A guide for current and prospective RDTI claimants navigating the proposed changes
If you run an R&D program in Australia, the 2026–27 Federal Budget handed down on 12 May 2026 is worth your attention. Nothing changes for this year's claim, but the Budget signals the most significant restructure of the R&D Tax Incentive (RDTI) since 2020. Buried in the papers is a proposal that would fundamentally narrow what counts as eligible R&D expenditure, and for many claimants (particularly in software, hardware, biotech, medtech and advanced manufacturing) it's a change worth planning for now rather than in 2028.
What was actually announced
The headline measure is the removal of eligibility for supporting R&D activities. Under the current rules, expenditure on activities undertaken for the dominant purpose of supporting a core R&D activity (things like building test rigs, running production trials, undertaking literature reviews, or maintaining specialist equipment) can be claimed alongside the core experimental work itself, at the same offset rate as core activities. Under the proposed framework, only core R&D activities (those involving a genuine experiment designed to test a hypothesis through a systematic progression of work) would remain eligible at all.
To offset the narrower scope, the Government has proposed a 4.5 percentage point increase to the offset rate for core R&D expenditure. Depending on a company's circumstances, this lifts the tax benefit on core spend by roughly 25 to 50%, with the refundable offset potentially reaching 48% for eligible companies, and sometimes as high as 53%! The intensity threshold for the premium non-refundable offset would also drop from 2% to 1.5% of total expenditure, making the higher rate accessible to more claimants.
A third proposed change is significant enough to deserve its own discussion: refundability of the offset would be limited to a company's first 10 income years. This isn't a minor mechanical adjustment, and it's covered in detail further below.
Two things are important to hold onto here. First, this is proposed to take effect from 1 July 2028, so there is no change to FY26, FY27 or FY28 claims. Second, and just as importantly, none of this is law yet. It's a Budget announcement, not legislation. Exposure draft legislation and a consultation process are expected to follow, and the detail (including exactly how the core and supporting boundary will be drawn) could still shift materially before it's enacted.
Why this matters more for some claimants than others
The impact of this change is not evenly spread. For businesses whose R&D is genuinely concentrated in core experimental work, with a tight hypothesis, test and evaluate cycle and minimal supporting infrastructure, the higher offset rate is a straightforward win.
But for many real-world claimants, supporting activities are not a peripheral add-on; they're a material share of the claim. There is no official ATO or Department of Industry, Science and Resources (DISR) breakdown of core versus supporting spend by industry, but the pattern practitioners consistently report is that supporting activity commonly represents a sizeable minority, and sometimes close to half, of total R&D spend in the sectors below:
• Software and SaaS: building test environments, provisioning infrastructure, and data preparation and integration work that enables the experimental development but isn't itself the experiment. Advisers commonly see supporting spend sitting anywhere from around 20% to 40% of the claim, depending on how much infrastructure and tooling work sits around the core build.
• Hardware and advanced manufacturing: producing prototypes and pilot production runs, calibrating and maintaining specialist equipment, and process-scale trials alongside the core design experiments. Supporting spend in this category can run higher again, often 30% to 50% of the claim, given how equipment and production heavy these programs tend to be.
• Biotech and medtech: literature reviews, assay development, sourcing and preparing materials, and regulatory or quality documentation supporting the experimental program without constituting the experiment itself. Supporting spend here is frequently in the 20% to 35% range, though highly regulated programs can push higher.
These figures are illustrative ranges drawn from adviser commentary and worked examples, not official statistics, since core versus supporting spend isn't separately published in the ATO's transparency reports. What is consistent across the commentary is a rough breakeven point: analysis from legal and advisory firms suggests the higher core rate only fully compensates claimants where supporting activity sits at around 10% or less of total R&D expenditure. For many programs in the sectors above, supporting spend runs well beyond that. If the measure proceeds in its current form, those businesses could see a lower total benefit from FY29 onward, even with a higher headline rate on the activity that remains eligible.
Getting clear on what actually counts as core
Once supporting activities fall out of scope, the core activity definition stops being one part of the eligibility test and becomes the whole test. It's worth being precise about the boundary now, well before it becomes the difference between a claim being eligible and not eligible at all. A core R&D activity is an experiment: it starts from a hypothesis that can't be resolved using existing knowledge, proceeds through a systematic, structured test of that hypothesis, and ends in observation, evaluation and a logical conclusion. Work that is planned, executed and evaluated with that structure is core. Work that enables or surrounds that structure, without itself being the test, is supporting. Work that involves no genuine technical uncertainty at all was never eligible and remains that way regardless of these reforms.
The table below works through a typical software R&D program (building a new recommendation engine) to illustrate where that boundary sits in practice. The core rows are deliberately broken down step by step through the hypothesis, experiment, observation and evaluation cycle, since that structure, not the label on the task, is what determines whether a given piece of work is actually core.
| Task | Classification | Why |
|---|---|---|
| Identifying a technical problem where the outcome can't be determined in advance from existing knowledge, e.g. whether a new ranking model can lift click‑through rate without degrading response latency | Core | Establishes the genuine technical uncertainty that a core activity must be directed at resolving |
| Formulating a specific, testable hypothesis about how to solve it, e.g. hypothesising that a graph‑based ranking model will outperform the existing collaborative‑filtering approach for cold‑start users | Core | A concrete, falsifiable hypothesis is the starting point the legislation requires |
| Designing the experiment or test protocol in advance, including control and variant groups, metrics and success criteria, before any testing begins | Core | Defining the test method upfront is part of the required systematic, structured approach |
| Building or configuring the experimental model or system variant(s) specifically constructed to test the hypothesis | Core | Directly implements the experiment; distinct from general‑purpose environment or tooling build |
| Running the controlled experiment or test under the defined protocol, e.g. executing the A/B test between the baseline and the new ranking model | Core | The experiment itself; this is the activity the whole test is built around |
| Observing and measuring results against the hypothesis and pre‑defined success criteria | Core | Observation of the experiment's outcome is an explicit step in the statutory test |
| Analysing results, drawing logical conclusions, and evaluating whether the hypothesis was confirmed, partially confirmed or rejected | Core | Evaluation and conclusion is the final required step of the hypothesis‑test‑evaluate cycle |
| Where results are inconclusive or negative, revising the hypothesis and repeating the cycle with a new test | Core | Iteration remains core provided each cycle still involves a genuine hypothesis and test |
| Building and configuring the test environment used to run the experiments | Supporting | Enables the experiment but isn't itself a test of a hypothesis |
| Preparing, cleaning and labelling data sets used in the experiments | Supporting | A necessary input to the experiment, not the experiment itself |
| Building internal dashboards or tooling to capture and log experiment results | Supporting | Supports measurement of the core activity's outcomes |
| Conducting a literature or prior‑art review before experimentation begins | Supporting | Establishes the technical knowledge gap but doesn't test a hypothesis itself |
| Fixing bugs using known, established solutions | Not eligible | No technical uncertainty; existing knowledge already resolves the issue |
| Building UI screens using established design patterns | Not eligible | Routine software development; not experimental in nature |
| Ongoing production support and patching after release | Not eligible | Business‑as‑usual maintenance work |
| Project management and business analysis activities | Not eligible | Administrative in nature, with no experimental content |
The pattern holds across sectors even though the specific tasks change: hardware and manufacturing claimants will see the same three-way split between design experiments (core), prototyping and pilot production (supporting), and routine production runs (not eligible); biotech and medtech claimants will see it between the actual experiments or trials (core), the literature reviews and assay preparation around them (supporting), and standard regulatory administration (not eligible). Getting comfortable classifying tasks this way now, at a task level and in real time, is the single most useful thing a claimant can do ahead of July 2028.
The nature of the risk is changing, not just the size of it
It's worth being precise about what's shifting. Today, whether an activity is classified as core or supporting doesn't change the benefit rate; both are eligible at the same offset. The classification matters for compliance and self-assessment purposes, but a misclassification today doesn't, by itself, reduce the cash benefit. From 1 July 2028, that changes completely: only core activities are eligible at all, so a supporting activity is no longer eligible at a different rate, it's simply not eligible. The consequence of getting the core and supporting boundary wrong is moving from a compliance issue to a straight loss of benefit.
The overcorrection risk: don't manufacture "core" activities
The natural response to this kind of change is to look at the activities currently claimed as supporting and ask whether they could instead be framed as core. That instinct is understandable, but it carries real risk, and the compliance environment makes it a risk worth taking seriously right now, not just from 2028.
A core R&D activity requires a genuine hypothesis, tested through a systematic progression of work from hypothesis to experiment to observation and evaluation, leading to logical conclusions, and undertaken for the purpose of generating new knowledge. Relabelling routine engineering, configuration, scale-up or production work as "core" without that underlying experimental logic doesn't change its substance. It just creates a claim that won't withstand scrutiny.
And scrutiny is intensifying well ahead of the 2028 start date. The Government has directed the ATO to undertake additional targeted compliance work on RDTI claims, tax agents and advisers are being asked to reinforce eligibility education with clients, and the DISR, which administers R&D activity registration and activity eligibility (the function previously associated with the AusIndustry name), has been conducting more formal reviews of higher-risk registrations, particularly where technical narratives, expenditure treatment or supporting records look inconsistent. Claimants who stretch definitions now, in anticipation of 2028, are more likely to attract review attention in the interim, not less.
The 10-year refundability limit: the other major change, and for some claimants the bigger one
The removal of supporting activity eligibility has understandably dominated the headlines, but the proposal to limit refundability to a company's first 10 income years is at least as significant for the businesses it affects, and arguably a more immediate threat to cashflow. It deserves equal attention when you're assessing what this Budget means for your business.
Under the current rules, eligible companies below the turnover threshold receive their R&D offset as a cash refund, regardless of how long they've been claiming. That refund is what allows loss-making R&D companies to convert R&D tax losses into cash today, rather than a deduction they can only use once profitable. Under the proposed changes, once a company passes 10 income years, it would still be entitled to the same higher core offset rate, but only as a non-refundable offset. That offset can only reduce tax payable or be carried forward to future years; it can no longer be paid out as cash. The turnover threshold for refundability would also rise to $50m, which helps some claimants, but the 10-year limit works against that improvement for anyone who has been in the program for a while.
The businesses most exposed are exactly the ones the RDTI was designed to support through the hardest part of the innovation cycle: R&D-intensive companies with long development and commercialisation timelines. Biotech and medtech companies routinely run development programs, including clinical trials and regulatory approval pathways, that extend well past 10 years from first R&D activity to any meaningful revenue. Deep tech, advanced manufacturing and hardware businesses with long qualification and scale-up cycles face a similar problem. For these companies, staying in tax losses past the 10-year mark isn't a sign of a failing business; it's often simply how long genuine, high-risk R&D takes to commercialise. Under the proposed rule, exactly these companies would lose access to the cash that has, for many of them, funded the next stage of the R&D program.
The practical effect is a timing and value problem, not just an eligibility one. A dollar of offset paid as cash this year can be reinvested in R&D immediately. The same dollar, available only as a non-refundable offset carried forward until the company is tax payable, could be worth substantially less in present-value terms, particularly for a company that might not be tax payable for several more years. Worked examples from tax advisers illustrate the point: a company with identical R&D spend before and after its 10-year mark can see its offset swing from a cash refund in the tens or hundreds of thousands of dollars to the same nominal amount locked up as a carry-forward offset, usable only once profitable and subject to the ordinary loss and offset utilisation rules.
There is also real uncertainty in the detail. Advisers have flagged that it isn't yet settled whether the 10-year clock runs from incorporation, from first R&D registration, or from some other trigger, and whether it's measured at the individual entity level or across a consolidated group. Each of those design choices materially changes who is caught and when. Businesses approaching the 10-year mark, or structured as part of a group, should treat this as a genuine open question rather than assume the most favourable interpretation.
For claimants who are both past the 10-year mark and carry meaningful supporting-activity spend, the two proposed changes compound. Not only would the supporting portion of the claim fall away entirely, but the core portion that remains would also lose its refundable cash value. Modelling both changes together, rather than in isolation, is the only way to get an accurate picture of the FY29 impact.
What claimants should be doing now
With the change still two Budgets and a legislative process away, there's no need to restructure a claim overnight. But there is real value in using the next 18 to 24 months to get the underlying evidence and claim architecture in better shape. Three things stand out:
• Classify tasks against the three-way test now, not just at claim time. Use a core, supporting, or not-eligible classification (see the worked example above) for each task on your R&D projects as the work happens, rather than reconstructing it in hindsight. This is the practical foundation for everything else on this list.
• Tighten task-level record-keeping. Make sure time and cost records can isolate core-only activity by resource, project and task, not just at the project level. If the eligible base narrows to core activities alone, claims that can already cleanly separate core from supporting spend will be far easier to substantiate and defend, regardless of when, or whether, the reform proceeds in its current form.
• Strengthen contemporaneous evidence of the hypothesis-test-evaluate cycle. Documentation created at the time, such as hypotheses set out before experiments run, records of what was tested and observed, and how conclusions were reached, is consistently what regulators say separates strong claims from weak ones. Retrospective, year-end narratives are the first thing reviewers push back on.
• Review current claim structures ahead of FY29, modelling both changes together. Map out what proportion of your current claim sits in supporting activities, and separately check where your company sits relative to the 10-year mark (and how it's structured, if part of a group). Model the combined effect of losing supporting-activity eligibility and, where relevant, losing refundability, at the proposed higher core rate. For some businesses this modelling will show a net benefit; for others, particularly long-cycle, R&D-intensive businesses past the 10-year mark, a materially reduced one. Either way, knowing the number now gives you time to adjust R&D planning, resourcing, funding strategy or even structuring decisions well before the rules take effect.
A note on where things stand
It bears repeating: this is a Budget announcement, not legislation. The Government has indicated a consultation process will follow, likely including exposure draft legislation that stakeholders can respond to, and industry bodies have already flagged concerns about the pace and scope of the change relative to the more incremental adjustments recommended by the recent Strategic Examination of R&D. The final design, including exactly how "core" and "supporting" activities are defined, and whether transitional rules are introduced, could look different by the time it's enacted.
That uncertainty is a reason to stay engaged, not a reason to wait. The businesses best placed to adapt, whichever way the detail lands, will be the ones already running disciplined, well-evidenced claims. If you'd like help assessing how your current claim would fare under a core-only regime, or want a second set of eyes on your documentation practices ahead of the transition, now is a good time to start that conversation.