Hardware R&D runs through physical things: prototypes, test rigs, fixtures, moulds. Each one raises two separate questions. Is making it an R&D activity at all? And if it is, how does its cost enter the claim? The answers often differ for items built in the same workshop in the same month.
Is the build part of the experiment?
| What was built | Usual treatment |
|---|---|
| A prototype that is the variant being tested | Part of the core activity, to the extent the test needs it |
| A test rig, fixture or jig built so experiments can run | Usually supporting, if directly related to a core activity |
| A rig or fixture whose own design is the unknown being tested | Assessed on the ordinary test; can be core |
| Further units, or finishing a prototype for sale or service | Produces goods, so supporting only if it passes the dominant purpose test |
| A demonstration unit for investors, customers or a trade show | Typically neither core nor supporting |
| Production tooling for a settled design | Not R&D; applying known methods to make a product |
The first row is the one most easily under-claimed. When the hypothesis concerns a geometry, a material or a mechanism, building that variant is the experiment, not preparation for it. The same applies to the third row: a fixture that has to hold a part in a way nobody has managed before is tested like any other variant. The demonstration row is the one most easily over-claimed. A unit polished for a pitch is commercial, whoever built it.
Tribunals look for a design that tests something. In Active Sports Management the Tribunal found no systematic progression and no new knowledge purpose in a customised basketball shoe: the work modified existing designs to suit a player, and no records showed hypotheses being formed and tested. The Federal Court dismissed the appeal in 2024. A record made before each build, saying what the build is meant to find out and what result would count as failure, is what separates a prototype from a product revision.
How the cost is claimed
Being an R&D activity does not mean every dollar is a deduction in the year it is spent. The ATO draws three lines that matter here. Which side of them an item falls on depends on the facts: whether it lasts beyond the test and is held for use, is used up, or is later sold or kept in service. The same mould insert can be consumed in one program and held as a tool in another.
- Tools and rigs held as assets. Expenditure included in the cost of a tangible depreciating asset is not a notional R&D deduction. Its decline in value is claimed instead, to the extent the asset is used for R&D.
- The people who build them. Section 355-225 excludes salary expenditure incurred on constructing depreciating assets, so building time is not claimed as R&D salary. It generally forms part of the asset's cost instead.
- Prototypes consumed by testing. Materials and labour used up in a test, and not held as an asset, are generally expenditure on the activity in the usual way.
If a prototype is later sold or put to use, the feedstock rules can claw back the incentive on the inputs it consumed. Treatment of particular items is for the company's accountant or tax adviser to confirm. See notional deductions.
Neighbouring questions
Building larger equipment to test a process is covered on pilot plants, and running a process at production scale on manufacturing scale-up. Software prototypes follow the same principle with different traps, set out on MVPs and prototypes. The robotics page applies these rules to a whole sector.