Picture a self-managed super fund that owns the family farm outright — a common structure for Australians who want their retirement savings and their livelihood under one roof. Under the tax the government announced in the May 2023 Budget, a good season and a rising land valuation could have generated a tax bill the following year, even though not one dollar of that gain had reached the farmer's bank account. The bill would have been calculated off a number on a valuer's page, not cash in hand. If the land's value fell back the following year, there was no cash refund — only a credit carried forward against future gains. That specific tax never actually landed on anyone. It was rewritten before it commenced. But for more than two years it was the law the government was trying to pass through Parliament, and the story of how close it came — and why it was eventually pulled back — reveals more about the risks in this tax than the version that finally took effect.
The problem was never really the tax rate. It was the mismatch between when the bill would fall due and when the cash to pay it existed. A farm, unlike a share portfolio, cannot be sold in ten-thousand-dollar slices. If a single year's valuation gain had pushed the fund's tax bill higher than the cash sitting in its account, the trustees would have faced a choice between borrowing against the property, drawing down other assets inside the fund, or selling land they had no intention of selling, to pay tax on a gain that existed only until the next valuation. That mismatch, more than the headline rate, is why accountants, actuaries and self-managed fund trustees spent two years arguing the original design was unworkable, not merely unwelcome.
That version, known as Division 296, became law on 1 July 2026. It imposes an extra tax on the earnings attached to very large superannuation balances. Understood narrowly, it affects a small number of very wealthy Australians and raises a legitimate question about a genuinely costly tax concession. Understood as a precedent, it is the closest Australia's tax system has come to taxing wealth that exists only on paper — and the $3 million line it starts at is a lot less permanent than it sounds.
What the government did: a new kind of tax for superannuation
Superannuation earnings are ordinarily taxed inside the fund at a concessional rate of 15% — well below most people's marginal income-tax rate, which is the entire point of the concession: it rewards people for locking savings away until retirement. In the 2023–24 Budget (9 May 2023), the Treasurer announced "Better Targeted Superannuation Concessions" — a new, additional 15% tax on the portion of an individual's earnings attributable to a Total Superannuation Balance (TSB) above $3 million, lifting the effective rate on that slice to 30%. Treasury opened consultation on draft legislation on 3 October 2023 and closed it on 18 October — a twelve-working-day window that professional bodies including the Tax Institute criticised at the time as too short for a measure this novel (Treasury exposure draft consultation, Oct-2023).
The concession itself exists for a simple reason: without a lower tax rate inside super, most people would be no better off saving there than saving the same amount outside it, which would undermine the point of running a compulsory retirement system at all. The difficulty is that the same flat 15% rate applies whether a balance is $50,000 or $50 million, so every extra dollar earned inside a very large balance gets the same concession as the first dollar saved by someone in their twenties. Treasury's stated justification for drawing a new line at $3 million was that balances above it are large enough, and rare enough, that an additional tax there would not touch the ordinary saver the concession is meant to help (Treasury, Better Targeted Superannuation Concessions fact sheet, May-2023, approx.).
The design choice that mattered most was how "earnings" was defined. Rather than taxing income the fund actually received — dividends, interest, rent, realised capital gains — the original formula measured the year-on-year movement in a member's Total Superannuation Balance. That measure rises and falls with the market value of everything the fund holds, sold or not. A share portfolio that went up on paper, a commercial property that was revalued, a farm that appreciated — all of it counted as "earnings" for this tax, whether or not the fund had actually sold anything and received cash. This is what became known, accurately, as a tax on unrealised gains: a first for the Australian tax system, which has otherwise always waited for an asset to be sold before taxing the gain on it (ITAA 1997, existing CGT event framework; Treasury exposure-draft Explanatory Materials, Oct-2023).
The second design choice was that the $3 million threshold would not be indexed. Every other major superannuation threshold in the system — the transfer balance cap, the concessional and non-concessional contribution caps — moves periodically in line with inflation. This one was drafted to stay fixed in dollar terms indefinitely, a deliberate choice the government did not attempt to hide (Treasury Laws Amendment (Better Targeted Superannuation Concessions and Other Measures) Bill 2023, as introduced, Nov-2023).
The Treasury Laws Amendment (Better Targeted Superannuation Concessions and Other Measures) Bill 2023 and its companion Imposition Bill were introduced into the House of Representatives in November 2023 and passed the House unamended. They then stalled in the Senate for close to two years. The government did not control the Senate outright and needed either the Coalition or the Greens, plus the crossbench, to pass the Bills. The Coalition opposed the unrealised-gains mechanism outright; the Greens, rather than opposing it, pushed to make it more aggressive — reportedly seeking a lower $2 million threshold and unrelated concessions as the price of their support (approx.; Senate Economics Legislation Committee proceedings and reporting, 2024). Neither position gave the government a path to the votes it needed, and the Bills spent 2024 and into 2025 without a Senate vote, including a Senate Economics Legislation Committee hearing on 18 April 2024 at which superannuation and tax bodies gave evidence against the design.
The Senate agreed in December 2024 to split the package into two separate Bills, isolating the contested Division 296 measure from unrelated superannuation provisions so the uncontroversial parts could proceed while the argument over unrealised gains continued (Senate, procedural motion, Dec-2024, approx.). Industry analysis through this period argued that self-managed super funds — which hold a disproportionate share of the country's largest balances, and are far more likely than APRA-regulated funds to hold illiquid, hard-to-value assets such as property and private business interests — would bear the brunt of any forced-sale problem, even though they represented a minority of the accounts Treasury expected the measure to touch (SMSF Association and industry commentary, 2024, approx.). The Bills remained unresolved through the May 2025 federal election, after which the government returned to office still without control of the Senate in its own right.
On 13 October 2025, the Treasurer announced a substantial redesign. The unrealised-gains formula was dropped in favour of taxing realised earnings only — dividends, interest, rent and realised capital gains, measured against a new notional cost base set at market value on 30 June 2026. A second, higher tier was added for balances above $10 million, taxed at an extra 25% (a 40% total rate on that slice), against the extra 15% (30% total) that continued to apply between $3 million and $10 million. Critically, both thresholds would now be indexed — in $150,000 increments for the $3 million line and $500,000 increments for the $10 million line, tracking the same broad method used for the transfer balance cap. The start date moved a year, to 1 July 2026 (Treasurer the Hon Jim Chalmers MP, media statement, 13-Oct-2025).
The rewritten Bills — the Treasury Laws Amendment (Building a Stronger and Fairer Super System) Bill 2026 and the companion Imposition Bill — passed the House of Representatives on 5 March 2026, passed the Senate on 10 March 2026, and received Royal Assent on 13 March 2026. Division 296 tax is now law, applying from the 2026–27 income year, with the first assessments expected once that year's fund earnings are finalised.
The strongest case for the policy
Put the government's defence at its best
The concession on very large superannuation balances is genuinely costly to the budget, and it is genuinely regressive. Every dollar of foregone tax on earnings inside a $5 million or $20 million balance is a dollar of tax expenditure that has to be made up somewhere else — by other taxpayers, by lower services, or by future borrowing. Superannuation's tax concessions exist to reward ordinary Australians for saving toward a comfortable retirement, not to shelter estates that will never be run down. A member who has already amassed several times the balance a comfortable retirement requires does not need the same 15% concessional rate applied to every extra dollar of earnings as someone building their first $200,000. On that reading, an additional tax on the earnings above $3 million is not an attack on the retirement system — it is a repair to a concession that has drifted a long way from its original purpose, targeted narrowly enough that it reaches, by Treasury's own estimate, roughly 80,000 people, or about 0.5% of Australians with a superannuation account, in its first year (approx.; Treasury estimate, 2025). Measured against that small, high-balance group, the case for asking them to pay a higher rate on earnings above an already generous threshold is not an unreasonable one, and the government can fairly say it listened when the design was shown to be flawed and rewrote it before it became law. Independent analysts go further still: the Grattan Institute has estimated the combined value of superannuation's tax concessions on contributions and earnings at tens of billions of dollars a year (approx.; Grattan Institute, 2023–2025), skewed heavily toward the largest balances, and has argued that cost is on track to exceed the age pension. On that view, Division 296 is not an aggressive intervention at all — some economists would say the government has ended up too cautious, not too bold.
For two years, the government was trying to pass a tax calculated off a number on a valuer's page, not cash in a bank account.
The evidence: what nearly happened, and what actually becomes law
The gap between the 2023 design and the 2026 law is the whole story of this topic, and it is worth setting out in numbers.
The 2023 design: earnings meant paper gains
Under the original formula, "earnings" for Division 296 purposes was the increase in a member's Total Superannuation Balance over the year, adjusted for contributions and withdrawals. Because that balance is a mark-to-market figure, any unrealised appreciation in the fund's assets was captured as taxable earnings, in the same year it occurred, whether the fund had sold anything or not. For funds holding illiquid assets — a farm, a commercial building, a private business, all common inside self-managed super funds — a large valuation gain in one year could mean a tax bill payable in cash the following year, with no guarantee the asset that drove the gain could be sold, or sold at a fair price, to fund it. The Tax Institute put the underlying objection bluntly: the design set what it called a legal precedent that Australians could be taxed on money they do not have and may never have (The Tax Institute, submission and commentary, 2023–2025).
Two years in Parliament
The Bills passed the House of Representatives in 2023 but could not secure Senate passage through 2024 and into 2025. The Senate agreed to split the Division 296 measure out from unrelated Bill provisions in late 2024 to isolate the contested part, and the Economics Legislation Committee heard evidence against the unrealised-gains mechanism from superannuation and tax professional bodies over successive hearings. Industry estimates during this period suggested the number of self-managed funds likely to be affected was materially larger than Treasury's original modelling implied, because SMSF balances are more concentrated at the top of the distribution than the broader APRA-regulated fund sector (industry commentary, 2024, approx.). The Bills remained unresolved through the 2025 federal election.
The October 2025 rewrite — and what actually becomes law
The redesigned Division 296 taxes only realised earnings: dividends, interest, rent, and realised capital gains net of realised losses, measured against a notional cost base fixed at each asset's market value on 30 June 2026. Unrealised gains are excluded entirely. A worked illustration shows how the two-tier structure applies: a member with a $15 million balance and $900,000 of realised earnings in a year would have roughly $7 million of that balance sitting in the $3–10 million band and $5 million above $10 million. Split proportionally, about $420,000 of the year's earnings falls in the lower band, taxed at the extra 15% (about $63,000), and about $300,000 falls in the upper band, taxed at the extra 25% (about $75,000) — a combined Division 296 bill of roughly $138,000 (illustrative worked example based on published Treasury/ATO methodology, approx.), on top of the 15% the fund already pays on that income. Both thresholds now index in line with the transfer balance cap methodology, in $150,000 steps for the $3 million line and $500,000 steps for the $10 million line, so the dollar lines will move upward over time rather than staying fixed.
The redesign also fixed the payment mechanism, not just the tax base. Members can pay a Division 296 assessment personally or elect, within 60 days of assessment, for their fund to release the money and pay the Australian Taxation Office directly — the same release-authority process already used for Division 293, the separate additional tax on high-income earners' concessional contributions (ATO, Division 296 administration guidance, 2025–2026). That mechanism genuinely solves the cash-flow problem for a realised-earnings tax, because realised earnings are, by definition, cash the fund has actually received. It would have done nothing to solve the cash-flow problem under the original design, because a fund cannot release cash from a valuation gain it has not banked — which is exactly the mismatch the farm example at the top of this article illustrates, and exactly why the release-authority fix only works because the unrealised-gains mechanism was dropped first.
The precedent test: how other thresholds have moved
Division 296 is not the first dollar-figure threshold in the superannuation system, and its own thresholds are explicitly indexed using the same method as an existing one: the transfer balance cap, introduced on 1 July 2017 at $1.6 million to limit how much a member could move into a tax-free retirement pension. That cap has already been indexed three times — to $1.7 million in 2021, $1.9 million in 2023, and $2 million from the 2025–26 year — moving in $100,000 steps tied to the Consumer Price Index (ATO, transfer balance cap indexation history, 2017–2025). Over those eight years the cap rose by 25%, broadly tracking cumulative inflation. But SMSF advisers have long observed that members with strongly performing funds continue to bump against, and exceed, their personal transfer balance cap despite that indexation, because fund earnings in a typical year routinely outpace CPI, particularly when equity and property markets perform well (industry commentary, approx.). A CPI-indexed threshold protects against inflation eroding its real value. It does nothing to stop a threshold being overtaken by a balance growing for reasons that have nothing to do with the cost of living — investment returns, and, for Division 296's $3 million and $10 million lines, the compounding effect of decades of compulsory contributions. The same dynamic that has pushed a meaningful number of retirees over the transfer balance cap despite indexation is available, in slower motion, to Division 296.
How large a group, eventually
Treasury's ≈80,000-person, ≈0.5% estimate describes the tax's first year only. Two forces work on it every year after that: existing large balances keep compounding, and new balances keep arriving at the top, as savers who have spent an entire career under compulsory superannuation reach retirement with balances well above what earlier cohorts accumulated — helped along by a Superannuation Guarantee rate that has itself risen over time, from 9% when it was introduced in the early 1990s to 12% from 1 July 2025 (approx.). No official long-run projection of the number of Australians Division 296 will eventually capture has been published as a single, citable figure. What is on the record is an acknowledgement, from Treasury and echoed by independent commentators, that the affected population is expected to grow from its first-year base as balances compound and new savers move through a system that has now been compulsory, and growing, for more than three decades (approx.; Treasury and industry commentary, 2025). A threshold billed today as reaching only the wealthiest half of one per cent of account holders is not a promise about who it reaches in twenty years.
One design feature does cut the other way, and fairness requires stating it plainly: Division 296, like the transfer balance cap before it, is assessed on each individual's own Total Superannuation Balance, not on a couple's combined savings. A husband and wife can each hold close to $3 million before either of them is touched by the tax, meaning a household can accumulate close to $6 million across two accounts and still sit under the line. Couples who split contributions and balances evenly between spouses over a working life — a strategy financial advisers already recommend for other reasons — are, in effect, given twice the room of a single member. That structural feature will keep a meaningful number of larger households out of Division 296 altogether, and it is a genuine, if quiet, limit on how far the tax reaches even as balances grow.
The critique: what the backdown does not fix
The redesign removed the two features critics attacked hardest — taxation of unrealised gains, and a threshold with no indexation at all — and it deserves to be assessed on the law that actually exists, not the one that never commenced. But three problems survive the rewrite.
The first is precedent. For more than two years, an Australian government drafted, consulted on, introduced, and pushed through the House of Representatives a tax base built from the change in an asset's paper value, applied to superannuation — the one part of the tax system built on the promise that money inside it is locked away and taxed lightly precisely because it cannot be touched until retirement. That the mechanism was eventually abandoned does not erase the fact that it was seriously pursued, that it had the numbers in the House, and that only a Senate the government did not control forced the retreat. The administrative machinery needed to measure a member's Total Superannuation Balance movement year on year already exists inside Division 296, because that same TSB concept still determines which tier a member sits in. A future government facing a revenue shortfall does not need to invent a new mechanism to revive the unrealised-gains formula; it needs only to change which figure the tax is calculated against, inside a Division that already exists in the Act.
Part of what made the 2023 design a live precedent, rather than a technical footnote, is that nothing directly comparable exists elsewhere in Australia's tax system. Capital gains tax, in every other context, waits for an asset to be sold. Land tax and council rates are levied on holding an asset, not on its year-to-year change in value. A tax that moved with the mark-to-market swing in an asset's value, inside the one part of the system built on the promise of concessional, locked-away savings, had no direct predecessor to draw on for how it would behave in a downturn, a valuation dispute, or a genuinely illiquid holding — and industry submissions argued nothing directly comparable exists in the retirement systems of Australia's usual peer comparators either (approx.; Tax Institute submission, 2023–2025). That absence of precedent was itself evidence the design was untested, not merely unpopular.
The second is that, as the transfer balance cap history above shows, indexation slows bracket creep without stopping it. The $150,000 and $500,000 step increments are tied to CPI movement, but superannuation balances do not grow on CPI alone — they grow through compulsory contributions stacked on top of investment returns that, over long periods, have historically outpaced inflation by a real margin. Indexation changes the slope of the line in the chart above; on the transfer balance cap's own eight-year record, it does not flatten it.
A smaller, more technical problem sits inside the notional cost base itself. Because Division 296 only measures gains realised after each asset's value is fixed at 30 June 2026, all of the appreciation that happened before that date — potentially decades of it, for a long-held asset — escapes the tax entirely. Growth after commencement is treated differently from growth before it, purely because of when it occurred, not because of anything about the asset or the member. That is a reasonable, even generous, way to start a new tax without reaching back into the past. It also means the measure's revenue, and its bite on any individual member, will build slowly over years as post-2026 gains accumulate, rather than applying in full from day one — a detail easy to lose in a debate that has mostly been about who the tax reaches, not when it actually starts reaching them.
A backdown is not a guarantee. It is one Parliament's choice, and the next Parliament can choose again.
The third is the cost of the uncertainty itself. Self-managed super fund trustees, financial advisers and tax practitioners spent more than two years planning around a tax that, in its announced form, never actually applied to anyone. Advice given, restructuring undertaken, and asset decisions made between 2023 and 2025 in anticipation of an unrealised-gains tax that was ultimately abandoned represent a real, if unquantified, cost of poor initial policy design — a twelve-working-day consultation window on a genuinely novel tax base, followed by two years of legislative limbo (approx.; Treasury consultation timeline, Oct-2023; industry commentary, 2023–2025).
What they got right, and the caveats
Credit is due where it is due. The government held two open, publicly scrutinised design flaws — an unrealised-gains base and a permanently frozen threshold — for over two years, and when the political arithmetic finally forced its hand, it did not simply abandon the measure; it rewrote the mechanism to remove both flaws before the tax became law. That is a genuine course-correction, driven by sustained, evidence-based criticism from professional and industry bodies working through the ordinary parliamentary process, which is exactly how the system is supposed to work. The final design is also closer to established tax principle than the original: taxing realised income, at a still-concessional rate even for the largest balances, is a materially different proposition from taxing paper gains. The process itself also improved: where the 2023 exposure draft was consulted on for twelve working days, the redesigned Bills followed a longer period of engagement with industry and professional bodies through late 2025 before being reintroduced in early 2026 (approx.; industry commentary, 2025–2026) — a small but real sign that the criticism of the original process, and not just the original design, was heard.
The underlying case for some reform also survives scrutiny. Superannuation was designed, and has repeatedly been described by independent reviews going back to the 2010 Cooper Review and the 2020 Retirement Income Review, as a vehicle for funding a dignified retirement — not as a low-tax structure for growing an estate indefinitely. A balance well beyond what any retirement plausibly requires drawing an extra concession every year is a legitimate target for reform, on the numbers alone. The caveat is that the final design still relies on an entirely new administrative concept — a notional Division 296 cost base, tracked separately from each asset's ordinary CGT cost base, reset at 30 June 2026 — adding a layer of complexity to fund administration that did not exist before, even without the unrealised-gains element that made the original version unworkable for illiquid holdings.
It is also worth recording that the redesign has critics on the other side of the argument. The Grattan Institute, among others, has argued for years that superannuation's largest concessions should be trimmed further and faster than Division 296 does — proposing, at various points, a lower $2 million cap on the tax break and a reduced pre-tax contributions cap, changes it estimates could save the budget more than $10 billion a year combined (approx.; Grattan Institute, 2023–2025). That a reform can be criticised simultaneously as too radical by the SMSF and tax-advice sector and as too timid by some economists is not proof either side is wrong. It is a reasonable indication that the government picked a genuine middle position — which is a defensible place to land, even if it means the final design satisfies neither the sharpest critics nor the most ambitious reformers.
The risk to the future
A template, not a one-off
The immediate risk from Division 296 as it now stands is modest: it reaches a small, high-balance minority, taxes income the fund has actually received, and indexes its thresholds. The larger risk sits one level up. Australia's Parliament has now shown, in detail and at length, that a government can draft, consult on, and push through one house of Parliament a tax on unrealised gains inside a compulsory retirement system — and that only a Senate numbers problem, not a settled principle, stopped it becoming law. Nothing in the Act that did pass prevents a future government, of any political colour, facing a tighter budget or a friendlier Senate, from reviving that formula, lowering the threshold, or pausing indexation, using the same TSB-based machinery that Division 296 already relies on to decide who sits in which tier. History with other thresholds is not reassuring on this point: "only the very wealthy" is exactly how every un-indexed or under-indexed tax line in Australia's system has been sold at the start, and exactly how each one has, in time, reached further down the income and wealth distribution than advertised. No one can say with certainty that this will happen to Division 296. What can be said is that the precedent now exists, the mechanism now exists, and the political cost of using it has already been shown to be survivable, provided the Senate numbers line up. The farm at the top of this article was never actually taxed on a gain it had not banked. The next asset like it, held by someone whose balance has grown past an indexed line that once looked comfortably out of reach, may not be so fortunate.