Up to 43.5% back as cash for a company under $20 million turnover — and one deadline that cannot be extended.
For an Australian startup that is losing money and spending it on engineering, the R&D Tax Incentive is usually the largest single source of non-dilutive funding available. It is a refundable offset, which means a company with no tax to pay gets it as cash rather than as a credit against future profits.
It is also the programme most often lost on administration rather than on eligibility. The registration deadline is absolute, the record-keeping requirement is contemporaneous, and the definition of eligible work is narrower than most founders assume.
General information, not advice. CatenaHQ is software. These guides are not legal, financial, tax or investment advice, and none of them takes account of your circumstances. Rates and thresholds are from the ATO and current as at October 2026. Eligibility of specific activities is a technical question — get it assessed.
The rate is expressed as your company tax rate plus a premium, which is why you see different headline numbers quoted.
Your company tax rate plus an 18.5% premium. For a base rate entity taxed at 25%, that is 43.5%, and it is refundable — paid as cash to a company in loss.
Available to entities with aggregated turnover under $20 million that are not controlled by income tax exempt entities.
Company tax rate plus a premium that steps with R&D intensity:
8.5% on R&D expenditure up to 2% of total expenditure, and 16.5% on expenditure above that 2% threshold. Non-refundable: it reduces tax payable and carries forward.
$20,000 of eligible R&D expenditure in the income year, unless the work is done through a registered Research Service Provider.
Above $150 million of notional deductions in an income year, the offset drops back to the plain company tax rate. Not a concern for most, but it exists.
You must register your R&D activities with AusIndustry within 10 months of the end of your income year. For a 30 June year end that is 30 April. The deadline is absolute — there are no extensions.
Registration comes first and the tax claim follows. Miss 30 April and the entire year's claim is gone, regardless of how eligible the work was or how good the records are. This is the single most common way the incentive is lost, and it has nothing to do with the merits.
It is worth putting the date in the compliance calendar the day the financial year ends, not the month before it falls due.
The programme distinguishes core R&D activities — experimental work whose outcome cannot be known in advance without applying a systematic progression from hypothesis to experiment to observation to conclusion, conducted to generate new knowledge — from supporting activities, which are directly related to core activities.
The distinction that catches software companies is between building something difficult and resolving genuine technical uncertainty. Writing a complicated feature using known techniques is engineering, not R&D. Testing a hypothesis whose outcome a competent professional in the field could not have predicted is R&D. The record has to show the uncertainty, the hypothesis and the experiment, and it has to have been written at the time rather than reconstructed at registration.
Significant changes were announced in the 2026 Budget. They are announced measures, not yet law, and all of them apply from 1 July 2028 — the 2028–29 income year. Nothing below changes a claim you are making now, but two of them change how you would plan:
The one worth watching is supporting activities. For a software company a meaningful share of a typical claim sits there, and a company whose claim depends heavily on supporting activities should understand what its claim looks like without them before 2028–29 arrives.
Contemporaneous means written while the work happened. A narrative assembled in April to describe the previous July is the weakest possible position in a review.
"This was hard and took a long time" is not the test. The test is whether the outcome could have been known in advance by a competent professional.
Ten months, no extensions, whole year gone. It is the most expensive calendar entry a startup can fail to make.
Time allocation across staff, contractors and overheads has to be defensible. "Roughly 70% of the dev team" is not an apportionment method.
It is a self-assessed programme subject to review. Budgeting the cash before the claim is lodged, and spending it before it arrives, has sunk companies.
The $20 million test is aggregated, counting connected and affiliated entities. A founder with other businesses can fail it without realising.
CatenaHQ tracks R&D projects and expenditure through the year rather than at the end of it, with the registration deadline sitting in the same compliance calendar as everything else that has a date. The point is not the arithmetic — your accountant does that — it is that the activity records exist with the dates they were actually written.
An R&D claim is also a question every investor asks in due diligence, because a refundable offset is a real and recurring line in the cash forecast. Being able to show the registrations, the activity records and the amounts received is part of being raise-ready.
Finding the programs you qualify for and keeping the evidence as the work happens is what the grant discovery module is for.
All of the CatenaHQ guides are written the same way: from the Act and the regulations, with the section numbers shown.
The other tax programme worth knowing: 20% back for your investors, and R&D expenditure is one of the things that scores points on the innovation test.
The aggregated turnover test counts connected entities. Knowing who owns what, across which companies, starts with the register.