout, FEA, test carrier, measurement, iteration, documented by calculation reports and test protocols.
Scaling a proven machine to a new size classBorderline
Borderline. Routine as long as the design rules transfer. It becomes eligible when stiffness, thermal behaviour or drive sizing demonstrably no longer scale linearly at the new size and have to be re-derived. The evidence is the calculation showing where the old rule fails.
Variant design for a load requirement outside the previous design rangeBorderline
Borderline. Eligible is the part for which no validated design exists and which is resolved in testing. Not eligible is adapting known variants within the catalogue, even if it takes weeks.
Adaptation design to customer request within the known scopeNot eligible
Not eligible. Interfaces, dimensions, options: the result is known before the start, the procedure is documented. Effort is not a criterion of the FZulG.

Drives, control & regulation

Own control strategy for coupled axes or process variablesEligible
Eligible. Anyone who designs a controller because standard controllers do not achieve path accuracy, damping or process stability must explain why. That names the state of the art. Keep the rejected controller structures and their measurements; that is the decisive evidence later.
Sensor fusion or in-process measurement to close the loop on a manufacturing processEligible
Eligible. Whether a process can be controlled in closed loop with the available signals at all is open beforehand. Measurement series showing signal quality and control performance over iterations are systematic approach and technical risk in one.
Retrofitting an existing machine with a new control systemBorderline
Borderline. Eligible if the interaction between legacy mechanics and new control is unresolved and worked out in testing. Routine if manufacturer documentation and a parameterisation guide prescribe the way.
PLC programming to specification, visualisation, parameterisationNot eligible
Not eligible. Intended use of the control platform. Only when the platform demonstrably cannot map the requirement and a separate method is developed alongside it does an open technical question exist.

Manufacturing processes & materials

New parameter window or process for a hard-to-machine materialEligible
Eligible. Tool life, surface quality and dimensional accuracy conflict and are measurable. That makes the technical risk quantifiable, the strongest starting position an application can have.
Additive manufacturing of load-bearing parts with open fatigue strengthEligible
Eligible. Whether the anisotropic properties of the layer-built material reach the required values under cyclic load cannot be determined before the test series. The series across parameters, orientation and post-treatment is your evidence of systematic approach.
Process optimisation on an existing lineBorderline
Borderline. Eligible as soon as a target value is required that the known process demonstrably fails to reach, and the cause is unclear beforehand. Without that measurement the certification office reads only a demanding requirement, and demanding is not a criterion of the FZulG.
Tool, supplier or machine change with run-in to manufacturer specificationNot eligible
Not eligible. Even elaborate run-in remains commissioning. The way is documented, the result predictable.

Prototype, testing & commissioning

Test carrier to resolve a question that was open beforehand, with measurement plan and abort criteriaEligible
Eligible. The clearest case in this group. Whoever defines in advance what is measured and when the trial counts as failed has already documented systematic approach and technical risk before the certification office asks.
First commissioning of a special-purpose machine at the customerBorderline
Borderline. Eligible is the share in which behaviour not previously mastered is worked out on the real process, documented by change states and measurements. Acceptance to specification is delivery.
Series ramp-up, acceptance, training, documentationNot eligible
Not eligible. Necessary and often laborious work, but no gain of knowledge in the sense of the FZulG. You hand over a result that is technically settled.

Digitalisation & Industry 4.0

Physics-based digital twin or condition model with own modellingEligible
Eligible. Whether a reduced model predicts the dynamic behaviour of the machine with sufficient accuracy is open beforehand. Comparing model and measurement over several iterations is the evidence.
Connecting legacy machines with unclear signal and data qualityBorderline
Borderline. Eligible if it is documented that the available signals do not support the required statement and own methods for conditioning had to be developed. Pure wiring, protocol mapping and dashboards are integration.
Introducing MES, ERP, dashboards or interfaces to standardNot eligible
Not eligible. Even demanding integration work remains integration. The tools are built for exactly this purpose, the way is documented, the result predictable.

What is regularly rejected

Quotation and order design to specificationNot eligible
Not eligible. The requirement is defined, the solution from the modular kit known. The exception is a clearly separable part for which no validated design exists. Then we apply for that part, not the order.
CE conformity, risk assessment, technical documentationNot eligible
Not eligible. Mandatory work under the Machinery Regulation whose result is settled. The test protocols created along the way can, however, serve as evidence for an eligible project.
Contract manufacturing and supply without own development riskNot eligible
Not eligible. Whoever manufactures to someone else's drawing bears no technical risk. The case is different when the supplier develops the component itself and is accountable for its behaviour.
Spare-part, variant and catalogue maintenanceNot eligible
Not eligible. The result is fixed beforehand: the same functional scope in a new form. Occasionally submitted by advisers along with everything else, and then it drags the whole project into scrutiny.

Your project is not listed or sits between two lines? That is the normal case; eligibility is decided on the specific facts, not on the heading. Describe the project to us and we will tell you where it stands.

What evidence does the BSFZ expect in mechanical engineering?

Mechanical engineering companies practically never introduce an additional system for the research allowance; they have the evidence already. We evaluate what exists and assign each source to the criterion it supports.

What evidence does the BSFZ expect in mechanical engineering?
Source	What it proves	Criterion
CAD revisions · PDM/PLM	Design revisions show attempt, rejection and new approach, the strongest evidence of experimental development a design department can deliver.	Systematic approach · Technical risk
Test rig and trial protocols	Aborted runs and measurement series date what was tried when and what it was measured against. A failed run is no blemish; it is proof of an open outcome.	Systematic approach · Technical risk
FEA · CFD · multibody simulation	Calculation reports with boundary conditions, convergence studies and comparison against measurement show that the target value was not trivially achievable.	Technical risk
Change requests · ECR/ECO	Change histories name the rejected solutions, exactly the information with which the state of the art can be demonstrated.	Novelty
Design meetings · design reviews	Minutes record where a solution was uncertain and which assumption had to be tested. Rarely read, often decisive.	Technical risk
Commissioning protocols · measurement reports	Measurements before and after the intervention document the difference between running in to instructions and working out previously unknown behaviour.	Scope
Time recording · project hours	For the tax office application the hour counts. We map an existing time recording onto the projects instead of introducing a parallel system.	Cost side
Standard tests · materials test reports	Tensile, fatigue and wear tests to DIN and ISO document the experimental character when they reflect iterations, not just acceptance.	Systematic approach

The most common mistake in assisted applications is a development process constructed after the fact with phases that never existed at your company. Your real test rigs and change histories are the better evidence; they provide a complete, dated record of the procedure.

What comes out of it?

Two calculated cases for projects that started after 31 December 2025, once at the SME rate of 35%, once at the standard rate of 25%.

Special-purpose machine builder, 50 employees · SME

Employer gross, 12 design and test engineers: EUR 960,000. Twelve design and test engineers, R&D share documented via project hours, plus a testing contract placed with a materials laboratory

R&D personnel costs (employer gross)EUR 432,000

Contract research, 70 % of EUR 60,000+ EUR 42,000

Overhead flat rate 20 %+ EUR 94,800

Eligible expensesEUR 568,800

Funding rate35 %

Research allowance per yearEUR 199,080

Per year. Applied for retroactively over three open financial years: around EUR 0.5 million, without the overhead allowance, because the project started before 2026.

Machine tool manufacturer, 300 employees

Employer gross, 30 developers: EUR 2,700,000. Thirty developers working on a new spindle and axis generation, not an SME. Standard rate 25%, contract research with a university institute

R&D personnel costs (employer gross)EUR 1,620,000

Contract research, 70 % of EUR 200,000+ EUR 140,000

Overhead flat rate 20 %+ EUR 352,000

Eligible expensesEUR 2,112,000

Funding rate25 %

Research allowance per yearEUR 528,000

Per year. The assessment base is capped at EUR 12 million, a maximum of EUR 3.0 million allowance at the standard rate, EUR 4.2 million as an SME.

As a rule of thumb: a design engineer with an employer gross of EUR 80,000 who spends half of the time on eligible projects yields around EUR 16,800 research allowance per year (SME rate, with the overhead allowance — the project starts after 31 December 2025). The whole is capped at an assessment base of EUR 12 million and an allowance of EUR 4.2 million per year. Go to the calculator Approved projects show the order of magnitude this reaches at machine and plant builders.

Guide: R&D tax credit for machine and plant builders
The eligibility rules, the sources of evidence and a template for the project description, as a PDF.

What mechanical engineers ask us most often

Are machine prototypes and physical samples fully eligible?+

Prototypes are eligible as an instrument of experimental development insofar as they serve to gain knowledge. The direct material cost of the prototype cannot be claimed, however. Eligible are the personnel costs of the employees involved, 70% of commissioned contract research and, since 2024, the depreciation of movable assets used exclusively in the project. The finished series product is not part of the funding.

Is special-purpose machine building for a specific customer eligible?+

Yes, if the company itself bears the technical risk and part of the development was open beforehand. Order design to specification is not. The scope decides: we apply for the part of the plant for which no validated design existed and leave quotation design, acceptance and documentation out.

Can machine-building suppliers have their component development funded?+

Yes, suppliers can have the R&D work for new components and systems funded, even if these are later installed in machines or vehicles by OEMs. What matters is that the supplier itself bears the technical risk and performs the development work. Pure contract manufacturing to someone else's drawing is not eligible.

Is simulation software (FEA, CFD) eligible as R&D expense?+

The personnel costs of the engineers who run simulations are fully eligible. Licence costs for the simulation software itself are not provided for as eligible expenses under the FZulG; they can be taken into account under other programmes such as ZIM. The calculation reports are at the same time some of the strongest evidence of technical risk.

Is a retrofit or modernisation of existing machines eligible?+

Only the part in which the interaction of legacy mechanics with new control or sensor technology was unresolved beforehand and was worked out in testing. If the modernisation follows the manufacturer documentation and a parameterisation guide, it is commissioning, not development.

Where is the line between design and development?+

What matters is whether the behaviour of the design was open before testing. Whoever applies known design rules to a new case designs. Whoever does not know whether stiffness, thermal balance or dynamics land in the target range, and only resolves this through simulation and test carrier, develops. This line runs within a project, not between projects.

How far back can machine builders apply?+

Applications for closed financial years from 2022 are possible; the assessment period of the respective year is decisive. For many machine builders the largest single amount lies precisely in these earlier years, because several development generations sit there unclaimed. Retroactive application

How much effort arises in our design department?+

For a typical mechanical engineering project we reckon with two to four hours on the part of the head of development: a technical conversation in which we take up the project, and one approval round on the finished project description. We take over the evaluation of CAD revisions, test protocols and time recording as well as both applications.

How does your documentation become an application?

Your design office and your test field already document everything the certification office wants to see, just in a different language. Here is the same set of facts in four registers, from the test rig protocol to the application sentence.

In the test field — What gets written anyway

Hybrid cell HBZ-540, series 2
Distortion 11 µm after 45 min laser
  // target ±3 µm, cooled plate not enough

- table sensors ...... 4 pcs
- compensation ....... model v3
- series 3 ........... 2.8 µm / 8 h ok

In conversation — What the team says about it

"After the laser process the table drifts away. The cooled plate did not do it, so we fitted four sensors and compensated against them, and on the third run we were below three microns."

Technically complete and worthless as application text. There is no reference to the state of the art or to the measured variable.

At the BSFZ — What is actually examined

Novelty
Technical risk
Systematic approach

For each of the three criteria the application must contain a verifiable statement. What is described is the knowledge gained on the way to the finished spindle.

In the application — What we formulate from it

"For the combination of laser cladding and milling in one setup no machine concept existed whose thermal distortion had been demonstrated below the required tolerance of ±3 µm. Whether the heat input of the laser process can be compensated far enough by design or by calculation could not be determined in advance. The project established this in three measurement series against a defined target."

This is the work. The same facts, four formulations, and only one of them is an application. This translation is our actual service: building, from what your test field produces anyway, the statement the certification office can examine. Not writing more beautifully. Writing differently.

Why do applications fail in mechanical engineering?

Rejected projects from mechanical engineering rarely fail on their technology. They fail on five recurring patterns, and we see them in self-submitted applications just as in externally assisted ones.

The state of the art is missing

The application describes what was built, not why it could not be done with what was available. Without named competitor solutions, standards or catalogue components the certification office cannot establish novelty. This is the most common reason in our own mandates.

How we do it. We specifically name the machine concepts, components and processes that were examined and rejected, each with the technical property on which they fail.

The customer order is submitted as the project

A special-purpose machine for a customer is not excluded per se, but an application covering "the plant" forces the certification office to assess order design and development together. The routine shares drag the whole project down.

How we do it. We scope the project to the part for which no validated design existed and cleanly separate order design, acceptance and documentation.

Project risk instead of technical risk

Delivery dates, budget and skills shortage are not technical risks. Anyone arguing this way describes a management problem and gets it decided as such.

How we do it. We phrase the risk as an open technical question: which property of the machine was not predictable before the start, and exactly where the project would have failed.

The prototype is described as a product

The application reads like a data sheet: performance, accuracy, cycle time. The certification office is looking for the way there, the trials that did not work and the measurement that proves the difference.

How we do it. We derive the systematic approach from your own artefacts: layout in the calculation report, trial in the test protocol, decision in the change request.

Cost side and facts do not match

The BSFZ certifies the project, the tax office checks the costs. Anyone who claims material costs for the prototype, or hours that contradict the project description, loses in the second step, after the certificate has already been issued.

How we do it. We run project description, time recording and research allowance application (ELSTER) against the same scope from the start. The cost side is not created afterwards.

If you already have a rejection, the first step is analysing the decision: which of the five points carries it. That determines whether objection or new application is the right way.

Which other industries use the research allowance?

The sector never decides eligibility — Section 2 FZulG does. These pages show the same three criteria against the activities of each field.

Automotive — Drivetrain, battery, driving function
Medical technology — Operating principle, SaMD, pre-approval
Pharma & life sciences — Compound, formulation, bioprocess
Construction — Methods, materials, component tests
Energy & cleantech — Cell, storage, grid control
Aerospace — Structure, propulsion, avionics
Agriculture & FoodTech — Process, formulation, field trial
Software & IT — Algorithm, model, architecture

Send us a drawing, not a questionnaire

In the first consultation we talk about your machine, not about your legal form. After thirty minutes you know which part of your development is eligible, in what order of magnitude, and whether a retroactive application is worthwhile. The first consultation is free; 30 % of the fee falls due with the BSFZ decision, 70 % only after the allowance has been paid out.

Book a first consultation

Official sources

Read on

Max Nodes

Founder and Managing Director of NOVARIS Consulting. Specialised in the German R&D tax credit (FZulG) since 2021; 25 of 25 accompanied BSFZ applications approved (as of 09/2026). Max Nodes