Division 296, the additional tax on large superannuation balances, is now law and takes effect from 1 July 2026. That single sentence carries a lot of history, because this measure has been through years of consultation, a lapsed bill, and significant redesign, and much of the commentary still circulating describes a version that no longer exists. For any firm with SMSF clients, getting the current, legislated position right is the whole job here, because the final law is materially different from, and in important ways kinder than, the original proposal that alarmed everyone. The measure passed Parliament in March 2026 and commences for the 2026-27 financial year, with the first assessments arriving after 30 June 2027, so the planning window is open now.
How Division 296 actually works in its final form
Division 296 adds a second layer of tax on superannuation earnings for individuals whose total superannuation balance exceeds $3 million. It is important to be precise about what it does and does not do. It is a personal tax, assessed on the individual rather than the fund, and it sits entirely on top of existing fund-level tax, so a member in accumulation phase still pays the usual 15% inside the fund and Division 296 adds to that. The additional rate is 15% on the earnings attributable to the proportion of a balance between $3 million and $10 million, which brings the combined rate on those earnings to around 30%, and an additional 25% on earnings attributable to the proportion above $10 million, for a combined rate of around 40% at that level. Critically, it is a tax on earnings, not a wealth tax on the whole balance, and it only applies to the proportion of earnings relating to the balance above the threshold. So a member with $3.5 million is taxed on roughly the one-seventh of their earnings that sits above $3 million, not on the full balance and not on the full earnings.
Two features of the final legislation matter enormously and correct the most common misconceptions. The first is that Division 296 does not tax unrealised capital gains. This was the single most controversial element of the original proposal, where paper gains on unsold assets would have been taxed and could have forced SMSFs to sell assets just to fund a tax bill on gains they had not received. That was removed. The final law taxes realised earnings only, meaning dividends, interest, rent, and gains on assets actually sold. The second is that both the $3 million and $10 million thresholds are indexed to CPI, in $150,000 and $500,000 increments respectively, which contains the bracket creep that worried people about an unindexed threshold. If a client or a colleague tells you Division 296 taxes unrealised gains or has a frozen threshold, they are describing the old proposal, not the law that passed.
The SMSF cost base reset and the traps around it
For SMSF clients specifically, the most important planning feature is the transitional CGT cost base reset, and it comes with sharp edges that need careful handling. An SMSF can elect to reset the cost base of its assets to their market value at 30 June 2026, so that only gains accruing after that date count toward Division 296 earnings when those assets are eventually sold. This is genuinely valuable for a fund holding assets with large accrued gains, but the rules around it are unforgiving. The election is all-or-nothing across every asset in the fund, so a trustee cannot reset the cost base on the appreciated property while leaving the shares alone, which forces a real decision where a fund holds a mix of large-gain and neutral or loss-making assets. It is irreversible once made, and it must be elected by the due date of the fund's 2026-27 tax return. There is also a further wrinkle worth flagging: any SMSF can opt in even if no member is over $3 million today, which can be the right call for a fund whose members are expected to cross the threshold later and which already holds assets sitting on large unrealised gains. Getting this election right, or right for the client's likely future position, is exactly the kind of decision that needs modelling rather than a default.
One further point that surprises clients: even though the tax itself is on realised earnings, a member's total superannuation balance, which determines whether they are caught at all, still reflects the full market value of their assets, including unrealised gains. So valuation evidence for SMSF assets will come under more scrutiny, because the year-end valuation decides whether the threshold is crossed. That administrative reality is worth building into the fund's annual process now.
What firms should be advising SMSF clients now
The planning window runs from now to the first assessment after 30 June 2027, and the useful work is identification and modelling rather than reflexive action. The first step is to flag which clients have, or are approaching, a total superannuation balance over $3 million, remembering that indexation softens but does not remove the long-run creep, so a client comfortably under today may be caught in a decade. For those clients, the questions worth modelling include whether the SMSF cost base reset is beneficial given the specific mix of assets in the fund, whether spousal balance equalisation through contribution splitting could keep both members under the threshold, and how the timing of realising gains interacts with the new tax now that only realised earnings count. What warrants real caution is the instinct to pull money out of super to duck the tax. Withdrawing can trigger capital gains tax, strip away creditor protection, and push future earnings into the member's personal marginal rates of up to 47%, so for most affected clients superannuation remains the most tax-effective structure available even under Division 296, and the question is how to structure what is held inside it rather than whether to abandon it. The measure is law, the numbers are knowable, and the value a firm adds is to have each affected client modelled and their cost-base-reset decision made deliberately before the 2026-27 return falls due, rather than discovering the exposure when the first assessment lands.
