The AFR called this private credit’s cockroach moment. The metaphor fits, but not for the reason given. The real issue is not whether more Bathlas are hiding. It is whether the funding model—open-ended funds promising monthly withdrawals while lending for years—ever made sense. These funds were sold to retail and superannuation investors, often relying on paid-for research ratings. Bathla did not invent this mismatch. It only exposed it.

The collateral

Administrators have reported that Bathla-related loans are concentrated in medium-density housing across a small number of fast-growing suburbs in north-west and western Sydney, with some properties securing more than one loan.

Table 1: The Bathla property book

Two points matter. First, there is a big gap between the $1.4 billion in real projects and the $15 billion in the sales pitch. Second, location. Some properties are in Melbourne’s outer west, but most are in a few postcodes in north-west Sydney. About 40 lenders could try to sell into this small area at once. Loan-to-value ratios of 60 to 65 per cent seem safe, until many lenders try to sell at the same time. Spreading loans across 40 funds does not mean the security is spread out.

The lenders, and where they touch your clients

The list of lenders is becoming clear. PAG has the largest exposure, at about $340 million. CVS Lane Capital Partners is next, with $250 million spread over nine land loans. These loans are not for construction. At year-end, 18 per cent of CVS Lane’s First Mortgage Fund and 24 per cent of its Property Finance Fund were tied up in Bathla. Both funds stopped accepting new money and froze withdrawals on 27 August after too many requests depleted a small cash pool. Centuria Bass has $278 million across six projects. Its two frozen funds have been paused since 14 August. Management says they hope to reopen within six months and have not yet written down any loans. Ray White Capital has $200 million in several sites. La Trobe Financial holds $38 million, a small share of its total, and says it will not freeze withdrawals. 360 Capital’s Mortgage REIT has $32 million and expects to recover all of it. Centuria’s own balance sheet holds $4.5 million. Balmain, Keyview, Credit Connect, Trilogy and MaxCap are also involved, but their numbers are not public. Alceon got out in January, selling $670 million. MA Financial has no Bathla loans but still limits withdrawals on its $2.3 billion fund to 1 per cent per month as a precaution.

Here is the part that matters for anyone advising into platform superannuation. The frozen funds, Centuria Bass and CVS Lane, are wholesale vehicles that sit off the retail menus. The funds that do sit on the big platform menus, and therefore inside several of the 25 largest APRA-regulated super entities as member-directed choice options, are La Trobe’s Australian Credit Fund (available via Macquarie Wrap, CFS, Insignia’s Expand, AMP’s North, BT Panorama, HUB24, Netwealth, Praemium and others), Trilogy’s Monthly Income Trust (Netwealth, HUB24, North, CFS Edge, Praemium and others) and MA’s Secured Real Estate Income Fund (BT Panorama, HUB24, Macquarie Wrap, Netwealth, Praemium and others). Of the three, only the MA fund is currently restricted, and it has no Bathla exposure.

The ratings that keep these funds on investment menus are thin and come from just one or two sources. When SQM cut Centuria Bass’s rating below investment grade, it alone triggered a rush for the exits and a freeze. Morningstar notes that a fund with a low rating is likely to be dropped from big investment platforms. La Trobe still has top ratings from SQM and Lonsec, but these were given before 25 August. Trilogy is rated only by SQM. The review of these ratings is yet to come, and the research firms are paid by the fund managers. When a single paid-for rating decides if a fund stays on a $100 billion super fund’s list, it is not research. It is infrastructure, and no one checks if it is sound.

What ASIC does next, and who should wear this

ASIC identified these risks before the collapse. Its June notice put private credit managers on watch ahead of 30 June valuations and referred to the La Trobe Australian Credit Fund in connection with its surveillance. REP 816 also found inconsistencies in the categorisation and disclosure of unlisted investments in super fund financial reports, as well as audit gaps. Last week, deputy chair Sarah Court described the “first significant cracks” in Australian private credit, while Commissioner Simone Constant called the episode the FSC private markets standard’s “first real test”. Against that backdrop, ASIC is likely to focus on four areas.

First, liquidity. The main problem in the frozen funds is the promise of monthly withdrawals, even though construction loans cannot be sold quickly. ASIC will check whether the rules for withdrawals matched what was promised in the fund documents, and whether pauses of two to six months were properly explained. Second, valuations. Loans to a borrower who stopped paying bills in mid-year should have been written down before August. Funds that kept Bathla loans at full value will have to explain why. Third, distribution. ASIC will ask if funds meant for monthly income were suitable for people in retirement accounts. Fourth, research houses. The conflict of interest in paid-for ratings has been ignored for years, even though one rating can now move hundreds of millions of dollars.

Responsibility is shared. First come the fund managers and trustees. They set up the promise of easy withdrawals and priced the loans. No amount of security justifies an open-ended fund lending for 30 months. Next are the platform trustees. They decide what products super members can buy, and they are paid for this. Passing the job to a single paid-for rating is not real oversight. Third are the research houses, for reasons already given. Last are the advisers, who sometimes ignore the liquidity rules when recommending these funds.

Regulators also share some blame. APRA’s rules on lending for property development pushed these loans out of banks and into non-bank funds. ASIC has not provided clear rules for valuing unlisted assets in managed funds. The structure that allows retail money to flow into wholesale funds remains allowed and is rarely checked. The people who built the system are now reviewing it. They should admit the design has flaws.

The superannuation question nobody has answered

To be clear about what the record shows: no large superannuation fund has been named as holding Bathla exposure, none of the frozen funds appears in the portfolio holdings disclosures of the large industry funds, and their private credit programs are dominated by offshore corporate direct lending. The exposure that does exist inside big super is member-directed, through the platform choice menus above. On the evidence available, trustee-directed member money is not in this event.

Bathla also shows a deeper problem for industry super funds. Some large funds now have more people taking money out than putting it in, just as they hold record amounts of hard-to-sell assets. When money is flowing in, these assets can boost returns. When money is flowing out, they become a problem. Each withdrawal is paid at a price based on values that cannot be checked against real sales.

This leads to a question of fairness. Super members can switch options almost instantly, at today’s price. But some assets behind that price, like private credit, can take months to turn into cash. When trouble hits, the pattern is clear. The informed and advised members move first, locking in prices that may not reflect reality. If values later fall, the loss does not vanish. It shifts to those who stayed put—the less engaged and less aware. In open-ended funds with hard-to-sell assets, this is not just theory. It is a real transfer of wealth from the least sophisticated to the most sophisticated, built into the system whenever asset values lag events.

The system has seen this before: in 2020, early withdrawals collided with unlisted-asset valuations; in 2023, Hostplus closed its dedicated property and infrastructure options as valuations fell. The responses were improvised—out-of-cycle revaluations, option closures and buffer management—and ASIC’s REP 816 indicates that reporting on unlisted assets remains inconsistent. If private credit valuations now lag events, fairness will become a live issue because trustees must protect all members, not only those who move quickly. The system promises daily access to money even when the underlying assets can take months to sell. Bathla is a small case, but the promise is much bigger. Allocators should ask every trustee two questions: how quickly do your unlisted marks move when the evidence moves, and who pays when they do not?

If I wanted more information, I had to reply to the email and agree to a two-year confidentiality undertaking that would bind me and any related entities.

I did not reply. The teaser itself was enough to work with. Everything I am about to describe was visible before the data room opened. The same approach works whether you are looking at a syndicated loan, a mortgage fund, a contributory scheme, or a credit note. Teasers are their own genre, and the genre has tells.

First read: whose collateral is it?

The teaser says the transaction is secured by $2 billion of freehold property. It also says the offer is for the mezzanine tranche.

These two statements serve different purposes. The property secures the whole capital stack. The mezzanine sits inside that stack, behind the senior lenders. If things go wrong, the seniors are paid from the collateral first. The mezzanine gets whatever is left. In marketing, ‘asset-backed’ often just means the seniors are asset-backed and you are behind them.

The first questions are about structure, and the teaser does not answer any of them. How much senior debt is ahead of the mezzanine? Where does the mezzanine attach and detach in the capital structure? Is it actually secured, or is it sitting behind a holding company with second-ranking security, or maybe no security at all? A $2 billion collateral pool means nothing for your recovery until you know how much is already spoken for before you get a turn.

Anyone who has invested in mortgage funds should spot this straight away. A first mortgage at 50 percent LVR and a second mortgage behind a construction facility are both ‘secured by property.’ They are not the same investment.

Second read: the creditor you cannot see

Every asset class has a creditor class that does not appear in the headline security description. In residential aged care it is refundable accommodation deposits. A large operator carries hundreds of millions, often billions, in RADs: lump sums paid by residents, repayable when they leave, sitting inside the operating entities as an interest-free, effectively at-call liability. The Commonwealth guarantees residents get their money back, then stands in their shoes as a creditor of the provider. Current reform settings add policy uncertainty to that liquidity. The $2 billion freehold portfolio is not a clean equity cushion. A big chunk is already committed to refunding deposits before anything gets to a subordinated lender.

In mortgage lending, the invisible creditors are different, but the discipline is the same. Prior ranking charges. Unpaid council rates and land tax, which rank ahead of the mortgagee. GST on a sale the borrower did not remit. Builders with security of payment claims on an unfinished project. The question is always the same: what claims are out there that the security description does not mention, and where do they rank compared to me?

Third read: is the strength a deal feature or a sector feature?

Seventy percent of cash flows from the Federal Government might sound like credit enhancement. Really, it just describes the industry. Every residential aged care operator in the country gets most of its revenue from Commonwealth subsidies. This statistic does not set this borrower apart.

Government revenue also brings its own risks. The borrower’s margin depends on policy. Mandated care minutes, nurse requirements, and wage decisions can push costs up faster than subsidy indexation increases revenue. The sector’s history of thin or negative per-bed operating margins is well known. That is one reason the current owner is selling.

The mortgage fund equivalent is the teaser that leads with ‘all loans secured by registered mortgages.’ So are everyone else’s. A feature that every participant in a market shares is a reason not to pick this one: we need to ask whether the arithmetic reconciles.

BBSY is currently in the high threes. Add 5.75 percent, and you get about 9.4 to 9.6 percent. The teaser quotes 10.25 to 10.5.

That gap has to come from somewhere. Maybe it is upfront fees, original issue discount, or a blended assumption that includes the higher PIK margin. None of this is disclosed. When the quoted yield is higher than the quoted coupon, you should ask why before you sign anything. The difference usually shows how the arranger expects the loan to behave. The PIK toggle is a clue. The teaser says interest is ‘expected’ to be serviced mostly in cash, but also allows it to be capitalised at a higher margin. If the structure is built from day one to allow flexibility in debt service, that tells you what the arrangers really think about cash flow certainty, no matter what the covering email says.

Fifth read: who is selling, and why is that presented as comfort?

The teaser says a global investment bank has underwritten the deal and will co-invest at the mezzanine level alongside investors. This is meant to be a sign of confidence. Sometimes it is. Sometimes it just means the underwriter could not sell the tranche. A leftover hold is not an endorsement. The real question is whether the bank’s position is a choice or just what they are stuck with, and whether their economics match yours or include fees you do not get. The same goes for the manager ‘co-investing alongside investors.’ At what level, in which unit class, with what fee rebates, and does the co-investment rank equally or get repaid first?

Sixth read: the mechanics of the offer are themselves disclosure

Finally, treat the offer document as evidence about the people making the offer. In this case, I was asked to accept a two-year confidentiality undertaking by return email, binding my related entities, before they would even name the borrower, the sponsor, or the underwriting bank. The same document says they do not guarantee the information you get is accurate, reliable, or complete, and they have no obligation to correct it. It also says a public bookbuild is expected soon. You are being asked to decide quickly, based on information nobody stands behind, about parties nobody has named.

None of these terms is unusual on its own. Taken together, they show you how the transaction is structured to treat you. Urgency, asymmetry, and non-reliance are not just quirks of retail-facing private credit. They are the business model.

The point

None of the above needed the information memorandum. It just took ten minutes, a teaser, and six questions: where do I rank, what claims can I not see, is the headline strength just a sector commonplace, does the arithmetic reconcile, whose co-investment is a choice rather than a leftover, and what does the offer’s own behaviour tell me about the people making it.

Most of the private credit and mortgage fund blow-ups now in the courts and ASIC reports were visible at this stage. Not the specific failure, but the shape of it. The data room is where you check the details. The teaser is where you decide if checking is even worth your time. Most of the time, it is not.

Nothing here is financial advice. It is a description of how one practitioner reads marketing documents.

The version that survives

Even without a conspiracy, something real remains. This is talking your book at scale.

The essay made the manager famous. The fame brought in capital. The fund’s size and performance made the thesis look credible. That credibility attracted more capital, which lifted the same companies the thesis relied on. This improved the pricing for every private AI position the early backers held. None of this needs coordination or intent. It only needs everyone to believe the thesis and benefit when others believe it too.

George Soros called this reflexivity. The new part is the packaging. The essay worked as a pitch for capital. The hedge fund served as proof of concept for the thesis. The backers’ private holdings were tied to the same story the fund promoted in public markets. This is legal. It is common on a small scale. It is rare to see it done this quickly or at this size.

What should allocators take from it?

Map the seed capital. When reviewing a new fund, identify whether its backers hold private stakes tied to the fund’s public thesis. That overlap is not automatically disqualifying, but it belongs in due diligence because it can amplify marketing and reduce independent challenge from the investor base. Then assess whether the manager’s operating experience matches the vehicle’s risks: an inexperienced manager running long-only, unlevered money is a talent bet, while the same manager running leveraged concentration through prime brokers is a survival bet. The August episode was not necessarily a failure of stock selection; the positions may yet prove right. It was a failure of liability management, the discipline that experience is meant to provide.

Reflexive flows can reverse quickly. Capital that comes for a story leaves when the story weakens. Funds built on fame face redemptions and collateral calls that process-driven funds avoid. The speed of the near-collapse, from record assets to margin calls in days, is the warning sign.

Watch what the smart private money does when IPOs arrive. The key question for the next 18 months is not whether the AI trade is right. It is whether the early investors in the loudest public funds are buying or selling when their private stakes list. That answer will matter more than any essay.

The young founder may rebuild. The thesis may yet prove right. But the structure around him is a case study for allocators. Narrative, capital and self-interest form a loop. The risk management came last.

That run is over. The reversal has been abrupt.

The American evidence

Since 30 June, MTUM has fallen 10.1 per cent. The S&P 500 has risen 3.2 per cent. MTUM is now 10.7 per cent below its late June peak. In the past three months, momentum returned just 0.7 per cent. The broad market returned 4.5 per cent. Value (IWD) returned 10.3 per cent. The factor that led the rise now leads the fall. Capital is moving to the stocks it once avoided.

Growth shows the same pattern. The Russell 1000 Growth ETF (IWF) returned 11.5 per cent over twelve months, about half the S&P 500. Since May, it has been flat. This year, gains have not come from expensive stocks. American small caps (IWM) returned 40.7 per cent in the year to June. That is 18 percentage points ahead of the S&P 500. The small cap re-rating that strategists predicted has already happened in America.

The Australian evidence

Australia followed the same pattern, but with different stocks. Here, momentum was in gold, resources and banks, not technology. These assets were not bought for earnings growth, which was missing. They were bought because prices were rising. CBA trading at more than 30 times earnings was momentum investing in a blue suit.

Since the end of February, when the trade peaked, Westpac has fallen 14.4 per cent. NAB is down 14.0 per cent. ANZ is down 6.5 per cent. CBA is flat, as is the ASX 200. Over twelve months, CBA has returned just 0.2 per cent. The multiple expansion that once lifted the index has stopped. Earnings have not risen to fill the gap. The stocks nobody wanted remain untouched. The Small Ordinaries has returned 2.6 per cent over twelve months and is 11 per cent below its January peak. It still trades at about 14 times forward earnings. The market leaders trade at 17 to 18 times. In America, the neglected segment has already re-rated by 40 per cent. In Australia, it has not moved.

The trend followers agree.

A third sign comes from systematic trend-following funds. These are momentum in its purest form. They rely only on price direction across many futures markets. They ignore stories and valuations. After a strong start to 2026, they have struggled since mid-year. Energy has been a problem. Downtrends reversed sharply, leaving systems on the wrong side. When machines built to ride trends lose money, it is usually a sign that the trends have broken.

What history says happens next

Momentum unwinds rarely end quietly. When capital leaves crowded winners, it often moves to neglected stocks. The adjustment is sharpest in the thin parts of the market, where small trades can move prices. In 2009, 2012 and 2020, extremes in the Australian small cap discount faded over three to five years. Small caps beat large caps by nearly 10 per cent a year in the five years after the trough.

This does not mean the rotation starts now. Momentum can fade slowly. Markets may drift sideways for months as leadership changes. Yet evidence from three sources now points the same way. American factor ETFs, Australian banks and trend-following funds all agree. The question for allocators is no longer whether the momentum trade has ended. It is where the capital moves next, and whether they are leading or following.

Sources: price data from Yahoo Finance to 11 August 2026 (ASX index data to 10 August); returns are price returns for ASX indices and total returns for US ETFs.

It is instructive to observe what was absent from Westpac’s list of causes. Net overseas migration remains in the hundreds of thousands annually, and the structural undersupply of housing persists. If population growth were the primary determinant of house prices, mortgage demand would not be contracting amid such a substantial migration intake. Westpac now characterises population growth and undersupply as factors that may only partially mitigate the impact of interest rates and tax policy. This ordering is correct, and it stands in direct opposition to the hierarchy that has shaped public discourse for the past decade.

Over the past three months, I have advanced a consistent argument across three analyses. Westpac’s update now provides empirical confirmation of that thesis within its loan book.

The data: rates dominate, migration explains essentially zero

“What Actually Moved Australian House Prices” (June) put 36 years of data on one chart: Sydney and Melbourne median prices, the standard variable mortgage rate, and net overseas migration, 1990 to 2026.

The results are unambiguous. The standard variable mortgage rate declined from 17 per cent in 1990 to 3.5 per cent in 2021. During this interval, Sydney house prices increased by a factor of 7.9 and Melbourne by 7.0. The price trajectories closely track interest rate movements. A model incorporating only the mortgage rate and trend income growth, excluding policy variables and migration, accounts for 23 per cent of Sydney’s price variation and 35 per cent of Melbourne’s. Migration, regardless of lag or transformation, contributes no explanatory power.

The period offers two natural experiments, both yielding the same conclusion. In 2020-21, net migration turned negative for the first time in modern Australian history, yet house prices in Sydney and Melbourne rose by 25 per cent as interest rates approached 3.5 per cent. In 2023, migration reached a record 538,000, but prices declined as rates increased. When interest rates and migration move in the same direction, attribution is ambiguous. When they diverge, as has occurred twice in five years, the influence of interest rates prevails in both instances.

The mechanism: prices are borrowing capacity in disguise

“The Maths Australia Doesn’t Want to Do” provided the underlying arithmetic. Residential property is purchased with mortgage finance, and mortgage finance is constrained by monthly repayment capacity. A buyer able to allocate $5,000 per month to repayments could service a 30-year loan of approximately $1.19 million at the 3 per cent rates prevailing in late 2021. At the 6.5 per cent rates projected for mid-2026, following recent increases, the same repayment supports a loan of only $790,000. The buyer, with unchanged income and employment, has lost one third of their purchasing power.

New Zealand provides a further empirical test. Despite persistent supply constraints, robust population growth, and a history of favourable tax policy, real house prices have declined by 31 per cent from their 2021 peak. Auckland has fallen 37 per cent and Wellington 39 per cent, a direct consequence of interest rates rising earlier and more sharply than elsewhere. The outcome for first-home buyers is instructive: Wellington’s 39 per cent price decline was offset by a 41 per cent reduction in borrowing capacity. Lower prices have not improved affordability; they simply reflect the contraction of credit that previously fuelled price inflation.

The repricing: from growth asset to income asset

“Australian Property Now Priced on Income” examined the implications of the May Budget changes for property investors. The removal of negative gearing for established dwellings and the replacement of the 50 per cent capital gains tax discount with inflation indexation at a minimum rate of 30 per cent reduces the after-tax return on a leveraged investment property from approximately 14 per cent to around 5 per cent, which is below the cost of debt. Negative cash flow, previously a subsidised strategy, now represents a compounding loss. The rational investor’s calculus shifts from accepting subsidised losses to requiring positive cash flow from the outset, compelling the asset to be valued on income, as with any other investment. A discounted cash flow analysis implies a 21 per cent price decline if rents remain unchanged, or a 5 per cent decline if rents increase by 20 per cent, with intermediate outcomes along this spectrum.

Westpac’s forecast that investor credit growth will halve to 4.5 per cent by 2027 indicates that this repricing is now being reflected in the loan book. A 20 per cent decline in applications over three months cannot be attributed to demographic change. It is the result of altered financial arithmetic.

What to watch

The forty-year period of declining interest rates was the principal driver of rising house prices. The same arithmetic, now operating in reverse through higher rates and revised tax policy, will drive the adjustment downward. Capital gains tax, negative gearing, and immigration policy each merit consideration on their own merits. However, the data does not support the proposition that immigration is the determinant of housing affordability in Australia, and as of yesterday, neither does the country’s second-largest mortgage lender. The most informative variable to monitor remains the path of mortgage rates, followed by the quarterly mortgage application data reported by the banks.

Sources: Westpac 3Q26 trading update, ASX, 10 August 2026; Reuters and Stockhead reporting of same; NAB July trading update; prior analysis and data in “What Actually Moved Australian House Prices”, “The Maths Australia Doesn’t Want to Do” and “Australian Property Now Priced on Income”, danielliptak.substack.com.

The Australian, meanwhile, reported that landlords are lifting rents sharply or selling up. Advertised increases include a Townsville house going from $500 to $570 a week, a rise of 14 per cent, and a Brisbane apartment going from $590 to $680, a rise of 15 per cent. The Agency reports a 15 per cent jump in investors requesting sales appraisals in four weeks, rents up 3 to 10 per cent across its networks since the May Budget, and rental open-home attendance up 30 per cent in Melbourne. PropTrack has national rents at a record $670 a week. Treasury modelled the Budget’s rental impact at $2 a week. That figure has already been overtaken.

Falling prices and rising rents are consistent with a broader repricing of Australian residential investment property from a capital-gains asset towards an income asset. Under the assumptions used here, the two outcomes follow from the same adjustment mechanism. The following analysis sets out the numbers.

How investment property used to be priced

For two decades the return on an Australian investment property had three components: a small net rental yield, an expected capital gain, and a tax subsidy that converted operating losses into refunds and taxed the eventual gain at half rates. Nobody ran a discounted cash flow on the rent, because the rent was not the point. The rent was a holding cost offset while the investor waited for the gain, and the tax system paid a large share of the wait.

Take a representative established dwelling: price $800,000, rent $650 a week, or $33,800 a year. Allow 29 per cent of gross rent for agent fees, rates or strata, insurance, maintenance and vacancy, leaving net operating income of about $24,000, a net yield of 3 per cent.

Now price it as an investor did in early 2026, before the Budget, at an 80 per cent loan-to-value ratio and a 6 per cent interest-only rate.

• Interest on $640,000 of debt: $38,400

• Pre-tax cash flow: $24,000 less $38,400 = a loss of $14,400

• Negative gearing refund at a 47 per cent marginal rate: $6,768, cutting the after-tax holding cost to $7,632

• Expected capital growth at 5 per cent: $40,000, worth about $30,600 after CGT at the discounted effective rate of 23.5 per cent

After-tax return is approximately $23,000 on $160,000 of equity, or about 14 per cent. The return is derived entirely from capital growth and tax benefits, as the income component is negative. Under these conditions, a 3 per cent net yield was rational; it was the appropriate DCF outcome given the growth assumption and tax treatment.

What the Budget removed

For established dwellings acquired after 7.30 pm on 12 May 2026, negative gearing is gone: losses are quarantined against the property rather than deductible against wages. From July 2027, the 50 per cent CGT discount is replaced with inflation indexation and a minimum 30 per cent rate on real gains. And the RBA has hiked three times this year, taking variable rates to around 6.5 per cent.

Run the same property for a buyer today, granting a generous assumption that prices grow at 3.5 per cent nominal against 2.5 per cent inflation.

• Interest at 6.5 per cent: $41,600

• Pre-tax cash flow: a loss of $17,600, now quarantined, with no refund

• Capital growth of $28,000 nominal, of which the real component of $8,000 is taxed at 30 per cent, leaving about $25,600 after tax

At $800,000, this cash-flow profile is unlikely to meet the return requirements of a rational marginal investor. The previous price reflected the DCF outcome under the old assumptions; with the inputs now changed, the clearing price must adjust.

The more significant change concerns negative cash flow itself. Previously, operating a property at a loss was a deliberate strategy, with the tax office subsidising nearly half the shortfall through negative gearing, and the discounted capital gain compensating for the remainder. Investors could rationally hold loss-making assets for extended periods because losses were subsidised during ownership and taxed at concessional rates upon exit. Under current settings, negative net cash flow is simply a recurring loss, with no refund and no reliable capital growth to offset it. The rational approach now requires positive cash flow from the outset. Achieving this on an established dwelling at a 6.5 per cent cost of debt requires either a substantially lower purchase price or a significantly higher rent.

The inversion: what price, or what rent, makes it work again

A practical application follows. If a 9 per cent after-tax return on equity is required, which is a modest expectation for geared property risk given a term deposit yields 4.5 per cent, and all other assumptions are held constant, the DCF can be solved by adjusting either the price or the rent.

The estimates below are illustrative rather than forecasts and depend on the stated financing, tax, inflation, growth and required-return assumptions; different inputs will shift the clearing price and rent, but not the direction of the repricing mechanism.

Using the above example, hold the rent at $650 a week, and the price that clears the required return is about $630,000. That is a fall of 21 per cent.

Hold the price at a 10 per cent fall, $720,000, and the rent that clears it is about $740 a week. That is a rise of 14 per cent.

Between those poles sits a frontier of combinations, each of which restores the investment case:

The market will settle at some point along this frontier, with the implied equilibrium gross yield in the range of 5.2 to 5.4 per cent, compared to approximately 4.2 per cent currently. Yields in the low fours reflected the previous tax settings. An income-priced market requires yields in the mid-fives.

Recent reports of advertised rent increases of 14 and 15 per cent in Townsville and Brisbane align precisely with this frontier. Landlords may not formally calculate a DCF, but the pursuit of yield to achieve cash-flow positivity effectively mirrors the same analysis, albeit informally. Treasury’s estimate of a $2 per week impact assumed the tax change would be absorbed within investor returns. The frontier demonstrates why this is not the case: returns were already at the minimum acceptable to marginal investors, so the adjustment necessarily occurs through prices and rents.

Why both margins move, not one

Whether the burden falls on prices or rents depends on who sets each.

Prices are determined by the marginal buyer. For established dwellings, investors are no longer the marginal buyers, as the after-tax calculations outlined above have removed their incentive, a trend confirmed by exodus data. The marginal buyer is now the owner-occupier, whose borrowing capacity has been reduced by the interest rate cycle, as previously discussed. Neither group supports current price levels, resulting in price declines. Grandfathering slows but does not prevent this process: while existing investors are not compelled to sell, appraisal data and the actions of pre-1985 CGT-exempt holders selling ahead of the 2027 changes indicate a significant cohort is exiting, with no replacement.

Rents are determined by the balance between renter households and available rental stock, and this balance is tightening from both directions. Each landlord sale to an owner-occupier reduces the rental pool, and while some renters become owners, the exchange is not one-to-one. Purchasers of this stock are more often existing owners and upgraders than tenants. Simultaneously, the pool of renters increases, as the same rate cycle that reduced prices has further constrained borrowing capacity, preventing marginal tenants from transitioning to ownership. Rental supply declines, demand remains, and rents increase. Living Here Cush Partners’ warning that the next twelve months will see no new investors and net investor selling accurately describes this dynamic.

SQM’s Louis Christopher contends there will be no crash because Australia, unlike markets that have experienced crashes, faces a chronic shortage of dwellings rather than a speculative construction surplus. This assessment is correct, but the conclusion requires careful interpretation. The shortage establishes a floor under prices, but not under rents. In such a shortage, adjustments that cannot be fully absorbed by prices are instead reflected in rents. The supply argument does not contradict this analysis; it explains why the rent component of the frontier bears more of the adjustment.

The Budget’s approach, redirecting negative gearing to new builds only, is logically consistent in theory but slow to take effect in practice. At current interest rates, construction feasibility is impaired, with builders exiting the market rather than expanding. The New Zealand experience indicates that capacity lost during downturns takes years to restore. The carve-out will be relevant in the next cycle, but it does not increase dwelling supply in the current one.

What this means

For allocators and advisers, the practical implications are clear. Established residential investment property should now be assessed as an income asset, requiring a gross yield in the mid-5 per cent range, with quarantined losses, indexed capital gains tax at 30 per cent or higher, and no reliance on capital growth unsupported by the current rate environment. Portfolios containing recent-vintage geared residential property will experience compressed returns from multiple directions, and while grandfathered tax treatment mitigates holding costs, it does not restore exit economics. Rental inflation of 10 to 15 per cent during the adjustment period, several times Treasury’s estimate, should be incorporated into cost-of-living and wage pressure assumptions, with all the implications that holds for the RBA.

In summary, the Budget’s changes represent defensible tax policy, but the transition is regressive. The removal of the subsidy is initially borne by falling prices for leveraged owners and rising rents for tenants, the two groups least able to absorb the impact. This is not an argument against the reform, but rather an argument for acknowledging who bears the adjustment and for avoiding the pretence, as in the $2 a week estimate, that such straightforward arithmetic would not take effect.

In an earlier Substack post, I argued that house prices are interest rates in disguise. This one adds the corollary. Rents are the discount rate applied to a market that can no longer pay its investors in capital gains. Both are now moving the way the arithmetic says they must.

The Senate committee examining property development heard this month that something in the order of 13 million bedrooms sit empty in Australia every night, largely for want of sensible downsizing options.

That is the story most people noticed. It is not the important one. The real headline was buried further down, and it is why I am back for a third round on land tax instead of moving on.

The rivers of gold have stopped flowing.

In the June budget, the New South Wales government wrote down its expected take from transfer duty and land tax by $8.4bn against what it had forecast only a year earlier. Total taxation revenue over the four years to 2029-30 came down by $7.1bn. Stamp duty for 2026-27 was cut to $12.6bn, a slump of nearly $2bn on the prior year and part of a $5.3bn decline across the forward estimates.

SQM Research estimates the states collectively stand to lose as much as $9bn as transaction volumes fall by around 30 per cent. On its numbers, the damage runs $3.2bn in New South Wales, $2.8bn in Victoria, $1.8bn in Queensland, $830m in Western Australia and $510m in South Australia. Loan Market data has mortgage applications down 25 per cent in New South Wales and 27 per cent in Queensland since their February peaks, with first-home buyers down 23 per cent, investors down 31 per cent and upgraders down 14 per cent.

The e61 Institute called this before it happened, describing property tax as the canary in the coal mine ahead of the NSW budget. Its point was that stamp duty is uniquely exposed because it depends on both price and volume, and volume is the more violent of the two. In December, the half-yearly review had revised NSW stamp duty up, with a further $800m of growth forecast for 2026-27. Six months later, the same line item was gutted.

Look at what is actually causing the pain. Sydney’s median is down about 3.7 per cent since January, now sitting at $1.26m. Melbourne is down 2.6 per cent to about $800,000. Hardly dramatic. The real collapse is in volume, not price, and volume is exactly what a transaction tax cannot handle.

The circularity nobody is naming.

Here is the chain of events, as simply as I can put it, since no one else seems to have done so.

In May, the Commonwealth reformed negative gearing and the capital gains tax discount to improve housing affordability. Transaction volumes fell. The states’ second-largest tax is levied on transactions. So a federal housing reform is being partly financed, without anyone deciding this, by the revenue base of eight governments that were not consulted and cannot control it.

Every serious look at moving from stamp duty to land tax finds the same thing: the states need a financial bridge during the switch, because duty revenue drops off before the new base builds up. That bridge is only affordable when duty revenue is strong. It becomes impossible just when you need it most.

In other words, the only time you can realistically escape a procyclical tax is at the top of the cycle. Australia just had that chance and did nothing. The states are now poorer, more exposed, and even less able to pay for the fix. That is what a trap looks like.

Ted Quinlan wrote this down in 2012

None of this is new. We have seen this movie before.

The ACT’s 2012 review, which I wrote about last week, concluded that the territory’s revenue streams were unfair, unstable and therefore unsustainable. Its specific finding was that conveyance duty was about a quarter of the ACT’s total tax take and was paid in any given year by roughly 9 per cent of households. A quarter of the base resting on a small and arbitrarily selected group, swinging with the cycle, unforecastable.

The ACT acted on that finding. Fourteen years later, Prosper Australia’s evaluation records that the average Treasury forecast error on ACT revenue fell from 7.9 per cent before the reform to 2.6 per cent after it.

A note of precision, since the numbers get quoted loosely. The AFR describes stamp duty as close to 30 per cent of NSW taxation revenue; e61 puts it at about 10 per cent of NSW revenue overall. Both are right. The first is the share of tax collections; the second is the share of the whole budget, including grants. The distinction matters because the first number tells you how exposed the tax base is and the second tells you how survivable the shock is. New South Wales is very exposed and, just about, survives it.

The patches are making it worse.

Faced with a tax that fails on both efficiency and revenue, the states have reached for concessions.

Queensland introduced relief of up to $103,830 for downsizers aged 60 and over from 25 March this year, available on new builds, off-the-plan apartments and vacant land. The ACT abolished duty for all first-home buyers, a national first. New South Wales replaced the Perrottet government’s first-home buyer choice between duty and an annual property tax with a straightforward waiver under $800,000. Victoria has extended its off-the-plan concession to April 2027. New South Wales is legislating surcharge purchaser duty relief for build-to-rent and retirement villages.

Each of these makes sense on its own. Put together, they miss the point entirely.

You cannot fix a volatile tax by carving out the bits that win sympathy. Every concession shrinks the base, piles more risk onto a smaller and more cyclical group, and makes the revenue swings worse. The states are treating the symptoms of an unfair, unstable tax by making it even shakier. And the concessions themselves distort the market, nudging buyers toward new builds instead of actually removing the barrier to moving.

Queensland’s downsizer measure is the clearest example. Policymakers have spotted that older owners are stuck in place. But the relief only applies to new builds. So a retiree who wants to move into an existing apartment two streets away, which is what most actually want, gets nothing.

Mobility lock-in is not an anecdote.

The man in the eastern suburbs is not unlucky. He is simply experiencing the policy exactly as intended.

The e61 Institute finds that a typical stamp duty payment now costs about five months of take-home income, roughly double the burden of the 2000s, and that each one percentage point increase in stamp duty reduces the rate of home moves by around 7.2 per cent. Matthew Cridland of Hamilton Locke, quoted in the same article, cites work putting the economic cost at about 81 cents for every dollar of duty raised.

That 81 cents should look familiar to anyone who read my first piece. It is essentially the Treasury’s own 2015 estimate of the marginal excess burden of conveyance duty, the worst-performing tax on the chart. Independent practitioners, the Commonwealth Treasury, the Productivity Commission and half a dozen university modelling teams have converged on the same figure from different directions over eleven years. There is no live empirical dispute here.

Nerida Conisbee of Ray White makes the necessary qualification, and it is a fair one. Removing stamp duty does not help if there is nothing to downsize into. Middle-ring suburbs across Australia have very little smaller-format housing, and developers such as Coronation Property describe a feasibility gap in the accessible mid-market, where the only apartments that stack up financially are the expensive ones.

Reforming stamp duty is not enough on its own. Tax underused land in good suburbs and developers face a simple choice: pay to sit on the site, or build. Transaction taxes keep people stuck in their homes. Cheap holding costs keep land stuck in low-value uses. A land tax deals with both, which is why it keeps cropping up as the answer to problems that seem unconnected.

In plain sight

Lay out the timeline, and it starts to look faintly ridiculous.

Adam Smith identified ground rent as the ideal object of taxation in 1776. Henry George built a global movement on it in 1879. Australia legislated a Commonwealth tax on unimproved values in 1910 and ran it until 1952. The Henry review made four recommendations on land tax and stamp duty in 2010. Ted Quinlan diagnosed the volatility problem in 2012, and the ACT began fixing it the same year. The Treasury published the marginal excess burden chart in 2015. Malcolm Turnbull, as prime minister, said out loud that no tax economist in the country would disagree with the swap. The Productivity Commission has recommended it. Grattan has costed it at up to $17bn a year of additional GDP. Infrastructure Victoria put it back on the table this year and estimated the implementation cost at between $1m and $5m.

This is not a question of missing information. No one is waiting for another study. The real problem is that the states own the tax base and pay the transition costs, while the Commonwealth holds the purse strings and collects most of the upside.

What has changed is that the Commonwealth now has something to lose. Its own housing reforms are being funded, in part, from a state revenue stream it does not control and has just weakened. A few years of transition funding from Canberra, tied to scrapping stamp duty, would cost little and buy every state a permanent upgrade in fiscal resilience. Canberra underwrites the states anyway. The Melbourne Institute made this exact point in May.

For allocators

Four things I would take from this.

Volume is what matters. State revenue is much more sensitive to turnover than to price. If you are forecasting state budgets or holding state bonds, watch transaction counts. Median prices are just a distraction. A 3.7 per cent price drop did not cause this. A 30 per cent fall in volume did.

The downgrade cycle is here, and it is not hitting everyone equally. The states with the highest duty rates have the most to lose and, conveniently, the most to gain from reform. Victoria’s rate is the highest. The ACT’s is the lowest. That tells you where the pressure to change will show up first.

Existing land taxes are not what is being proposed here. Notice that the NSW downgrade hit both transfer duty and land tax. Australia’s current land taxes are narrow, full of thresholds, and based on values that swing with the cycle, so they end up just as volatile. A broad, flat charge on unimproved land is a different animal. Do not confuse the two when you are modelling.

The politics have shifted. For forty years, the excuse for not moving to land tax was that stamp duty was too much of a cash cow. That argument has just been put out to pasture. Every treasurer is now staring at years of downgrades in the same budget line. The case for a stable base is no longer just economist chatter; it is now a credit rating issue. The economic case has been settled since 1776 and achieved nothing. The fiscal case is new, and fiscal arguments are the ones that get things done.

Someone in Sydney is sitting on four empty bedrooms because moving would cost $150,000. Someone in Treasury is writing down $8.4bn because not enough people are moving. The same tax is causing both problems, and the fix has been gathering dust on the shelf, fully specified and independently checked, for longer than any of us have been around.

Except for one. In 2012, the ACT launched a twenty-year plan to ditch conveyance and insurance duty in favour of general rates. Fourteen years on, it remains the only real-world evidence we have. The results are more interesting than either camp lets on. The reform did what it said on the tin, but the latest independent review says it will not finish on schedule.

Both points matter, but the second is not much to do with the first.

What Canberra was trying to fix

The intellectual foundation was Ted Quinlan’s 2012 review, which concluded that the territory’s main revenue streams were unfair, unstable and therefore unsustainable. The specific indictment of stamp duty is worth restating because it is the strongest sentence in the entire Australian tax debate and almost nobody quotes it.

Before the reforms, conveyance duty was about a quarter of the ACT’s total tax take. A quarter of the revenue rested on a small, random group, chosen not for their ability to pay but simply for moving house. The revenue bounced along with the property cycle, and forecasts became a game of pin-the-tail. The hardest hit were those least able to bear it: first-home buyers and anyone moving for work, divorce or bereavement.

The idea was to spread the same revenue across all landholders each year. The aim was not to squeeze more out, but to raise the same amount from a base that stays put.

What the evidence says

The ACT government commissioned a formal evaluation in 2019, covering the first seven years. It was overseen by an advisory group including Robert Breunig of the ANU, Richard Denniss and Robert Tanton, with the modelling split between the Centre of Policy Studies at Victoria University, the Tax and Transfer Policy Institute at the ANU and NATSEM at the University of Canberra. Publication was delayed by COVID, which is part of why the findings are less known; the promise was kept. Over seven years, general rates climbed by $793m, while the revenue lost from stamp duty and insurance duty was $855m. The territory ended up collecting $62m less than it would have without the reform. Whatever else you say, this was not a sneaky tax grab in disguise.

Every economic indicator nudged in the right direction, though none by much. Real gross state product rose by $302m, or 0.11 per cent. Real investment, employment and wages all crept up by similar small amounts. These gains came at no extra cost. The territory collected the same revenue in a new way. In the property market, turnover and prices both rose. In rentals, the evaluation found rents probably fell, and supply increased, especially at the lower end. Both duty and rates became more progressive. The bottom four income groups paid a smaller share than before. The average Treasury forecast error on revenue dropped from 7.9 per cent before the reforms to 2.6 per cent after. The base stopped lurching, which was the main goal, and it was achieved.

The prices question deserves a clear answer.

The loudest prediction in 2012 was that annual land taxation would be baked into land values and Canberra prices would tumble. They did not. Prices and turnover both rose, and the reform’s defenders have used this ever since as proof that the capitalisation fear was overcooked.

No one should lean too heavily on that, especially if you are allocating capital.

Two effects pull in opposite directions. Scrapping duty lets buyers pay more, since the money once sent to the Revenue Office now goes into the deposit and gets leveraged. Adding an annual charge lowers the land’s value by the present value of the future bill. In the ACT, the first effect won. Strong population growth, steady public-sector jobs and a fourteen-year house price boom did their bit. Almost any policy would have looked good.

Canberra is not a perfect test of the pure Georgist capitalisation theory. It shows something more practical. A transition this big, stretched over many years, does not cause messy repricing. That is a point about transition risk, not about the long-term impact of a full land value tax. Anyone modelling a national swap should measure the capitalisation effect properly. Canberra alone does not close the case. This year, Pegasus Economics delivered an analysis of the ACT budget to the Legislative Assembly’s estimates committee. Its conclusion was blunt. Unless there are dramatic changes to both conveyance duty and general rates in coming budgets, it is extremely unlikely the government will meet its original deadline for eliminating conveyance duty.

The details are gloomier than the headline. The report found the rate of substitution has basically stopped. General rates are expected to stay flat at about 29 per cent of own-source revenue from 2025-26 onwards. Conveyance duty is still projected to make up about 10 per cent in 2029-30, with just two years left on the twenty-year clock.

Fourteen years on, the ACT has trimmed stamp duty but is nowhere near scrapping it for the following reasons.

The rates base became too handy.

In 2011-12, the ACT collected $209m in general rates. By 2026-27, it expects $944.2m out of a total tax take of about $3.3bn. Some of that is the reform working as planned. Some is not. The clearest sign was the $250 health levy added to rates bills in 2025-26, then dropped in 2026-27. Once a government has a broad, stable base, it is hard to resist using it for other purposes. Each time that happens, the deal offered to voters in 2012 looks less like a swap and more like a ratchet.

The base never became neutral.

This gets little attention but matters most if you are allocating capital. The ACT did not create one broad land value tax. It built general rates on all land, then kept a separate land tax for property that is not an owner-occupied home. That tax is a fixed charge of $1,693 plus a valuation charge, assessed quarterly. So a rented house pays much more than an identical owner-occupied house next door.

That is the opposite of what theory suggests and the opposite of what the housing shortage needs. It also weakens the reform’s own rental-supply argument. If you want institutional and private capital to provide rental housing, taxing rentals more than owner-occupied homes is not going to help.

Concessions crept back in.

The 2026-27 budget scrapped stamp duty for all first-home buyers, a national first. It also removed duty for eligible pensioners and disability concession recipients, removed duty on new unit-titled properties bought by owner-occupiers, and raised the commercial threshold to $2.1m. Each is defensible on its own. Together, they make the remaining duty base narrower and more concentrated. That makes the last stretch of abolition harder and brings back the lumpiness the reform was meant to fix.

What to take from it

The economics held up. This is the most important fact, and it keeps getting overlooked. Fourteen years in, with a full independent evaluation, no one has shown that the swap hurt Canberra’s economy, property market or lower-income households. The predicted disasters never turned up. The modest gains did.

The politics held up better than anyone expected. The reform survived four territory elections. That is encouraging for a policy everyone assumes is electoral suicide, and it gets little coverage because it is the absence of a story.

The timetable failed. That was about fiscal discipline, not tax design. A twenty-year transition maximises the years voters feel the new tax and minimises the annual benefit anyone can point to. It also gives a government eighteen chances to use the new base for something else. Victoria’s commercial and industrial property tax is faster and more straightforward. There, a property pays duty one last time at its next sale and then starts a ten-year clock. This takes the government’s discretion over the pace off the table. Three design rules follow. Revenue neutrality needs to be legislated and audited, not just promised in a press release. The base should be flat and broad, with no extra charge on the tenure you want to encourage. The transition should be event-driven, not calendar-driven. That way, it cannot be slowed by a treasurer having a rough year.

The transition trade is narrower than it looks. In Canberra, turnover rose, and prices held up. The losers were not all property owners, only those caught by a surcharge: residential landlords, who paid rates on top of land tax. So the question in the next state is not whether property is exposed, but which owners the design hits. That is politics, not economics, and it will vary by jurisdiction.

The ACT’s interest bill is set to rise from $811m in 2026-27 to $1.2bn by 2029-30. The territory has gone from net creditor to carrying a hefty debt. None of that is due to the tax reform, which was revenue neutral by design. It is down to spending. But it is why the substitution has stalled. A government under fiscal pressure will not give up a working revenue line. State credit metrics are a good early warning for whether a promised transition will actually happen. Keep an eye on Victoria. The commercial and industrial transition is underway, with more than 12,000 properties in the mix at the last budget. Infrastructure Victoria has put residential land tax back on the agenda. If a big state moves on residential, the ACT stops being a curiosity and starts being a template. The repricing of land-heavy assets stops being hypothetical.

Fourteen years ago, Canberra asked a question nobody else dared. The answer was mostly positive. The experiment is stalling not because of the answer, but because a stable, unavoidable revenue base is the most useful thing a treasurer can have and the hardest thing to part with.

They did not touch land.

The demand-side tweaks got all the political airtime. The land question is the one that actually matters. It sits quietly in the state and local revenue base, waiting for someone to notice, but no one seems in much of a hurry.

That is the real reform gap. Australia has shown it can tinker with housing tax perks when the mood takes it. The bigger prize is sitting with the states: move revenue away from taxing transactions, payroll and capital, and toward the unimproved value of land.

The idea, briefly

Henry George published Progress and Poverty in 1879. He argued that the community creates the value of a location. Roads, schools, stations, neighbours and nearby demand all add value. That value, he said, should go to the community, not just to the title holder. His policy was a tax on the unimproved value of land, set high enough to replace taxes on labour, trade and consumption.

The economics behind it are older than George and have survived every school since. Adam Smith made the case in The Wealth of Nations. He noted that a tax on ground rents falls entirely on the owner and discourages no industry. Milton Friedman called the tax on unimproved land value the least bad tax. Joseph Stiglitz formalised the Henry George theorem: in an efficiently sized city, land rents raise exactly enough to fund the optimal level of public goods.

The mechanism is not exactly quantum physics. Tax labour and people work less or move. Tax capital and it disappears. Tax transactions and they dry up. Tax land and, well, the land is still sitting there, looking unbothered. The supply is fixed. That is why the tax cannot be shuffled onto tenants as higher rents, no matter how hard anyone tries.

Australia is hardly a stranger to all this. Victoria gave land tax a whirl in 1877. South Australia followed in 1884. The Fisher government brought in a progressive Commonwealth land tax in 1910, which lasted until 1952. Deakin, Fisher and Hughes all backed George’s ideas. South Australia even had a Single Tax League. So this is not some imported novelty. It is a path Australia once took, then quietly wandered away from, whistling as it went.

Marginal Excess

In March 2015 the Treasury published its tax discussion paper, Re: think. Buried in chapter two is a chart of the marginal excess burden of Australia’s major taxes, which is to say the economic cost imposed for each additional dollar of revenue raised.

The chart does not beat around the bush. The taxes Australia leans on most are often the ones that do the most harm, while broad-based land taxes are at the friendlier end. Stamp duty is the worst offender, costing about 80 cents in lost activity for every extra dollar raised. Company tax is next in line. GST and income tax are somewhere in the middle. Broad-based land taxes and municipal rates sit at the bottom, below zero. Treasury’s logic is simple: land cannot run away, so you can raise revenue without scaring off work, investment or transactions. In the model, some of the burden lands on foreign owners and is handed back to locals. The upshot is that raising a dollar this way actually leaves Australians better off before the dollar is even spent.

Nothing else on that chart manages this trick. Treasury pointed it out eleven years ago, and the world barely blinked.

The same paper notes that more than half of the potential land tax base by value is exempt from state land taxes, mostly owner-occupied housing and primary production land. Australia’s recurrent taxes on immovable property sit above the OECD average as a share of total taxation, but at roughly half the share collected in the United States, the United Kingdom and Canada. In 2013-14, land tax raised about 9 per cent of state tax revenue. The states fund themselves by taxing jobs and house moves. The one thing they could tax that cannot dash for the exit, land, gets off lightly.

Productivity benefits

Stamp duty is a tax on moving things to where they belong. When a worker stays put instead of moving closer to a better job, or a family stays in a house that no longer fits, or a retiree skips downsizing because the transaction cost eats the benefit, the economy ends up with people and houses in all the wrong places. It is a game of musical chairs where no one wants to get up.

Infrastructure Victoria put a number on it. Victorians spent about $12bn moving house in 2022-23. About $3bn was the actual cost of moving, including agents, conveyancers and removalists. The other $9bn was stamp duty. That is an effective tax rate of about 300 per cent on the real cost of relocating. Not many taxes can claim that sort of mark-up.

The evidence points in the same direction across jurisdictions and methods: replacing stamp duty with a broad land tax lifts activity by making housing and labour move more freely. Independent Economics and the 2011 NSW financial audit both estimated that swapping conveyance duty for a broad-based land tax would raise long-run economic activity by about 1.3 per cent. Grattan put the national gain at up to $17bn a year. Victoria University’s Centre of Policy Studies found the swap adds about 30 cents of national income per dollar of revenue shifted. A 2025 paper in Macroeconomic Dynamics found welfare gains of 3.57 per cent from reducing stamp duty.

Australia has spent five years holding summits about productivity. Here is a reform where Treasury’s own model says raising revenue actually pays you. Infrastructure Victoria puts the implementation cost somewhere between $1m and $5m, which is a rounding error in government terms. The productivity case is the familiar one. The less familiar case is competition: land tax changes who gets to play, not just who gets to pay.

Competition

This argument gets the least airtime and deserves a lot more. It is the wallflower at the tax reform dance.

Land price is a barrier to entry. If a good site costs a fortune to buy but next to nothing to hold, then the main qualification for competing is having a fat balance sheet, not running a better business. Site control becomes a competitive weapon. A big operator can sit on a corner it does not need to keep a rival out. It is Monopoly, but with real money.

Shift the burden from the upfront price to the annual holding cost and the logic flips. Land gets cheaper to buy and pricier to sit on. Capital that was locked up in the entry ticket is freed for equipment, staff, inventory and software, the things that actually generate returns. Suddenly, the barrier to entry drops for the operator with a better idea and not much spare change.

There is a second competition effect in how Australia levies land tax. State regimes are progressive on aggregate holdings within a state. The tax rate rises with the size of the portfolio. The OECD has pointed out, and the Treasury paper repeats, that this biases against large-scale residential investment and discourages institutional investors from the private rental market. Australia’s shortage of institutional-grade rental housing is partly a tax design choice.

Rent seeking

George’s main point was that land value is created by society and pocketed privately. Build a metro or train line, and the windfall goes to whoever owns land nearby. Rezone a block, and its value can double overnight. The rational move for anyone with capital is to spend it on chasing rezoning, planning influence, or getting ahead of infrastructure news.

That is rent seeking in its purest form, and it is not a moral failing of developers. It is simply the logical response to the incentives. When lobbying pays better than building, money knows where to go.

A land value tax scoops up the uplift automatically, as soon as it appears, with no need for planning gain negotiations, value capture levies or voluntary agreements. It takes away the prize. It also changes what banks do. When land appreciation is taxed as it happens, mortgage lending against land looks less appealing. Australia’s housing debate has become all about supply, and supply does matter. But the shortage is also a problem of holding costs, which gets less of a look-in.

Australia’s residential dwelling stock was worth $12.8 trillion as of March 2026. Most of that is land. On the last detailed breakdown, total Australian land was worth $9.2 trillion. Residential land was $7.7 trillion, or 84 per cent. When land is the asset and the building is just the depreciating inconvenience attached to it, the rational owner is more interested in holding the land than actually using it.

With a land value tax, an underused block in an established suburb gets a bill every year, no matter what sits on it. The owner either develops it, sells it to someone who will, or pays the tax from elsewhere. Sitting on a vacant lot in a good spot and waiting for the community to do the work starts to cost real money instead of being a free ride.

The transaction effect matters too. ACT data found that every 10 per cent cut in stamp duty rates led to a 6 per cent jump in transactions. More turnover means households are better matched to homes. This is a kind of supply that does not need a single new brick.

What the sceptics are right about

A land value tax is only as good as the assessment of unimproved value, and Hayek’s reservation was precisely that assessment disputes would produce unfair outcomes. Australia has an advantage here, in that state valuers-general already assess unimproved values annually and municipal rates already run on them. The infrastructure exists. It is not costless, but it is not novel.

Then there is the asset-rich, cash-poor household. This is a real issue, but it can be solved. The ACT and South Australia already have deferral schemes that work like a reverse mortgage. Eligible owners can let the liability build up against the property until it is sold. A deferred land tax is, in effect, a claim on the estate. That is what makes it a political headache.

Someone who paid $40,000 of stamp duty last year and then faces an annual land bill has been taxed twice. The ACT’s response was a twenty-year phase-in, beginning in 2012 and still running. Victoria’s commercial and industrial property tax, live since 1 July 2024, takes a different route. A property pays duty one final time on its next sale, then enters the annual regime after ten years. More than 12,000 properties had entered by the 2026 budget. It is a template.

There is also the federalism problem. This is the real reason nothing happens. The states own the base and wear the transition cost. The Commonwealth holds the purse strings and collects much of the growth dividend. Until someone bridges the revenue gap in the changeover years, a reform that almost every economist cheers on in theory remains stuck in practice.

Why asset allocators should care

For allocators, the investment point is simple. A real shift to land value taxation would squeeze land values and boost transaction volumes. Assets that rely on land appreciation rather than operating income would reprice downward. Assets that get their returns from the improvement, the business or the tenant covenant would look better than the dirt underneath. Development models built on sitting on cheap land for a decade would start to wobble. Models built on turnover, productive use and operating yield would get stronger.

The listed and unlisted property universe is not all the same here. Land-heavy, low-yield holdings are exposed. Improvement-heavy assets, and anything where the earnings come from what happens inside the building, are better placed. Institutional build-to-rent, currently penalised by progressive land tax, would benefit from the direction of travel. The May budget showed that the political constraint on housing tax reform is softer than it looked eighteen months ago. Victoria has started on commercial and industrial land. The ACT is fourteen years into a twenty-year transition, and the price collapse its critics forecast has yet to show up. Infrastructure Victoria has put residential land tax back on the table in a thirty-year plan. None of that is a reform.

The tax that cannot be dodged, shifted or sent offshore is sitting there, mostly exempt, under $7.7 trillion of residential land. Henry George spotted this in 1879. Treasury did the sums in 2015. The question is not whether the argument stacks up. It is who pays for the transition, and how long Australia is willing to keep taxing the things that can bolt for the exit instead of the one thing that cannot.

The valuation critique is fair. SpaceX lost US$4.3bn last quarter and has an accumulated deficit of US$41bn. It is priced at about 94 times revenue. Morningstar’s fair value estimate is US$780bn, less than half the US$1.77 trillion float valuation. But the bagholder argument misses the main point. It assumes retail investors will buy the unlocked stock. Most buyers will not be retail. They will be institutions buying because their rules require it. Often, these are the same institutions whose private-market arms are selling. Australian superannuation members are on both sides of this trade, often without realising it. Their funds’ statutory disclosures show this. I have reviewed them.

The float that was built for the index

The mechanics are the story. SpaceX sold 555.6m new Class A shares at US$135 on 11 June. It raised US$75bn, the largest capital raise in history. No insider sold shares in the offer. With about 13.08bn shares outstanding, the free float at listing was 4.3%. Every price since the US$161 first-day close, the US$225.64 intraday peak, and the drop below issue price was set within that small float.

The lock-up structure is unusual and not well understood. Instead of a single 180-day lock, the prospectus sets out a staged release. Insiders can sell up to 20% of eligible locked shares from the second full trading day after the first quarterly result as a public company, expected in early August. Another 10% is released if the stock closes at least 30% above the issue price of US$175.50 on five of the ten trading days into that result. Then 7% unlocks on each of these days after listing: 70, 90, 105, 120, and 135. These dates are 21 August, 10 September, 25 September, 10 October, and 25 October. Another 28% unlocks after the September quarter result. The rest is released at day 180, on 8 December. Only Elon Musk’s 6.4bn shares and some unnamed institutional block holders are locked for the full 366 days, until June 2027, with no early release. If the US$175.50 performance hurdle is not met, and at current prices it will not be, the 10% tranche is not lost. It moves into the later time-based releases. The supply comes anyway. The condition only changes the timing.

Why was it built this way? CNBC reported the answer at the time. Nasdaq changed its rules on 1 May to allow fast entry into the Nasdaq 100 for new listings with a market capitalisation above the index’s 40 largest members. The seasoning period is now as short as 15 trading days. A faster unlock schedule builds float quickly, which is needed for index inclusion. This is what the retail framing misses. SpaceX’s index weights are based on float-adjusted capitalisation, which started at about US$90bn, not the US$1.77 trillion headline. Each tranche released by early holders increases the amount index-tracking funds must buy at the next rebalance. The unlock calendar and the index weight calendar are the same. Sellers do not just find buyers. The structure forces them, at whatever the price is, on dates set in the prospectus.

Governance shows who runs this company. In April, two months before listing, a charter amendment created a dual class structure. For 24 years, all investors held one class of stock. Now, Class B holders get 10 votes per share and the right to elect most of the board as long as any Class B shares remain outstanding. Mr Musk holds about 92% of Class B and about 82% of total voting power. Removing him as chief executive or chairman needs a separate majority of Class B, which means his own consent. The company reincorporated in Texas under rules that make tender offers, proxy contests, and shareholder proposals harder to execute. The board approved up to 200m additional super voting shares tied to a US$7.5 trillion valuation and a one-million-person Mars colony, reportedly without an independent compensation committee. IPO participants had to waive the right to sue to get an allocation. None of this is hidden. All of it is, or should be, in the price.

Who is actually selling, and what they have told us

The sell side is not hypothetical. In several cases, it has made its intentions public.

ARK’s Venture Fund, which made SpaceX its largest holding and bought a further 3.3m shares in the days before listing, published an IPO guide stating that during the lock-up new inflows will be deployed into other private companies, and that once the lock-up expires the fund will have “full flexibility” to manage the position, including opportunistically reducing exposure and recycling proceeds into the next generation of private names. That is a distribution plan in plain sight: SpaceX is the liquidity event that funds the next vintage.

Baillie Gifford’s four trusts are the best documented case of an institution on both sides at once. Edinburgh Worldwide, US Growth and Schiehallion all trimmed SpaceX into the December 2025 tender at the US$800bn valuation, the same mark that reset carrying values across the industry. The firm then lifted its own valuation of the company three times in six months, from the 87% December write-up to US$1.25 trillion after the xAI merger rebased secondary prices, to roughly US$1.6 trillion when the float priced, each step disclosed to the market and each step below the deal mark, on the stated principle that it values off verifiable transactions rather than press speculation. Scottish Mortgage now carries SpaceX at 25.7% of assets; its private holdings have breached the trust’s 30% cap, forcing a shareholder vote to authorise further private investment despite the breach, and Schiehallion’s manager said on listing day that the question for the trusts is “what is the right position size” as the lock-ins ease. The manager’s public commentary remains committed to the name. The trust’s own arithmetic is a seller on any calendar. Meanwhile Scottish Mortgage’s register shows the rotation in real time: the shares fell more than 5% around the IPO and swung to a discount as investors pulled money from the wrapper to buy the listed line directly.

Fidelity, the largest single institutional holder with SpaceX at 4.7% of Contrafund and material weights in two other flagship funds, positions built from 2015 at a US$10bn valuation, has said nothing. As open-ended funds, their marks now simply track the tape, which, below US$135, means the largest holder in the market is carrying the position underwater relative to the float price.

The Australian leg: your members are on both sides

Now to the part that concerns readers directly. In the past week, I reviewed the 31 December 2025 portfolio holdings disclosures of AustralianSuper, Australian Retirement Trust, Aware Super, UniSuper, and Hostplus’s latest manager allocation disclosures. Two findings matter.

AustralianSuper’s statutory disclosure lists, under private equity in its High Growth, Conservative Balanced and Stable options, a holding in X.AI Corp, alongside OpenAI and Anthropic. That disclosure is dated 31 December. In February, SpaceX absorbed xAI in an all-stock deal. The country’s largest super fund held, in options including those its members retire in, paper that became pre-IPO SpaceX equity. ART told the Guardian its members’ exposure to SpaceX was about A$15 per member. ASFA put the system average near A$50. ART’s chief investment officer was reported by the AFR as braced for aftershocks. Hostplus discloses the sector’s largest private equity programme by proportion, 8.5% of its trust, through vehicles that carried late-stage SpaceX and xAI secondaries worldwide: Lexington co-investment funds, HarbourVest, Hamilton Lane, Partners Group secondaries, Pomona, Siguler Guff.

Second, members could not have seen any of this. Every fund’s disclosure stops at manager or vehicle level. Aware lists 48 external private equity managers and not one underlying company. ART lists about 61. UniSuper, the control case, invests almost no capital in private equity. Only AustralianSuper publishes a company-level look-through; even then, the list contains no values and cannot be linked to a manager, and the fund says some private equity managers do not allow disclosure. The December files are a floor on the system’s exposure, not a ceiling. A member reading their fund’s statutory holdings file on 31 December could not tell if their savings held SpaceX. The fund could.

On the other side is the buy side, which is not hidden. Aware’s international shares option tracks a custom MSCI World ex-Australia index. ART’s diversified options hold about 4,000 listed lines. This is benchmark breadth, not conviction. Hostplus’s largest international sleeve is IFM’s Indexed Global Equities, at 7.7% of the trust. Its Australian core is an enhanced index. Its member-choice-indexed options surpassed A$21bn last June and continue to grow. AustralianSuper’s Indexed Diversified option holds 1,922 listed lines and no unlisted assets. As SpaceX enters the Nasdaq 100 under the fast entry rule and then the MSCI family, and as each unlock increases its float, these books buy. Not because anyone chose to buy at 94 times revenue, but because replication is the rule. ART’s head of investment strategy put it plainly to the Guardian: new money does not get created. The key question is where the money flowing into SpaceX, Anthropic and OpenAI comes from. Some of it, as shown above, comes from the accounts of the same members whose private equity sleeves are selling.

The LPs pay twice

This brings us to the economics, and to who wins. The general partners have already banked their winnings in two ways that no post-IPO price can take back. First, realised carry. The tender chain, from US$210bn in mid-2024 to US$350bn that December and US$800bn a year later, was not just a valuation ladder. Each tender was a real sale. Funds that sold converted marks into cash and locked in performance fees at each step. Carry on those proceeds is earned and paid, whatever SPCX does now. Second, management fees. Where fees are based on net asset value, which is common for evergreen vehicles, pre-IPO feeder funds and many fund-of-funds layers, every mark-up in the chain increased fee income for years, at both the fund-of-funds layer and the layer below. None of it is refundable.

To be fair to the GPs, carry on stock still locked up is different. It will be set at sale or distribution prices through the unlock windows, not at the US$800bn mark. Managers holding residual positions have real exposure to the December tape. The flat-fee listed managers, Baillie Gifford and Fidelity among them, earn no carry at all. But the asymmetry stands.

The GP’s risk on the residual is a smaller bonus. The limited partners face more uncertainty, because their risk is the capital itself. Their payouts come last. They get cash if managers sell well through the unlock windows into the index bid. They get stock if managers distribute in kind into a falling market. In both cases, prices are set months after the marks that set years of fees. The video is right that someone overpaid. It just blames the wrong group. The real loser here is not the day trader who bought at US$225. It is the fee-paying LP and the index-tracking member, whose participation was never a choice.

Predictions

First, supply dominates until December. The performance hurdle fails, the 10% rolls forward, and about two-thirds of the locked shares become saleable between August and late October as index demand steps up. The stock spends most of that time below its issue price, with rallies around each unlock as index demand resets. The 8 December full release is likely the capitulation point and a better entry for those interested.

Second, expect supportive corporate news just before unlock windows from holders with stakes to defend. The video is polemical on this but right to mention it. Alphabet’s pre-IPO compute contract set the example. Watch for announcements around 21 August, 25 September and 8 December.

Third, the 30 June 2026 portfolio holdings disclosures, due by late September, will be the first statutory documents where Australian members can see SpaceX. It will appear as a listed equity line at market weight after the mechanical buying. Expect several funds to hold SPCX in their indexed and benchmarked sleeves at a higher dollar value than their private programmes ever held. That is the thesis of this note in one filing. I will publish the comparison when the files are released.

Fourth, the governance discount appears slowly. A controlled company with a 366-day insider lock, an unappealable board and a US$25bn bond stack with widening spreads does not re-rate on Starship launches alone. The valuation debate will still be active when Mr Musk’s own lock expires in June 2027. That is when the real overhang discussion starts.

The exit was engineered well. So was the bid. What nobody engineered is a reason, at these prices, to be either.

This note is general information only. It does not consider any reader’s objectives, financial situation, or needs, and it is not a recommendation regarding any financial product. Figures are drawn from the SpaceX prospectus and related filings, fund disclosures and cited press reporting as at 20 July 2026, and may contain errors; verify independently before relying on them.

This scrutiny is justified. ASIC’s findings are real problems. But neither regulator asks who designed the system. The growth of retail and superannuation exposure to private markets, and the weaknesses now under review, are largely the result of regulatory choices. Three stand out.

Exhibit one: APRA priced the banks out of development finance

APS 112, which began in January 2023, set a 150 per cent risk weight for land, development and construction loans. Residential development loans can get a lower 100 per cent weight, but only if pre-sales cover the full debt. Few projects meet this test when they need funding. For banks, this makes development lending very capital intensive. At 150 per cent risk weight, a development loan must earn much more than a standard corporate loan to deliver the same return on capital.

Banks did what the capital rules intended. They pulled back. But property development finance did not stop. Demand for housing remained. Funding shifted to non-bank lenders and private credit funds. These vehicles hold no regulatory capital and have no prudential supervisor. They now raise money from superannuation funds, family offices and, through feeder structures, retail investors.

This was not a market accident. It was the result of policy. APRA seems to recognise this. In June 2026 it began consulting on easing these rules, proposing to cut the pre-sales requirement for the 100 per cent risk weight from full debt coverage to 50 per cent. The aim is to let banks lend more. Regulators do not change rules that work. The consultation is an admission: the 2023 settings pushed development credit out of the banking system and into the hands of the unsupervised. The exposures now under review are the ones the capital rules sent them to.

Exhibit two: ASIC licensed the funds but never told them how to value anything

Valuation sits at the centre of ASIC’s private credit findings. Report 820 identified funds with incomplete valuation policies, infrequent valuations, committees that both approved loans and then valued them, and construction loans carried at “as if complete” values that flattered the book. The regulator’s June 2026 statement described valuations as an immediate point of action across private markets.

This finding is awkward for ASIC. It licenses responsible entities and registers retail managed investment schemes. It has done this for decades. Yet it has never issued clear rules on how to value unlisted assets in these schemes. There are no set frequencies, no independence rules, no standards for methods, and no guidance on stale or model-based marks. On the superannuation side, APRA’s SPS 530 at least requires trustees to have a valuation governance framework. Fund managers running registered schemes with private loans have had to rely on general fiduciary duty and accounting standards. In practice, they have been left to decide for themselves.

Industry valuation practices may be poor. But ASIC should not be surprised. When a regulator gives no guidance for twenty years, it is no shock that practice ranges from good to bad. This is not just a problem with the industry. It is a result of the lack of guidance. Funds that set up independent valuation committees did so by choice. Others were never told to do so.

Exhibit three: the retail-into-wholesale plumbing that ASIC permitted

The third design choice is the least discussed and arguably the most consequential. Australian law allows a registered retail fund to invest all of its assets substantially in an underlying wholesale fund. ASIC has long permitted these feeder arrangements, and they have become the standard architecture for distributing private credit to retail investors and self-managed super funds.

This structure has two effects that need more attention.

The first is regulatory. The retail investor protections, disclosure, design and distribution obligations, and ASIC’s product intervention powers attach to the feeder. The assets, valuations, related-party loans, and fee mechanics sit one level down, in a wholesale vehicle designed to be outside the retail regime. The investor holds a retail wrapper around a wholesale reality.

The second is commercial. The ban on conflicted remuneration introduced after the Future of Financial Advice reforms applies to retail financial products. It does not reach arrangements struck at the wholesale level. A wholesale fund can pay rebates, fee shares, and distribution payments to platforms, dealer groups, and intermediaries in ways that would be prohibited, or at least require disclosure, if made one level up. The feeder structure does not merely permit this. In some corners of the market, it is the reason the structure exists. Money flows to the wholesale layer as management fees, and flows back out to the people who gathered it, invisible to the end investor whose statement shows only the headline fee of the retail feeder.

ASIC knows about this. Its 2026 surveillance priorities include fees, margin structures and conflicts in wholesale private credit funds, and the distribution of private credit to retail clients. Its progress updates call for law reform to strengthen the wholesale funds regime. The architecture now under review has been visible for years, built on permissions granted by ASIC. The wholesale investor test has barely changed since 2001. The feeder structure was never banned, never restricted, and never required look-through disclosure. What changed is not the structure. What changed is its size.

None of this excuses the industry

Let me be clear about what this argument is not. It is not a defence of managers who valued construction loans as if the building were finished, or who let the committee that wrote the loan mark the loan. Managers who took advantage of a permissive environment own their conduct.

A market review that looks only at participants, and not at the incentives they faced, will not fix the problem. If APRA’s capital rules pushed development credit into unsupervised vehicles, the solution is to change the rules, not just watch funds more closely. If poor valuation practice is due to a lack of standards, ASIC should set a standard, even if it is late. If feeder structures exist to avoid the conflicted remuneration ban, the answer is to regulate the wholesale layer or change the wholesale test, not just issue stop orders on product disclosure statements.

Regulators often say that trust and confidence must be restored in private markets. That is true. A good start would be an honest account of how the system was built, including the parts designed by the regulators themselves.

The first piece is Craig Shepard’s client note on First Samuel, “Taxation Without Alignment.” Two hundred and fifty years ago this month, the American colonies went to war over a slogan: no taxation without representation. Shepard marks the anniversary by arguing that Australia’s problem is a quieter cousin of the same disease. We are represented, all right. But the level of government that collects the taxes which grow with the modern economy is not the level of government that must deliver the services, build the infrastructure and carry the debt that growth creates. Taxation without representation caused a revolution. Taxation without alignment, as Shepard puts it, causes something slower and more corrosive: underinvestment, resentment and drift.

The second is the Policy Institute Australia’s recent report on means testing, Home Truths, which underpins the ABC’s recent discussion on budget balance. Taken together, these analyses indicate that reform should focus on tax design and spending efficiency, rather than on austerity or population controls. This perspective warrants broader consideration among professional allocators, which is the purpose of this review.

It is important to clarify at the outset that this analysis does not concern immigration policy. On any long-term assessment, population growth is a net positive for the Australian economy, deepening the labour market, supporting demand, and mitigating demographic ageing. The issue lies in the fiscal architecture: migration is primarily a federal policy lever, yet its infrastructure consequences are borne by state balance sheets. Revenue generated by new arrivals accrues predominantly to the Commonwealth, while the associated costs, such as schools, hospitals, transport, and policing, fall to the states. A growth strategy layered onto a revenue-sharing model established in 1942 inevitably produces the fiscal strain now evident. Reform must address the system itself, not the intake.

The Commonwealth owns everything that grows.

Start with what grows. Company profits have risen materially as a share of GDP since the 1990s, with much of the rise tied to the resources boom, which lifted mining profits and shifted aggregate income toward the corporate sector. Company tax is a Commonwealth tax. As Shepard notes, the physical location of the mine, the legal location of the taxpayer, and the population pressure created by national economic policy are three distinct factors.

Personal income tax is the same story with a sharper edge. In a progressive system, bracket creep is not an accident; it is the design. Wages rise, nominal incomes rise, and the federal government collects more than proportionate revenue without ever announcing a tax increase. The Parliamentary Budget Office projects that personal income tax will make up nearly 54% of total tax receipts by 2032-33. Despite the tax cuts of the 2000s, the share of income tax in gross household income continues to climb. Every dollar of it is federal revenue.

The aggregate result is one of the largest vertical fiscal imbalances of any federation in the world1. The Commonwealth raises more than 80% of the federation’s taxes and controls the most important tax bases: income and consumption. On the Forum of Federations’ figures, it needs only around 61% of the pie to meet its own spending, while the states raise just 17% of revenue but require roughly twice that, about 33%, to meet theirs. The imbalance is largely the product of the Commonwealth’s takeover of income tax in 1942 and High Court rulings that struck down various state taxes. It has never been redesigned since, only patched.

The states, by contrast, are left with what Shepard calls narrower, uglier taxes. Payroll tax grows with wages but taxes employment. Stamp duty rises when property changes hands but is volatile and punishes mobility. Land tax is the economists’ favourite and the politicians’ nightmare. Fines and service charges raise money but are not growth taxes. The only broad state-shared tax that naturally grows with nominal spending is the GST, collected federally and distributed through a federal formula. The Henry Review counted at least 125 taxes in Australia, with 90% of revenue raised by just ten of them. The state-controlled entries on that list are consistently ranked among the least efficient taxes in the economy.

For two decades the misalignment mattered less than it does now, because Australia was rich enough to avoid hard choices. The terms of trade rose from a low of 47.5 in 1999 to a peak of 144.2 in 2022, and were still around 117 in the March quarter of 2026. That income boom did a lot of work. It allowed lower effective tax rates, funded catch-up wages for teachers and nurses, held down deficits and interest rates, and lifted asset prices. For almost two generations, prosperity without investment looked sustainable.

The pull-forward

It wasn’t, not entirely. Shepard’s sharpest section describes the past generation as, in part, a pull-forward. The generation that owned the assets received the compounding benefit of lower rates and higher prices. It also inherited infrastructure built by earlier generations: railways, ports, schools, hospitals, water systems. In Victoria’s case, that inheritance reaches back to the 1890s. The replacement cost was never paid as it accrued. The bill was deferred.

Then the population surge arrived, concentrated as usual in the two big eastern cities. In 2024, net overseas migration added around 107,000 people to New South Wales and 101,000 to Victoria. In 1996, the equivalent figures were less than half of those for NSW and less than a quarter of those for Victoria. No wonder, as Shepard observes, we didn’t build then.

Population growth drives housing construction, but it also requires trains, roads, hospitals, schools, police, water, parks and power. And here is the sentence that should be pinned above every federation reform working group: migration is largely a federal policy lever; the infrastructure consequences are state balance-sheet events.

Victoria: catching up, not overbuilding

Victoria’s Big Build, and the associated debt, should be assessed through this lens. While criticism of the costs is understandable, given inflation driven by weak productivity, supply constraints, industrial relations challenges, and suboptimal procurement and policy, the underlying requirement remains legitimate. Much of the expenditure reflects deferred replacement of legacy infrastructure and the additional capacity necessitated by three decades of sustained population growth.

Shepard’s numbers put a scale on it. On First Samuel’s estimates, Victoria needed to spend around $26bn a year for a decade simply to catch up on depreciation and the capital demanded by ten years of net overseas migration, a figure struck before the surge in construction costs. Victoria is currently spending about $21.4bn (2025-26), falling to an average of roughly $16.5bn across the forward estimates. NSW is in the same world: a $116.7bn four-year infrastructure programme, more than $30bn of it in 2026-27 alone. In other words, and this is the uncomfortable conclusion, Victoria is not overbuilding. Even at allegedly inflated prices, it is only just catching up, and the forward estimates show it giving up the chase.

The resulting budget arithmetic is now familiar. Net debt is expected to hit $175.6bn by the end of next financial year and top $199.3bn by mid-2030, by which point annual interest repayments are forecast to reach $11.8bn, more than $32m a day. The 2026-27 budget’s $1bn operating surplus captures only day-to-day revenue and expenditure; once debt-funded capital spending is included, Victoria is forecast to record more than $30bn in cumulative cash deficits over the next four years. As Saul Eslake put it, until you start running cash surpluses you can’t begin to repay debt; indeed you can’t stop adding to it. Victoria is already paying roughly $9bn a year in interest to fund a catch-up whose benefits (a bigger workforce, higher national output, more federal income and company tax) accrue substantially to the Commonwealth.

Why does a state in this position keep reaching for property and payroll taxes? Because, as Shepard lays out, a fast-growing state without access to the federal government’s growth taxes has exactly four choices: tax property harder, lift fees and charges, increase payroll tax, or borrow. The first is politically brutal. The second is regressive. The third is economically distorting. The fourth is inevitable. Victoria has done all four. It has lowered land tax thresholds, added COVID debt levies on property and payroll, expanded the windfall gains tax and layered on surcharges, and it has still ended up with the debt. That is vertical fiscal imbalance in practice, and it is the mechanism by which fiscal federalism shows up in your property portfolio returns and in the TCV-ACGB spread (the extra yield investors demand to hold Victorian government bonds over Commonwealth bonds of the same maturity).

The GST carve-out that made it worse

The system’s shock absorber was supposed to be horizontal fiscal equalisation: distributing the GST so that each state could deliver a broadly comparable standard of services, accounting for their different revenue bases (including mining royalties) and costs. Then came the 2018 GST changes. What looked like a technical adjustment of floors, relativities and transition payments has become one of the most consequential fiscal decisions in modern Australia.

Western Australia’s mineral royalties meant the original formula would have cut its GST share hard; instead, a floor was negotiated. The Grants Commission now estimates WA will receive around $6.6bn more in 2026-27: $5.5bn in “no worse off” payments plus $1.1bn in pool top-ups. WA’s GST distribution comes to $9.3bn against $26.1bn for NSW and $27.9bn for Victoria, but WA also banks around $11bn in mining royalties on top. Saul Eslake’s summary, “heads Western Australia wins, tails the other states and territories lose”, is hard to improve on; he has called it possibly the worst Australian public policy decision of the century so far.

The most striking aspect, as Shepard observes, is the disparity in fiscal capacity. Western Australia can fund electricity subsidies and invest beyond its infrastructure requirements, with approximately $6.5bn in necessary capital expenditure compared to over $12bn invested. Victoria, by contrast, requires $26bn and manages $21bn. On First Samuel’s simple citizen’s ledger (services delivered, minus contribution, plus capex and subsidies, minus GST returned and payroll tax), a West Australian nets out roughly $4,500 a head in their favour and a Victorian roughly zero. Once land tax, stamp duty and other state charges are added, the Victorian falls about $4,000 behind while the West Australian stays $1,000 ahead. The numbers are deliberately simple, but the point stands. Debt and GST inequity are not abstractions. They shape services, investment, and growth, and increasingly they determine where capital should be deployed.

The other side of the ledger: the Commonwealth’s targeting problem

If the states’ problem is that they cannot capture growth revenue, the Commonwealth’s problem is the mirror image: it captures the revenue and then sprays a growing share of it at households who do not need it. Shepard gestures at this. The federal government has no broad wealth tax, limited means of capturing asset-price uplift, and a grandfathered CGT discount that lightens the load when gains are finally realised, even as the asset-price boom delivered one of the largest wealth transfers in modern Australian history. The Policy Institute Australia’s June report, Home Truths, the data source behind the ABC’s means-testing story, quantifies where that leaves the transfer system.

The headline numbers: across the Child Care Subsidy, Parental Leave Pay, Aged Care, and the Age Pension, the best-off 20% of households will receive about $25.6bn this financial year, around 20% of total spending on those programs. Better-targeted means testing could free up around $21bn a year, while building on existing rules rather than replacing them.

The distributional detail is where it bites. A retired couple with a $5m home in Sydney can qualify for the full Age Pension worth almost $50,000 a year, and up to $150,000 for one spouse in aged care, while their neighbours who rent and hold $1.5m in super receive no pension at all and roughly half the aged care support. Just 1% of households have income too high to qualify for either the Child Care Subsidy or Parental Leave Pay; a family earning $400,000 a year may qualify for around $20,000 in Child Care Subsidy and a further $20,000 in Parental Leave Pay.

The mechanism behind the drift matters for anyone who thinks about balance sheets for a living. Income is more equitably distributed than wealth in Australia, and our means tests are overwhelmingly income tests, with the largest single store of household wealth, the owner-occupied home, exempt from the pension assets test entirely. As the population has aged and housing wealth has compounded, rules calibrated for a poorer, younger Australia now deliver transfers to older Australians that have grown by around $25,000 per adult in real terms since the 1990s. This is the generational transfer Shepard describes, running through the spending side. The cohort that received the asset-price uplift, the lower rates and the inherited infrastructure is also the cohort whose entitlements the income-taxed young are now funding.

What could $21bn a year do? On the Institute’s modelling, it could fund an additional $5,000 in benefits to each of the 4.4m households in the bottom 40% by income or wealth; or cut every income tax rate by 1.7 percentage points; or substantially expand JobSeeker or Rent Assistance; or eliminate 70% of this year’s fiscal deficit. Their most interesting single proposal is the Retirement Contribution Scheme (ReCS), a HECS-style, opt-in scheme that allows asset-rich, income-poor retirees to defer the cost of self-funding their retirement against their home, rather than forcing a sale or leaving the taxpayer to underwrite housing wealth accumulation for the next generation’s inheritance.

Joining the two halves

Considered together, these two analyses present a coherent case for reform.

First, give the states a growth tax, or a share of one. The options are well-rehearsed: a state surcharge on the income tax base with a matching Commonwealth reduction; broadening or raising the GST with the proceeds genuinely tracking population-driven need; or a Commonwealth-funded bridge to let the large states execute the stamp-duty-to-land-tax transition that the ACT has proven viable over twenty years. Any of these converts state revenue from a bet on property transaction volumes into a claim on the growth the states’ own population intake generates. None of them can be self-funded by the states, which is itself a symptom of the imbalance: the states cannot afford to fix their own tax bases.

Second, Commonwealth spending should be more effectively targeted. The $21bn identified in Home Truths does not represent a reduction in the safety net, but rather a return to its intended purpose: directing support to those in need. Redirecting even a portion of these funds toward deficit reduction or tax rate cuts would address the most corrosive trend in the federal fiscal mix—the increasing reliance on income tax, projected to account for 54% of total receipts by the early 2030s, which places a disproportionate burden on younger and working Australians to subsidise older and asset-rich cohorts.

Third, it is necessary to recognise that this is fundamentally a productivity agenda. Stamp duty impedes labour mobility and the efficient allocation of housing. Payroll tax thresholds constrain business growth. Bracket creep penalises additional effort. Poorly targeted transfers impose costs on future taxpayers. Each of these outcomes results from policy choices, all of which can be addressed without altering migration settings. Population growth increases the value of correcting these distortions, as the benefits of reform scale with the economy's size. As Shepard writes of tax design compounding: a CGT discount here, a GST floor there, bracket creep left untouched, state taxes unreformed, infrastructure deferred for a generation. None of it looks decisive in isolation. Together, they explain why public finances feel tight in the middle of one of the great terms-of-trade booms in our history.

What this means for allocators

Shepard concludes with implications for markets that merit attention because they challenge prevailing pessimism. In his assessment, weak ASX performance is not directly attributable to these structural issues, but rather to the entrenched expectation that reform is unlikely. Share prices reflect anticipated change, not current conditions. The prevailing malaise is a function of the belief that substantive reform will not occur. Any credible progress on GST reform, a broader federal tax base, or an intergenerational settlement focused on productivity would improve long-term growth expectations and re-rate equities, even before implementation. At present, the market assigns almost no value to reform optionality—a fact that will become significant if and when that changes.

Three further practical takeaways:

1. State credit is fundamentally a structural issue rather than a cyclical one. Semi-government spreads, particularly in Victoria, reflect a revenue model that lacks the capacity for self-correction. Without federation reform, the default response for fiscally constrained state treasuries will be increased reliance on property and payroll taxes, further asset recycling, and higher debt levels. Any narrative of a state’s ‘return to surplus’ should be approached with appropriate scepticism regarding the distinction between operating and cash outcomes.

2. Property tax risk has become a permanent consideration in Australian real asset underwriting. Land tax settings in the growth states have been adjusted repeatedly, always in the same direction, and the fiscal pressures underlying these changes persist. The logic of the four available choices ensures that such adjustments will recur. Assumptions of sovereign-level tax stability in long-term property and infrastructure models should be discounted, and the per capita disparity between Western Australia and Victoria is a material factor in cross-state capital allocation.

3. The debate on means testing now presents investable opportunities. If the Home Truths direction gains traction (and the political economy of a HECS-style ReCS is far more palatable than putting the family home into the assets test), the second-order effects will affect retirement income products, home equity release, aged care operators, and superannuation drawdown behaviour. The debate is at the stage where the unthinkable becomes discussable. That is usually when it pays to start doing the work.

Shepard concludes with the Ship of Theseus analogy: incremental changes, though individually small, have cumulatively reshaped the fiscal landscape. As noted at the outset, I do not claim expertise in this field, and a weekend’s reading does not confer it. However, it is clear that the two issues discussed are, in fact, facets of a single problem. The federation’s revenue structure, established in 1942 under emergency conditions, has been subject only to incremental adjustments rather than comprehensive redesign. Means tests were constructed for a nation with far less housing wealth than exists today. Neither arrangement is immutable. Victoria’s budget position is simply the most acute intersection of these two design flaws. The appropriate response is not to attribute blame to new arrivals or to equate an operating surplus with genuine fiscal repair, but to realign revenue collection with population growth and to ensure that support is directed to those in need.

References

1. Shepard, C., Taxation Without Alignment: The Fiscal Imbalance Reshaping Australia, First Samuel Investment Matters, July 10th 2026 — https://firstsamuel.com.au/taxation-without-alignment-the-fiscal-imbalance-reshaping-australia/

2. Policy Institute Australia, Home Truths: The Case for Rebalancing Toward Better Means Testing, Auster, A., Williams, H., Parta, I. & Tarrant, N., June 24th 2026 — https://www.policyinstitute.org.au/home-truths/ (full report, executive summary and chart pack available at that page).

3. ABC News, “Means testing key programs could save budget billions”, July 11th 2026 — https://www.abc.net.au/news/2026-07-11/means-testing-key-programs-could-save-budget-billions/106901412

4. Victorian Government, Budget 2026-27, Budget Papers, May 2026 (net debt, interest and capital programme projections; as reported by AAP, May 5th 2026).

5. Cameron Harrison, “Victorian Budget 2026: Surplus or Illusion?”, May 2026. Operating surplus vs cash deficit analysis; Eslake, S., comments to AAP.

6. Parliamentary Budget Office. Bracket creep analysis and projection of personal income tax share of total receipts (~54% by 2032-33), as cited in Shepard (2026).

7. Commonwealth Grants Commission. 2026-27 GST relativities and WA no-worse-off estimates, as cited in Shepard (2026).

8. Forum of Federations, Australia: Equity, Imbalance and Egalitarianism. Commonwealth and state revenue and expenditure shares.

9. Australia’s Future Tax System Review (Henry Review), Final Report, 2010. Tax concentration and state tax efficiency findings.

10. The Conversation, “FactCheck: how much of Australia’s tax is collected by states and territories?”. Commonwealth share of tax collections (~81%).

11. Varela et al. (2025), per-capita transfer growth by age cohort, cited in Policy Institute Australia (2026).

12. NSW Review of Federal Financial Relations (Thodey Review), 2020. Stamp duty to land tax transition analysis.

13. ABS. Net overseas migration by state (2024 and 1996 comparisons), as cited in Shepard (2026).

Momentum has pushed prices far above normal levels. The biggest stocks now rise because they have already risen. Firms with strong finances and steady profits are left behind and trade at discounts. In America, this is clearest among AI-related companies. In Australia, the pattern is the same, though less focused. The similarities are plain.

What the factor data actually says

The data are clear. In early 2026, America’s main momentum fund returned 8%, beating the S&P 500 by three points. Quality returned 6%. Value lagged at 2%. This is not new. Momentum leads, quality is overlooked, and value trails. By some estimates, half of its recent gains are attributable to this narrow cohort. Momentum in the United States has ceased to be a diversified strategy; it has become a concentrated wager on a handful of AI-adjacent giants. Investors purchase what has already risen, which increasingly means only a few names, and so the cycle perpetuates itself.

J.P. Morgan Asset Management’s factor team recently noted that dispersion within the momentum factor is at its widest since 1990, with the difference in returns between the strongest and weakest momentum stocks now at its widest since 1990. Momentum is experiencing its most sustained period of outperformance since the dot-com bubble. Historically, such extremes have served as warnings rather than endorsements.

The other side of the ledger

Long-term investors should take note. Momentum has soared, but quality has lost ground compared to its own history and the wider market. J.P. Morgan says quality in America is now as cheap as after the dot-com crash and the COVID crisis. Those were the best times to buy quality in the last 25 years. The firm is close to turning positive on quality shares. Quality means firms with steady profits, strong returns, and little debt. These companies grow value over time and do not need to raise money when conditions are poor. In a sensible market, such firms would be expensive because they are reliable. Their current low prices are not a sign of weakness. It shows the market is focused elsewhere: on momentum.

Why this happened: the AI overhang

The reason is clear. We are living through a real shift in technology, and no one knows what AI will mean for the economy. Faced with this, money moves to the winners and avoids the rest. J.P. Morgan puts it this way: investors have moved to the winners and left the underperformers behind due to uncertainty around AI. Momentum is now how the market deals with not knowing. Instead of judging which firms will do well in an AI world, investors let prices decide.

This pattern can last, but not forever. Buying just because prices have risen ignores real value. With so much money in a few stocks, any change could be sharp. The trade is crowded.

The Australian version: a bank instead of a chatbot

Australia shows the same pattern, but with different names. In America, momentum means AI and chip firms. In Australia, it is clearest in the Commonwealth Bank.

CBA is now the biggest stock on the ASX, making up about 12% of the index. Its share price over the last 18 months is a clear case of momentum. By mid-2025, it traded at about 31 times earnings, almost double its ten-year average of 16 and three times the global bank average of 10. For comparison, JPMorgan Chase trades at about 13 times earnings. CBA’s local rivals, ANZ, Westpac, and NAB, trade between 13 and 17 times. Australia’s top bank is valued at more than twice that of America’s biggest bank and nearly three times the global average.

There is little reason for this price. With earnings near A$5.58 per share, even a high multiple would not match today’s share price. Macquarie’s target is A$105, but the stock trades near A$180. The gap is not about business strength, but momentum. Index funds buy CBA for its size and rising price, not its business. As it grows, it draws in more money. The pattern is the same as in America, but focused on one Australian bank.

The rhyme

The links between America and Australia are clear. In both cases, a few big firms have soared, not because of profits but due to momentum and passive money. Morgan Stanley says Australian banks now trade at 19 times earnings, up from 13.5, which is above normal. The firm says prices now reflect all the good news, and the risk of a fall is greater than the risk of further gains. The recent rise in bank shares came from hopes of lower rates, which are now mostly priced in.

As in America, the focus on a few big firms has left other sectors behind. Morgan Stanley expects the materials and resources sector to lead the ASX in 2026, helped by a weaker dollar, strong metals demand, and better profits. This sector accounts for 19% of the market but generates 28% of earnings. Investors now pay less per dollar of profit here than they do for the index. Quality and value in Australia have not fallen. They have just been overlooked.

The real cost: when price stops allocating capital

The stock market decides which projects get money. High share prices let firms raise funds cheaply and invest. Low prices make money scarce, and good projects may not happen. Momentum has broken this link.

Momentum and passive money have made markets more concentrated. Seven firms now make up a third of the S&P 500. Index funds buy these firms for their size and rising prices, not for their quality. This cycle has split price from performance. The biggest firms get cheap money, even if they do not need it. Smaller, well-run firms pay more, not because they are risky, but because they are ignored.

Private markets show the same trend. In early 2026, deals over $100 million made up almost 90% of all venture capital. AI firms took most of this money, with three firms getting nearly half. This is not money going to the best ideas. It is money chasing momentum. Many good businesses miss out, not because they are weak, but because they do not fit the story.

It is hard to spot this misallocation in real time. The costs show up later. In Australia, passive money keeps pushing up one bank’s share price, while potentially better options are ignored. When markets reward size and rising prices over real returns, investment falls, and concentration grows.

What a mean reversion would look like — and why it’s not a prediction of doom

This is not a call for a crash. Momentum can last a long time. It is hard to fight and has paid off. Morgan Stanley still expects the ASX 200 to reach 9,250 in 2026, with returns of 10 to 12 percent. The U.S. market may also continue to rise. More importantly, the source of returns is likely to shift. When gains are focused in a few stocks, history shows that other firms catch up. Leaders do not always fall, but quality companies close the gap.

Long-term investors should focus here. A dramatic reversal is unnecessary; it suffices for the market to recognise the value inherent in robust finances. Right now, these traits are cheap. J.P. Morgan says quality was last this cheap after the dot-com crash and during COVID. Both were good times to buy top firms. The same forces that lifted momentum stocks have made quality firms cheap. Every dollar that goes to recent winners is a dollar not given to sound businesses. The key question is not which stocks are rising, but which firms are strong and undervalued. On this, America and Australia now offer rare choices.

This final assertion warrants immediate examination, as it is fundamentally incorrect. Its inversion is central to the economic viability of the entire proposition.

A data centre is a heater.

Begin with the first law of thermodynamics. A compute cluster converts essentially every watt of electricity it draws into heat. Not most of it, almost all of it. The theoretical minimum energy of computation (the Landauer limit, for the physicists in the audience) is about 15 orders of magnitude below what actual chips dissipate so that it can be ignored. A one-gigawatt AI training cluster is, thermodynamically speaking, a one-gigawatt bar heater that happens to produce model weights as a byproduct. That heat must leave the facility at exactly the rate at which it is generated, continuously, or the temperature rises without bound, and the silicon dies.

On Earth, the challenge of heat rejection is so trivial that it scarcely appears in a data centre’s capital allocation. The atmosphere and hydrological cycle provide essentially limitless, cost-free heat sinks. The evaporation of a single kilogram of water removes 2.26 megajoules of energy; a gigawatt-scale facility can dissipate its entire thermal output by evaporating several hundred kilograms of water per second, or by circulating ambient air across radiators with heat transfer coefficients in the hundreds to thousands of watts per square metre per degree. Cooling, in this context, is a minor engineering consideration and a routine expense, not a binding constraint.

In orbit, the absence of an atmosphere and water cycle removes conduction and convection as heat-transfer mechanisms, leaving radiation as the only option. Radiation can work without a medium, so the claim that “no particles mean heat cannot be removed, making space data centres impossible” is misleading. The real problem is more precise and more difficult: radiation is the only available heat-rejection channel, and it has limited capacity to dissipate heat at the gigawatt scale.

The fourth-power trap

The governing physics is one equation. The power radiated by a surface is P = εσAT⁴, where σ is the Stefan–Boltzmann constant (5.67 × 10⁻⁸ W/m²K⁴), ε is the surface’s emissivity, A is its area, and T is the radiator’s temperature, raised to the fourth power.

Read that carefully, because the crucial variable is not the temperature of space. The 2.7-kelvin cosmic background is the destination for the heat, but the rate at which heat leaves is set by how hot your radiator runs, not how cold the destination is. Vacuum is an insulator. Space is not a freezer that pulls heat out of your equipment; it is a thermos that traps whatever heat you generate until you can push it out photon by photon. This single misunderstanding that cold equals cooling drives the orbital data centre pitch more than any other.

At first glance, the fourth-power dependence appears advantageous: increasing the radiator temperature dramatically increases radiative flux. However, heat transfer is unidirectional; the radiator must remain cooler than the components it serves, and GPU silicon is limited to junction temperatures of approximately 85-95°C. Accounting for thermal gradients across the cold plate, coolant loop, and radiator fins, a realistic operating temperature for the radiating surface is in the range of 40 to 60°C. At 330 K with a high-emissivity coating, the theoretical maximum is approximately 600 watts per square metre per side, before environmental penalties. In Earth orbit, radiators also absorb significant energy: up to 1,361 W/m² from direct sunlight, approximately 240 W/m² from terrestrial infrared emission, and additional reflected albedo. Attempts to use heat pumps to raise radiator temperature are self-defeating, as the pump’s work is converted into additional heat that must also be rejected. The second law of thermodynamics imposes an unavoidable penalty.

The best benchmark we have is the International Space Station, whose actively pumped ammonia radiator system is the most sophisticated space thermal control ever flown. It rejects about 70 kilowatts across hundreds of square metres of deployed panels, a net performance around 100–150 W/m². Industry rules of thumb for low-Earth-orbit radiators sit at perhaps 150–350 W/m² net.

Scaling this requirement to an optimistic 300 W/m², a single gigawatt of compute requires approximately 3.3 square kilometres of deployed radiator area. For an idea of scale, this is equivalent to the area of Central Park, or twice the size of Melbourne’s Hoddle Grid. All of this requires precision-engineered, fluid-filled panels susceptible to micrometeoroid damage. At performance levels demonstrated by the ISS, the requirement approaches seven square kilometres. Notably, current industry plans do not contemplate one-gigawatt facilities, but rather campuses of five to ten gigawatts.

The mass requirement is equally prohibitive. ISS-class radiators operate at tens of kilograms per kilowatt of heat rejection. Even the most ambitious next-generation designs aim for 5-10 kg/kW. These are values that remain unproven in practice. At 5 kg/kW, a single gigawatt of heat rejection requires 5,000 tonnes of radiator, excluding any GPUs, solar panels, or structural elements. Furthermore, the approximately three square kilometres of solar arrays required to power such a facility generate additional waste heat, which must also be dissipated.

On Earth, heat rejection is a negligible consideration. In orbit, however, the Stefan–Boltzmann law dictates the mass, cost, and reliability of the entire system. This is the primary constraint. The critical question is whether the second variable—the cost of delivering mass to orbit—can offset it.

Sixty years of launch costs, and the plateau nobody mentions

The launch cost story is usually told as a triumphant decline. The honest version is stranger. Using the standard CSIS Aerospace Security dataset in inflation-adjusted dollars per kilogram to low Earth orbit: the Vanguard programme of the late 1950s cost on the order of $900,000/kg. By the mid-1960s, Titan-class vehicles had brought that cost down to roughly $25,000–30,000/kg, and the Saturn V, with its sheer scale, achieved about $5,400/kg in 1967.

For the subsequent forty-five years, there was no further progress. The Space Shuttle, marketed as a reusable breakthrough, delivered payloads at approximately $54,000 to $65,000 per kilogram—a tenfold regression from Saturn V. Titan IV and Delta II in the early 1990s achieved $30,000 to $39,000 per kilogram. The Atlas V and Delta IV vehicles of the 2000s reached $8,000 to $13,000 per kilogram. Between 1965 and 2010, the inflation-adjusted cost of reaching orbit remained within a narrow band, exhibiting no downward trend. Any projection of a declining cost curve across this period must first account for this prolonged stagnation.

Reusability finally broke the plateau. Falcon 9 debuted in 2010 around $2,700/kg; Falcon Heavy reached roughly $1,500/kg by 2018. This is a reduction of more than 90 per cent from the Shuttle era in under a decade, and the foundation of the entire modern commercial space economy.

This leads to a significant complication in the 2026 data: the price per kilogram has ceased declining and is now increasing. In February, SpaceX raised the dedicated Falcon 9 launch price to US$74 million, or $3,245 per kilogram at full payload, and simultaneously increased rideshare rates to $7,000 per kilogram. Concurrently, analyst estimates place SpaceX’s internal marginal cost for a reused Falcon 9 launch at $15 to $28 million, or approximately $630 to $1,000 per kilogram. The divergence between cost and price is now pronounced. While the underlying cost of reaching orbit continues to decrease, the market price has decoupled and is rising. The cost curve and the price curve are now distinct, and any business case predicated on ‘collapsing launch costs’ must specify which metric it refers to.

Is $100 per kilogram real?

The number that makes orbital data centres pencil is the Starship target: $100–200/kg. The honest status of that number is a target with improving supporting evidence and zero demonstrated pricing. Starship has flown thirteen times with a genuinely mixed record. Through to late 2025, eleven launches resulted in six successes and five failures, and the first flight of the upgraded V3 vehicle in June this year was a partial success on a suborbital trajectory. The V3 specification matters: roughly 200 tonnes to orbit, fully reusable, meaning a reused V3 would match what the original Starship could only do expendably. Third-party analysts at Payload Research have estimated an internal cost of around $500/kg for the expendable first-generation configuration, and report that SpaceX’s internal roadmap projects a path toward $100/kg as the fleet scales past 70 flights. None of this is a rate card. No customer has bought a kilogram at these prices.

Is there a floor? Two; one physical, one economic. The physical floor comes from the rocket equation and propellant chemistry: reaching orbit requires about 9.4 km/s of velocity change, which, at methane-oxygen efficiency, means burning 20–45 kilograms of propellant per kilogram delivered. Starship’s propellant load costs on the order of a million dollars per flight, implying $5–10/kg in fuel alone at full payload. Mature transport industries, with airlines as the canonical case, provide an asymptote to total operating costs of two to four times fuel once vehicles fly thousands of cycles with minimal refurbishment. That implies a practical floor around $20–50/kg, and only in a situation where Starships turn around like A320S. So no, $100/kg does not violate physics. It requires airline operations on a vehicle that has not yet flown the same airframe twice.

The economic floor may prove more restrictive, as evidenced by Falcon 9: internal costs are approximately $630 per kilogram, while customer prices have risen to $3,245 per kilogram. In the absence of effective competition, any cost reductions achieved by Starship will primarily benefit SpaceX’s margins and its vertically integrated initiatives, such as Starlink and the very orbital compute projects under discussion. The broader market is unlikely to access $100 per kilogram pricing; only SpaceX may do so internally.

Multiply the constraints

The relevant calculation is straightforward. Assuming five thousand tonnes of radiators per gigawatt, which is a highly optimistic unproven figure and the current market price of $3,000 per kilogram, launching the cooling system alone would cost $15 billion per gigawatt. This renders the proposition unviable relative to terrestrial data centres, where cooling costs are negligible. At an internal Starship cost of $100 to $150 per kilogram, the same radiators would cost $500 to $750 million, which is still substantial, but not implausible when compared to multi-billion-dollar terrestrial projects, especially if the orbital solar advantage (five to eight times the annual energy yield per panel, with no land, water, or grid constraints) is fully accounted for.

The case for orbital data centres is thus neither a wager on physics nor on general launch costs. It is a compound bet: first, that radiator specific mass improves by several multiples beyond any system yet flown; second, that Starship achieves true airline-like operational cadence; and third, that SpaceX utilises this capability internally rather than offering it at market rates, all this before terrestrial constraints on power and cooling are alleviated through incremental improvements in generation, chip efficiency, and grid expansion. Each assumption is individually plausible. Their combination, however, is precisely the mechanism by which venture returns are occasionally realised and infrastructure projections are routinely invalidated.

For allocators, the appropriate response is rigorous due diligence, not premature endorsement or dismissal. When orbital compute is presented in a TAM slide or offer document, as it inevitably will be, the relevant questions are precise: what net radiative flux is assumed, in watts per square metre, and how does this compare to the 100 to 150 W/m² demonstrated by the ISS? What radiator mass per kilowatt is projected, relative to the tens of kilograms achieved to date? Is the launch cost assumption based on market price, internal cost, or an aspirational target? As of this year, these are three distinct figures, each trending differently. The Stefan–Boltzmann law has been established physics since 1884. It is indifferent to narrative, and its compounding effect on leverage is as inexorable as interest’s.

The recurrence of this conversation reflects the persistence of the underlying phenomenon. Conventional commentary, however, tends to treat LIC discounts as a uniform issue across the sector, with solutions typically framed as improved board governance, structural reform, manager buybacks, or simply patience. This analysis contends that the discount issue is not uniform. Rather, it consists of two related but distinct problems, each associated with specific structural features of the listed vehicle. The resulting policy implications for investors differ materially from those suggested by standard commentary.

The two features are size and underlying-asset liquidity. The very large, very liquid, very well-known established Australian LICs such as the Australian Foundation Investment Company at approximately ten billion dollars, Argo at approximately seven, BKI and AUI and Whitefield each in the one-to-two-billion range, all trade, in the great majority of market environments, at or close to net tangible assets. Premiums are common during equity market weakness as investors rotate to perceived safety; modest discounts emerge during strength as the same investors rotate out. The cycle is bounded. AFIC trading at a ten per cent discount, as it briefly did during 2024, was described in industry commentary as “almost unheard of” — and it is.

The smaller LICs, broadly those with market capitalisation under three hundred million dollars, exhibit a different pattern. They trade at structural discounts that are wider, more persistent, and less responsive to market sentiment than the large-cap LICs. The discounts do not close on their own when markets recover. They are not principally a function of investor confidence in the manager. They are a function of the structural position the LIC occupies in the secondary market. The LICs holding illiquid underlying investments—particularly private equity, but also unlisted credit, infrastructure, and concentrated unlisted equity positions—exhibit a more extreme version of the same pattern. The Pengana Private Equity Trust, the Bailador Technology Investments LIC, and the Cordish Dixon Private Equity Fund series have, at various points over the past five years, traded at discounts of 15, 20, and 30 per cent. None of these discounts is principally about the underlying portfolio quality or the manager’s competence. All of them are principally about the secondary-market structure of the listed vehicle relative to the marginal buyer of its shares.

The following analysis examines each of these claims in sequence.

The size axis

The Bell Potter framework for LIC analysis, which is the closest thing the Australian market has to a generally accepted analytical structure for the sector, uses a market capitalisation threshold of $500 million to distinguish large-cap from small-cap LICs. Above the threshold, the LIC trades within a narrower band around NTA; below it, the discount widens and becomes more cyclical. The threshold itself is somewhat arbitrary with different market segments, and commentators use different cutoffs, ranging from 300 million to 1 billion — but the underlying empirical observation is robust. Smaller LICs trade at wider, more persistent discounts. The relationship is monotonic across the size distribution.

Three structural factors account for this pattern, and their effects are cumulative.

The first is secondary market liquidity. An institutional investor such as a superannuation fund, family office allocation, or dedicated LIC fund needs to deploy a meaningful position in an LIC and needs the secondary market to absorb the trade without significant price impact. For a one-billion-dollar LIC, a five-million-dollar position is half a per cent of the vehicle’s market capitalisation and can be accumulated over a small number of days without materially moving the price. For a one-hundred-million-dollar LIC, the same five-million position is five per cent of market cap, and either takes weeks to accumulate or moves the price substantially. The institutional buyer is, in practice, capped out of the smaller LIC by the secondary market microstructure. The marginal buyer pool shrinks, and the buyer who remains, usually a retail investor or a small-scale wholesale investor, is less price-insensitive and less able to arbitrage the discount.

The second is analyst coverage and broker market-making. The economics of providing equity research coverage on a sub-three-hundred-million-dollar LIC do not work for most broker firms. The fund management fees generated by the LIC are small relative to the cost of producing meaningful research; the trading volumes generated by the LIC’s shareholder base are small relative to the cost of providing market-making support. The major Australian broker firms, such as Bell Potter, Wilsons, Morgans, and MST Marquee, produce regular research on the larger LICs and a select subset of smaller ones, but the long tail of small LICs has minimal or no broker coverage. Without coverage, the LIC is not visible to the wholesale and institutional channels that would otherwise compete to narrow the discount.

The third is the cost-of-capital arithmetic for the managers themselves. A LIC trading at a sustained discount to NTA cannot raise additional capital at NTA without diluting existing shareholders. This means the management company cannot grow the LIC, so management fees cannot grow, diminishing the manager’s incentive to invest in research coverage, marketing, or board engagement. The LIC becomes, in financial-planning terms, a sunset asset for the manager; the manager’s commercial energy shifts to other products. The discount persists because no party has the economic incentive to close it.

These three factors reinforce one another. The institutional buyer is excluded, the broker coverage that might attract such buyers is lacking, and the manager has no economic incentive to address either. As LIC size decreases, the influence of each factor intensifies, resulting in more persistent discounts.

The illiquidity axis

The second structural feature that predicts LIC discounts is the liquidity of the underlying holdings. This is the dimension along which LICs investing in private equity, unlisted credit, infrastructure, and concentrated unlisted positions diverge most sharply from LICs holding broad listed equities.

The mechanism underlying this effect is analytically more nuanced than the size effect, yet the empirical pattern it produces is more pronounced.

For an LIC holding only listed Australian equities, the manager’s published monthly net tangible assets figure is a verifiable number. Every position in the portfolio has an observable market price at the relevant date. The NTA is the sum of those prices, less liabilities, divided by shares on issue. An investor who is sceptical of the manager’s NTA calculation can reconstruct it themselves from public information. The discount, if any, is a function of the LIC’s secondary market price relative to an NTA that the market accepts as accurate.

For an LIC holding private equity positions, the NTA figure is something different. The underlying holdings are unlisted. They are valued at the manager’s estimate of fair value, typically supported by a combination of recent transactions in the underlying companies, comparable public-market multiples, discounted cash flow analyses, and external valuation reviews. The manager’s valuation methodology may be entirely defensible, conducted by appropriately qualified parties, audited by reputable firms, and consistent with industry practice. It is also, structurally, the manager’s estimate. The market is being asked to take the NTA on the manager’s authority rather than to verify it independently.

The market’s response is to discount the NTA by some amount that reflects, in aggregate, three distinct concerns. The first is uncertainty about the valuation itself; this is the discount, the market’s haircut against the manager’s estimate. The second is illiquidity of exit because the LIC’s holdings cannot be sold quickly if the LIC needs to realise them, so the realised value at exit may be materially below the carrying value. The third is the asymmetry in disclosure. The market believes that the manager has more information about the holdings than it does, and discounts for this information asymmetry rather than relying on the disclosed NTA.

Each of these concerns is rational, and each operates independently. Combined, they produce structural discounts on PE-holding LICs that are markedly wider than the discounts on listed-equity-holding LICs of similar size. The Pengana Private Equity Trust, Bailador Technology Investments, and the Cordish Dixon Private Equity Fund series have all traded at discounts to NTA in the fifteen-to-thirty per cent range across various points in their listed lives. These discounts are not principally about the underlying portfolio quality or the manager’s track record. They are about the market’s pricing of the disclosure structure.

A structural arbitrage argument arises in this context. When a PE-LIT trades at a twenty-five per cent discount to NTA and the NTA is accurate, a long-term investor who holds until wind-up realises the discount as an enhancement to returns. If, instead, the NTA is overstated by twenty-five per cent, the discount reflects the market’s correct adjustment, and the investor receives only the underlying return. The expected return is thus the portfolio return plus an uncertain portion of the discount, weighted by the likelihood that the NTA is accurate. The investment case for a discounted PE-LIT is, in effect, a judgement on whether the manager’s NTA estimate is more accurate than the market assumes.

This is a different analytical question from the equivalent question for a listed-equity LIC, where the NTA is verifiable and the discount is primarily driven by secondary-market microstructure rather than valuation uncertainty.

When the two axes compound

The structural features compound when they coincide. A large LIC holding broad listed equities, AFIC, Argo, which sits at the favourable corner of the matrix and trades close to NTA in normal market conditions. A small LIC holding illiquid investments sits at the unfavourable corner and trades at the deepest, most persistent discounts in the sector. The Cordish Dixon Funds, with a sub-one-hundred-million-dollar market capitalisation and holding US private equity fund-of-fund positions, have at various points been the cleanest empirical illustration of the compound effect. So have the smaller hedge-fund and concentrated-equity LICs like the VGI Partners vehicles before their restructure, the Salter Brothers Emerging Companies fund, and the various small-cap-focused boutique LICs across the sector.

The mid-cases are mid-cases. WAM Capital and the Wilson Asset Management family, the Whitefield Industrials LIC, and the smaller LICs holding broad listed equities typically trade at discounts that are real but bounded in the five-to-fifteen per cent range, with cyclical variation around that base.

The small-company-focused LICs — those investing in Australian small-cap listed equities all sit interestingly within the matrix. They are typically sub-three-hundred-million-dollar vehicles by market cap (the size axis), but they hold listed positions (the liquidity axis is favourable). The Acorn Capital Investment Fund (ACQ), the Naos Emerging Opportunities LIC, the Glennon Small Companies LIC, and the Concentrated Leaders Fund all sit in the mid-range of the discount distribution. Wider than the very large LICs, narrower than the PE-LITs. The fact that their holdings are listed and verifiable bounds the discount; the fact that they are small bounds the marginal buyer pool.

Within this framework, the cross-sectional empirical pattern is predictable to a first approximation. The position of a given LIC on the size and liquidity axes largely determines the level at which its discount is likely to stabilise.

What does this imply for investors

Three implications follow from the two-factor framework, and they differ from the implications that the standard commentary tends to draw.

The first implication is that purchasing a small or illiquid LIC at a discount, with the expectation that the discount will close over time or as sentiment improves, rests on a weak structural premise. The discount is not primarily a sentiment-driven effect that will mean-revert. Rather, it reflects the structural pricing of the marginal buyer’s position in the secondary market, and it endures as long as the underlying structural conditions persist. While the discount may narrow during periods of strong sector inflows or in response to corporate actions such as share buybacks, wind-up announcements, or takeover bids, it does not narrow in the absence of such catalysts.

The second implication is that the most effective way to realise the discount on a small or illiquid LIC is to hold the position when a catalyst occurs. Activist investors targeting persistently discounted LICs, such as Sandon Capital and Affluence Funds in the Australian market, have at times generated significant returns by compelling boards to undertake corporate action. Vehicles such as Hearts and Minds, VGI Partners Global Investments, Pengana International Equities, and Salter Brothers Emerging Companies have all attracted activist attention with the explicit objective of forcing actions to close the discount. The activist thesis is not that the discount will close due to undervaluation, but that it will close as a result of direct intervention. These are distinct theses, each with different probabilities and capital requirements.

The third implication is that the structural disadvantages inherent in the small and illiquid LIC segment account for the sector’s long-term contraction. For much of the twentieth century, listed investment companies were the standard vehicle for Australian retail investors seeking diversified equity exposure with professional management. The subsequent growth of unlisted unit trusts, the introduction of ETFs in the late 1990s, and the proliferation of exchange-traded managed funds since 2015 have progressively eroded the structural rationale for the closed-end listed vehicle. Structures that converge to NTA, such as ETFs through arbitrage and unit trusts through daily creation and redemption, do not experience the discount issue because their disclosure architecture compels the market price to track the underlying portfolio. The LIC structure remains relevant only where its closed-end nature provides a genuine advantage, such as for managers running concentrated or unlisted positions who require a stable capital base. Where no such advantage exists, as with managers running broadly listed strategies, the closed-end structure is structurally disadvantaged relative to open-ended alternatives.

The structural sorting that has been occurring over the past decade, with large established LICs maintaining their position, smaller LICs gradually being wound up or converted to open-ended structures, and the PE-LIT segment producing few new listings, is consistent with the framework above. The sorting is not finished. The market microstructure is doing what it is expected to do.

What this implies for managers and boards

For managers and boards of small or illiquid LICs, the policy implications are correspondingly specific. The structural discount is not primarily a function of investor sentiment that can be addressed through communication, manager engagement, or shareholder events. Rather, it is determined by the relationship between the listed structure and the marginal buyer. The policy levers that materially influence the discount, in approximate order of structural impact, are as follows.

The first is wind-up or conversion. The most direct way to close a sustained structural discount is to remove the listed vehicle structure that produces it by winding up the LIC and returning capital, converting to an open-ended unit trust, or being acquired by a competitor. Many of the small Australian LICs that have closed over the past decade have done so via one of these mechanisms. The discount closure on the path to wind-up is typically substantial.

The second is share buybacks executed at NTA. Buybacks that accrete NTA per share are the policy lever most directly available to the board and most consistently effective. They require the LIC to use available cash to repurchase shares at the market discount, which is a counterintuitive use of capital for boards focused on growth, but it is the highest-return capital deployment available to the LIC as long as the discount persists.

The third is a dual-class or open / closed structure. The Magellan Global Fund’s restructuring into a dual-class structure with both open and closed share classes — preserving the closed-end capital structure for the manager while offering investors a redemption option through the open class- is one example of a structural compromise that addressed the persistent discount on the closed-end class. The structure is not appropriate for all LIC strategies, but for managers whose investment approach benefits from a stable capital base, the dual-class approach is one of the few structural options that materially affects the discount.

The fourth lever is communication and disclosure. Standard commentary on LIC discounts often emphasises this approach, advocating for improved board engagement, greater transparency, and clearer articulation of manager strategy. However, the framework outlined above indicates that this lever has the least impact, as the discount is not fundamentally a communication issue. Enhanced disclosure may marginally narrow the discount in favourable market conditions, but it does not address the underlying structural microstructure.

The deeper point

Conventional commentary on LIC discounts often frames the phenomenon as a market puzzle, implying either a general dysfunction in the LIC sector or opportunities for patient investors. The two-factor framework presented here suggests that neither interpretation is accurate. The discounts are not a puzzle but rather a rational pricing mechanism in the secondary market microstructure for small and illiquid listed vehicles. Nor are they, in general, opportunities for patient investors, as the structural factors underlying the discounts do not mean-revert.

For the Australian adviser, family office, or wholesale investor allocating to the LIC sector, the practical implication is straightforward. The large, established listed-equity LICs are reasonable holdings, with discount-to-NTA dynamics bounded by the structure’s size and liquidity. The small and illiquid-holding LICs are either positions held with the expectation of corporate action catalysts or positions whose discounts should be understood as a structural feature of market microstructure rather than a sentiment-driven mispricing.

For managers and boards of LICs, the implication is that sentiment-management interventions such as annual general meeting roadshows, manager communications, and board engagement are largely unrelated to the structural drivers of the discount. Effective interventions are those that address the vehicle's structure rather than investor sentiment.

Whether the Australian LIC sector continues to contract, stabilises at a smaller equilibrium, or assumes a new structural role within broader portfolio architecture remains uncertain. What is clear from the available evidence is that the prevailing conversation about LIC discounts has lacked analytical precision. The discount is not a single phenomenon but rather the result of two related yet distinct structural features of a particular type of listed vehicle. The sector will continue to evolve in response to these underlying dynamics.

If you enjoy Daniel Liptak, it would mean the world to me if you invited friends to subscribe and read with us. If you refer friends, you will receive benefits that give you special access to Daniel Liptak.

How to participate

1. Share Daniel Liptak. When you use the referral link below, or the “Share” button on any post, you'll get credit for any new subscribers. Simply send the link in a text, email, or share it on social media with friends.

Refer a friend

2. Earn benefits. When more friends use your referral link to subscribe (free or paid), you’ll receive special benefits.

• Get a 1 month comp for 3 referrals

• Get a 3 month comp for 5 referrals

• Get a 6 month comp for 25 referrals

Visit the leaderboard

To learn more, check out Substack’s FAQ.

Thank you for helping get the word out about Daniel Liptak!

Before proceeding, it is important to be clear: ‘gaming the rules’ in this context does not imply illegality. Nearly everything described here is lawful, disclosed, and often anticipated by the very regulators who drafted the rules. This is the core of the problem. The central claim is that the asymmetry is not a loophole; it is a feature. Owners secure the upside and entrench their control, while members are left to absorb volatility and valuation risk. The regulatory framework treats this as something to be disclosed, not something to be prevented. That is the central issue under examination.

The mega-float as a one-way option for the founder

Start with the structure of the SpaceX listing itself, because the numbers make the asymmetry impossible to miss. The combined SpaceX-xAI-X entity floated at a valuation of roughly US$1.75–1.8 trillion, raising about US$75bn, making it the largest IPO in history. But the float represented only around 4.3 per cent of the company’s equity. Roughly 95 per cent stayed in private hands. And the shares sold to the public were mostly newly issued rather than insider stock cashed out, meaning Musk’s economic stake of roughly 42 per cent of the consolidated entity was barely touched by the listing. This is the first concrete example of the broader pattern.

Then comes the control mechanism. Through a dual-class structure, Class B super-voting shares carry 10 votes each against 1 vote for the Class A stock sold to the public, leaving Musk with somewhere in the region of 79–85 per cent of the voting power while owning a minority of the economic interests. Public shareholders, including every super fund and pension fund that bought in, have supplied a large share of the capital and, in governance terms, received almost nothing. Because mergers, acquisitions, capital raises, executive pay, and strategic pivots all rest with the insider, he can approve them unilaterally and without public or investor consent. This was disclosed prominently in the S-1 as a risk factor. Disclosure was the entire defence: tell them it’s asymmetric, and the asymmetry becomes permissible.

This structure produces an extraordinary result for the founder. He secures tens of billions in permanent capital at a record valuation, marginally dilutes his own economic interest, and preserves absolute control. The consolidation of xAI and X into the listed entity in February 2026 shifts an annual loss of approximately US$5 billion, along with a monthly cash burn of US$1 billion, onto the public shareholders’ balance sheet. Public investors may have believed they were acquiring a space company, but in reality, they purchased a sprawling conglomerate encompassing rockets, satellite internet, a loss-making AI compute business, and a social media platform, without any influence over whether these businesses should be combined. For the owner, this arrangement functions as a one-way option: he retains both the upside and control, while the public is left to shoulder the valuation risk as the company trades at nearly ninety times revenue and continues to report GAAP losses.

The next phase is already embedded in the structure. The lockup period for insiders and early investors lasts the standard 90 to 180 days, opening the exit window between September and December 2026. At that point, early employees, venture investors, and the underwriting syndicate can all sell into a public float that remains thin precisely because so little equity was released at IPO. This scarcity inflates prices on the way up as demand chases limited supply, but provides little support when the lockup ends, and a flood of stock hits the market. This sequence is not accidental; it is engineered. The recent 25 per cent decline in SpaceX’s share price within weeks of its listing suggests the market is beginning to anticipate the consequences of this design.

This is not the invention of a single founder. Meta, Alphabet, and Salesforce established the dual-class model; the practice of floating a minimal public tranche at a maximal valuation has become standard for the most coveted listings. The issue is structural: regulatory permission allows founders to monetise their holdings while retaining control, and to float only a small fraction at inflated prices. The central claim is that this creates a repeatable mechanism for converting public capital into private control, with the risks distributed across the buyers. Increasingly, those buyers are pension and superannuation funds, which brings us to the other side of the equation.

The other side of the trade: how the super system was built to be the buyer

For a founder to extract value through these mechanisms, a counterparty is required that is large, captive, and structurally compelled to continue purchasing. Australia’s A$4.3 trillion superannuation system fits this description with remarkable precision. Its regulatory architecture, not through malice but through the aggregation of individually rational rules, has produced a system that is collectively susceptible to exploitation. This is the necessary counterpart to the founder’s extraction apparatus.

The first defining feature is compulsion and scale. Mandatory contributions guarantee a continuous flow of capital that must be invested with each payday, irrespective of prevailing market conditions. APRA has accelerated consolidation by imposing an A$30 billion minimum fund size, concentrating capital in a diminishing number of very large funds. By 2024, 933 APRA-regulated funds managed more than twice as many assets as a decade earlier, even as the number of funds declined by a third. Large funds with global equity mandates are, by design, compelled to purchase whatever dominates the benchmark, which is now the AI sector. The system’s defining characteristic the compulsory inflows thus becomes its principal vulnerability: the capital must be deployed, even when prudence would counsel restraint.

The second feature is the use of the benchmark as a regulatory tool. The Your Future Your Super performance test measures funds against benchmark indices, publicly shaming and potentially closing those that underperform. The intention was to protect members from persistent underperformance and excessive fees. In practice, it punishes any deviation from the index. A fund that prudently underweights an overvalued, AI-heavy benchmark and is wrong for even a year risks failing the performance test, attracting regulatory scrutiny, and losing members. A fund that simply tracks the benchmark and rides the bubble down is, by definition, only following the index it is measured against. Regulation intended to protect members by enforcing a benchmark instead forces funds into the crowded trade and penalises the caution that would actually safeguard retirement balances in a downturn. AustralianSuper has acknowledged this tension, stating it will only reduce its international equities overweight if it determines there is a bubble, clearly an admission that the default is to remain exposed until the damage is done.

The third feature concerns the valuation and liquidity treatment of unlisted assets, where the harm to members becomes most tangible. Approximately 16 per cent of APRA-regulated super investments are in private-market assets, about half of which are offshore, or roughly A$500 billion. These assets are valued infrequently and with considerable discretion. APRA’s December 2024 thematic review identified widespread shortcomings in valuation governance, including conflicts of interest, insufficient oversight, infrequent revaluations, reliance on external managers, and inconsistent fair-value reporting. APRA found these issues serious enough to require 12 of the 23 reviewed funds to improve their valuation or liquidity frameworks, as weak governance can delay the recognition of losses and distort outcomes for members.

This is significant for the asymmetry because infrequent and discretionary valuation of illiquid assets introduces a timing inequity among members. As Morningstar and others have observed, the regulator’s primary concern is the equitable treatment of beneficiaries entering or exiting the fund. Since most super funds are defined-contribution, members may transfer between funds at any time, necessitating valuations to ensure fairness. If a private asset is maintained at an outdated, inflated value during a downturn, the member who exits at that price is overcompensated at the expense of those who remain. Those who stay absorb losses that were already real but not yet recognised. This is not a hypothetical risk: HESTA was required to compensate two groups of members after APRA determined that its valuation decisions at the onset of the Covid-19 pandemic were inadequate and inequitable. Ultimately, discretion over when to recognise a loss is discretion over which members are required to bear it.

Where the two structures meet

The argument becomes more acute when the founder’s extraction apparatus and the super system’s structural compulsion are considered together, as they are mutually reinforcing. The first creates the opportunity to externalise risk; the second guarantees the presence of a buyer able to absorb it. In combination, they transform disclosure into permission and compulsion into vulnerability. The central contention is that one system is constructed to extract risk, while the other is constructed to absorb it.

The mega-float presents public investors with a governance-light, loss-making, and highly valued asset, complete with a preordained post-lockup selling event. The superannuation system supplies a compelled, benchmark-driven, and scale-constrained buyer that must deploy capital and is penalised for deviating from the index now dominated by that asset. Simultaneously, the expanding private-market segment of the super system, including offshore private credit invested in AI data centres, as APRA has observed, is valued so infrequently that losses from any AI repricing can be deferred, smoothed, and redistributed among member cohorts depending on the timing of revaluation. The owner secures permanent capital at the peak of the cycle. The member provides it by compulsion, holds it under benchmark pressure, and, if conditions deteriorate, may absorb the loss on a delayed and uneven schedule that favours those who managed to exit at the outdated price.

APRA is aware of these dangers. Its 2026 system-risk review identifies the relevant vulnerabilities: it continues to monitor entities exposed to private markets, noting that while domestic private credit risks remain contained, regulated entities are globally interconnected and exposed to offshore developments. The review notes that some US private-credit providers have large, concentrated exposures to software companies, that concerns about AI’s impact on earnings have led to increased redemption requests, and that about half of super funds’ private-market exposures are international. The regulator understands the channels of contagion. But recognising these risks is not the same as closing them. The timing problem is structural: APRA’s response is to intensify supervision, raising expectations on valuation governance and liquidity risk management, rather than prohibiting the practices that create the asymmetry. Supervision may gradually lift standards, but it does not prevent a determined founder from floating a minimal, governance-light tranche at a record valuation, nor does it force a fund to revalue a deteriorating private holding before its scheduled cycle if it prefers not to.

This is the persistent pattern. Regulation is reactive, disclosure-based, and tethered to benchmarks. Extraction is proactive, structural, and acutely attuned to those benchmarks. The gap between a rule crafted to address the previous crisis and a structure designed to exploit current permissions is where value is transferred from members to owners. The dual-class share rule was initially a reasonable concession to founder-led innovation; it has evolved into a mechanism for separating control from accountability. The performance test was intended to protect members; it now enforces participation in crowded trades. Discretion in valuing unlisted assets was meant to accommodate genuinely hard-to-price holdings; it now operates as a lever to shift losses among member cohorts. In each instance, a well-intentioned rule, when confronted by a sophisticated counterparty, has produced an asymmetry the rule never anticipated.

What the bust does to the asymmetry

If the AI thesis collapses on the timeline anticipated by the BIS, the asymmetry will shift from theoretical to actual, resulting in tangible losses. The order in which those losses are distributed is the central issue.

The insider has already secured permanent capital at the peak and retains control irrespective of the share price. A 50 or 70 per cent decline affects only notional wealth, as the cash is secured and the votes are entrenched. Early investors and the syndicate, if they have timed the lockup appropriately, will have commenced selling into the public float before the worst materialises. Public shareholders, including super funds, pension funds, and individual members, hold Class A shares with no control, no preferential liquidity, and only residual economic exposure to a conglomerate now valued at a fraction of its purchase price. On the listed side, the loss is at least immediate and transparent. On the unlisted side, the situation is more difficult: private-credit and private-equity exposures to the same AI sector will be written down gradually and unevenly, and because the timing is discretionary, members who exit during the lag are subsidised by those who remain, as demonstrated by the HESTA precedent, now at system scale, in a downturn far larger than March 2020. APRA has already identified stress in private credit, with funds experiencing redemption requests and, in some cases, restricting withdrawals. Restricting withdrawals is the mechanism by which illiquidity becomes the member’s problem precisely when liquidity is most needed.

The most fundamental question is who selected the risk and who is left to bear it. The founder determined the structure and captured its rewards. The fund adopted a benchmark-tracking allocation under regulatory compulsion. The member whose compulsory contributions were invested in a benchmark dominated by governance-light mega-caps, and then partially redirected into infrequently valued private credit lent to AI data centres, chose none of it and is left last in line when the structure fails. The retirement saver is the residual claimant in a sequence of decisions made by others, each insulated from the full extent of the downside. This is the meaning of the assertion that the rules were constructed so insiders prevail, and members pay: not that anyone violated the law, but that the legal architecture channels the upside to those who designed it and the downside to those who were merely required to participate.

What follows

The practical conclusions are necessarily narrower than the critique, as the critique is structural and no individual can reconfigure the superannuation system. Nonetheless, three clear implications emerge.

For members, the lesson is that ‘diversified default option’ and ‘prudently managed’ are not synonymous. The regulatory protections surrounding superannuation, particularly the performance test, guard against fee-driven underperformance, but not against correlated bubble risk imposed by the benchmark. A member who recognises that their default option is compelled by the benchmark into the AI sector can at least make an informed decision about whether this aligns with their own risk tolerance and investment horizon, rather than presuming the regulatory framework has addressed it.

For funds and their advisers, the lesson is that the timing of valuation has become a primary governance concern. The discretion APRA identified is as much a liability as a convenience: a fund that maintains outdated private-market valuations during an AI repricing is, whether intentionally or not, determining which members it benefits. Proactively addressing revaluation through quarterly independent assessments, prescriptive and asset-specific triggers, and genuine separation between investment and valuation functions is the distinction between equitable treatment of members and a future compensation liability on the scale of HESTA, or greater.

For those assessing the broader system, the lesson is that asymmetry will not be remedied by disclosure, because disclosure is the very mechanism that enables it. A risk factor in an S-1 does not protect the buyer; it protects the seller. The structures that transfer value from members to owners are lawful, disclosed, and proliferating, while the regulatory response, such as supervision, expectations, and thematic reviews, will always trail a generation behind the practices it seeks to oversee. The honest conclusion is not that regulators are negligent; APRA’s analysis of contagion channels is genuinely perceptive. The difficulty is that a reactive, disclosure-based, benchmark-anchored rulebook is structurally outpaced by sophisticated actors who interpret it as a set of permissions. Until the rules treat asymmetry as something to be prevented, rather than merely disclosed, compulsory contributions will continue to finance founders’ permanent capital, and members will remain last in line when the reckoning arrives.

The significance of these warnings lies not in their novelty. Bubble alarms are both common and inexpensive. What matters is what they reveal about market positioning. The issue is no longer confined to the high valuation of AI stocks. Institutions and retail investors are now structurally overweight a single, correlated, and leveraged theme. This crowding has diverted capital from other sectors of the economy, and the mechanics of any subsequent unwind, should returns disappoint, are likely to be far more damaging than a routine correction. This analysis addresses three problems in sequence: the concentration itself, the leverage embedded within it, the capital diverted from other areas, and finally, the likely contours of the reversal.

How concentrated the bet has become

Start with the raw arithmetic of exposure, because it is the part most investors underestimate about their own books. The five largest hyperscalers are expected to deploy more than US$1 trillion in capex between 2025 and the end of 2026. That spending is concentrated among a handful of names that also dominate the major equity indices. A passive investor who believes they are diversified by holding a broad US index fund is, in practice, holding a portfolio whose returns are increasingly determined by the AI capex cycle. The S&P 500 has become, to a meaningful degree, a leveraged bet on a single thesis wearing the costume of a diversified benchmark.

This is the first trap, and it is insidious precisely because it is not immediately visible. The retail investor allocating to an index fund, the super fund member defaulted into global equities, and the family office pursuing a so-called balanced 60/40 portfolio all carry significantly more idiosyncratic AI risk than their asset-allocation statements would indicate. Concentration at the index level has quietly undermined the diversification investors believe they possess. When a handful of stocks account for the majority of index returns, the index ceases to function as a diversified instrument and instead becomes a thematic vehicle. One can hold 500 names and still be making a single bet.

For active institutions, the dynamics differ but are no less hazardous. Career risk and benchmark risk drive managers toward the crowded trade rather than away from it. A manager who underweights AI and is wrong, even briefly, faces redemptions and underperformance relative to peers. Conversely, a manager who overweights and is wrong finds safety in numbers and a convenient rationale. The incentive structure thus rewards crowding. This is how the so-called smart money and the so-called dumb money end up in identical positions. The convergence is not a matter of collective folly, but rather the result of institutional incentives to track the benchmark and retail investors' tendency to focus on recent performance. When all participants are positioned on the same side, there are no buyers left when the market turns.

The market behaviour of mid-2026 provides clear evidence of crowding. SpaceX’s approximately US$86 billion IPO attracted intense demand, only for the shares to decline by about 25 percent from their peak within weeks. The company then issued a US$25 billion bond, prompting the ‘bubble territory’ remark. South Korea’s chip-heavy index has experienced swings exceeding 10 percent in a single session. The Nasdaq fell sharply when Apple signalled higher chip costs. These are the tremors of a market where positioning is overwhelmingly one-sided, and conviction is shallow. Large price movements in response to relatively minor news are characteristic of a crowded, over-owned trade, as there is no longer sufficient diversity of opinion to absorb shocks.

The leverage running underneath

Concentration alone produces volatility. Concentration plus leverage produces forced selling, and forced selling is what turns a correction into a crisis. The AI build-out has been financed to an unusual degree with debt, and the structure of that debt is where systemic risk is concentrated.

Technology companies have flooded the corporate credit market, raising hundreds of billions to finance AI projects and exploiting credit spreads near their lowest levels in decades. This detail is critical. Spreads at such lows mean lenders receive minimal compensation for the risks they assume, which is a classic precursor to a credit repricing. When spreads are compressed to this extent, there is no buffer. Any perceived increase in risk triggers a rapid widening, and the cost of capital for the entire AI sector adjusts abruptly.

Then there is the opacity of financing, which the BIS repeatedly singled out. The AI ecosystem is held together by what the central bankers described as a complex web of financial ties between AI giants, shadow banks and data-centre builders. Some of these ties are genuinely circular: chipmakers extending loans to developers who use the money to buy the chipmaker’s chips. Vendor financing of this kind inflates demand, as it lets a supplier book revenue it has effectively funded itself, and it means apparent end-demand may be softer than headline numbers imply. Circular financing is a hallmark of late-cycle booms precisely because it manufactures the appearance of organic demand while quietly concentrating credit risk on the balance sheets of the very firms whose valuations depend on that demand being real.

Private credit or the so-called “shadow-banking channel” is the part that should worry anyone who remembers 2008, and it is why the central bankers explicitly reached for the second-GFC comparison. Private credit funds have piled into AI data centres to capture the yield on offer. The shadow-banking sector has expanded rapidly precisely because post-crisis regulation tightened the rules on mainstream banks, pushing risk into less-regulated, less-transparent vehicles. The result is a large, opaque pool of leverage extending into the most speculative corner of the cycle, sitting outside the regulatory perimeter that was built after the last crisis specifically to contain this kind of risk.

This stress is not theoretical. The BIS observed that signs of strain are already apparent in private credit, with many funds overwhelmed by redemption requests and, in some cases, compelled to gate withdrawals, thereby preventing investors from accessing their capital. This is the most under-appreciated mechanism in the entire structure, and it is important to be precise about why it is dangerous.

Many private credit vehicles offer investors periodic liquidity while holding fundamentally illiquid assets, such as long-dated, hard-to-value loans to data-centre projects and AI-related borrowers. This creates a liquidity mismatch, the same structural flaw that has caused the collapse of vehicles in every modern liquidity crisis from 2008 to the 2020 dash-for-cash. When sentiment shifts, investors rush to redeem first, knowing the assets cannot be sold quickly enough to satisfy all requests. Gating may temporarily halt the run, but it also signals to other investors that the exit is narrow, which increases pressure on other funds with similar exposures. Gating one fund becomes a signal to redeem from the next. Stress propagates through the system not in spite of the gates, but partly because of them.

For the leveraged holder of AI equities, the sequence is mechanical and rapid. A drawdown triggers margin calls, which in turn force sales into a declining market. Falling prices prompt further calls and breach covenants on the debt side. Lenders, observing spreads widen, withdraw financing. This is precisely the ‘sudden pullback in financing’ the BIS warned could transform the capex boom into a bust. Leverage removes the investor’s ability to wait out a drawdown. It converts a paper loss into a realised one at the worst possible moment, and it does so for many holders at once, since they are all margined against the same correlated collateral. The deleveraging of a crowded, leveraged trade is never orderly. It resembles a stampede through a single exit.

The capital that went somewhere else — or nowhere

The third problem receives the least attention, as it is a story defined by absence. When a disproportionate share of available capital is allocated to a single theme, it becomes unavailable for other uses. The AI boom has not unfolded in isolation; it has taken place against a backdrop of finite investable capital, and the opportunity cost, though difficult to observe, is nonetheless real.

Consider the destination of the marginal dollar. Record-high technology share prices have channelled equity issuance toward AI-linked companies. The corporate credit market has been dominated by technology borrowers taking advantage of tight spreads. Private credit has shifted toward data centres. Each of these flows represents capital that did not reach manufacturing, infrastructure outside the AI sector, smaller companies lacking an AI narrative, healthcare, energy transition projects, or the less glamorous productive economy that does not command a thematic premium. Capital allocated according to narrative rather than return on capital is, by definition, misallocated. The broader the narrative’s gravitational pull, the more investment across the economy is directed by a single story rather than by dispersed judgement across thousands of individual opportunities.

This is the classic signature of a bubble that the BIS identified by analogy. The canals of the 1830s, the British railway mania of the 1840s, the dotcom build-out of the late 1990s, each featured, in the BIS’s framing, a genuine technological breakthrough that attracted capital in excess of what commercial returns could ultimately justify. The technology was real in every case. Canals worked. Railways transformed Britain. The internet changed everything. AI may well raise productivity significantly over the coming decade. Whether the breakthrough is real is not the question. The question is whether the capital committed can earn a return commensurate with its cost, and history’s answer in these episodes was—not because the technology failed, but because too much money chased it too quickly, building capacity faster than profitable demand could absorb. Each episode, the BIS noted, ended with a reversal in investment.

There is a particular harshness in how this resolves for the sectors deprived of capital. During the boom, these capital-starved sectors underperform, which appears to validate the decision to avoid them and further restricts their access to funding. This self-reinforcing cycle makes the concentration seem ever more justified until it abruptly reverses. Companies unable to raise capital due to the absence of an AI narrative were not necessarily inferior businesses; they were simply excluded by prevailing sentiment. When the narrative collapses, the rebound can be abrupt in the opposite direction, but much value will have been destroyed in the interim, and some businesses will not survive to benefit from the eventual rotation.

For the smaller end of the equity market, this dynamic has been particularly punishing. Capital concentrated in mega-cap AI names has left smaller companies trading at historically wide discounts, irrespective of their individual fundamentals. This is the troubling hallmark of a top-heavy market: widespread underperformance beneath the index leaders means that the so-called diversified investor is, once again, less diversified than assumed. The index return is supported by a handful of names while the median holding stagnates. The index appears robust. The underlying market is not.

What the reversal looks like

So suppose the central bankers are right and the AI thesis disappoints- not that AI fails, but that returns arrive more slowly or are smaller than the assumed capex. What actually happens, and why is it worse than an ordinary correction?

The immediate effect is a repricing of the AI sector itself. Disappointing returns prompt investors to restrict financing for AI companies, which is the central mechanism identified by the BIS. Equity multiples contract, and because these names comprise such a large share of the index, the broader market declines alongside them. The retail investor who believed in diversification discovers otherwise. The super fund member’s so-called balanced default suffers a loss tied to a single theme. There is no refuge within the index, because the index itself was the bet.

The next effect is the previously described leverage unwind, and this is where a correction escalates into a crisis. Margin calls force sales. Tight credit spreads widen rapidly because there is no buffer. Lenders withdraw financing across the AI supply chain. As the BIS noted, if hyperscalers reduce or halt their capital expenditure, borrowers throughout the chain may struggle to replace lost revenue and service their debt. The data-centre builder, the chip supplier, and the private-credit fund that lent to both are all exposed to the same slowdown, and their outcomes are correlated rather than independent. The opacity of the financing structure means no one knows precisely where the losses reside, and uncertainty about the concentration of risk becomes contagious. When it is unclear which counterparties are impaired, participants withdraw from all of them. This is the process by which a private-credit issue becomes a banking problem, and then a systemic one. It is the 2008 pattern applied to a new asset class.

The third effect is macroeconomic and affects individuals who have never owned an AI stock. Optimism about AI has provided a significant tailwind to global growth. Remove this, and a substantial portion of recent economic momentum disappears. The BIS made clear that reversals following historical investment booms have triggered economy-wide recessions, and it warned that a major AI-linked equity correction could have a greater impact today than in previous episodes, as households now hold a larger share of their wealth in equities relative to income. When equity wealth contracts, household spending declines, and the slowdown spreads from the financial sector to the broader economy. The wealth effect operates in reverse. The individual who loses employment in the downturn may have no connection to AI at all. This is what transforms a concentrated bubble into a public problem, rather than merely an issue for those who participated.

This reversal would occur against an unusually fragile macroeconomic backdrop, eliminating the shock absorbers that typically mitigate a downturn. The BIS highlighted persistent inflationary pressures, exacerbated by energy disruptions near the Strait of Hormuz, alongside deteriorating fiscal positions, as many governments have failed to restore public finances during the recent period of growth. Elevated inflation limits the extent to which central banks can reduce rates to support markets, while strained public balance sheets restrict governments’ ability to spend in response to a downturn. The policy tools that softened previous busts are now partly depleted. An AI unwind would strike an economy with less capacity to respond than usual, which is precisely the combination that can transform a recession into a prolonged one.

What follows from all this

None of this requires the belief that AI is fraudulent or that the technology will fail to deliver. The honest position is the uncomfortable one: the technology may be genuinely transformative, while the financial structure constructed around it may be dangerously excessive, and these two realities unfold on different timescales. The technological breakthrough develops over a decade. The capital cycle unfolds over quarters. It is entirely possible, and historically the norm, for long-term technology to prove its worth even as the near-term investment boom destroys substantial capital. Railways transformed Britain; railway investors suffered ruin.

For investors, the practical implications arise from these three problems rather than from any attempt at market timing, since the timing is genuinely unknowable and those who claim otherwise are not to be trusted. The concentration problem suggests that investors should look beyond the index to assess true thematic exposure, rather than relying on the diversification claimed by asset-allocation statements. The leverage problem requires treating liquidity terms as a primary risk, understanding exactly what can be sold and when, and maintaining deep scepticism toward vehicles that promise liquidity against illiquid collateral, as such promises are most likely to be broken at the worst possible moment. The capital-starvation problem presents an opportunity as well as a warning: sectors and smaller companies deprived of capital during the boom are, on a return-on-capital basis, precisely where neglected value tends to accumulate, even if they remain difficult to hold while the concentration persists.

The more fundamental issue concerns positioning rather than prediction. The argument for caution does not depend on forecasting the precise timing of a bubble’s collapse. It is based on the observation that the market’s structure- the concentration, leverage, opacity, and crowding has rendered the cost of error asymmetric. If the AI thesis delivers as expected, the overweight investor benefits, but only modestly more than a balanced one. If it disappoints, the overweight, leveraged, index-tracking investor faces forced selling into an illiquid, correlated unwind with limited policy support. When the payoff is skewed in this way, prudence does not require betting against the technology. It requires avoiding a position where even a delay in returns, rather than outright failure, can inflict permanent damage. The central bankers are not forecasting a crash. They are highlighting that the market is now structured so that any correction would be unusually severe. These are distinct points, and the latter merits attention.

The instinctive reading is comforting. Comparable mortgage offerings in the same market quote net returns of 11-13 per cent. A lower rate, on similar collateral and with the same security ranking, appears to indicate stronger credit. Lower yields are supposed to reflect lower risk. This relationship is embedded in every finance textbook.

In Australian property private credit, however, the relationship does not hold. This analysis examines why, what the structure conceals, and how to look beyond it.

What “after fees and costs” hides

The most important phrase in the offer is the one most likely to be skimmed past. “After fees and costs” indicates what the investor will receive. It does not tell them what the borrower is paying, what the manager is taking, or how the difference between those two numbers compares with the market’s pricing of the underlying risk.

The typical assumption is that the manager charges an annual management fee on the fund, perhaps 1.5 to 2.5 per cent per annum, and that the difference between the borrower’s rate and the investor’s net is roughly equal to this fee. If this were the entire picture, the calculation would be simple: a 9.4 per cent net to the investor would imply a gross interest rate to the borrower of approximately 10.9 to 11.9 per cent.

This is not the complete picture. The fund manager’s revenue from the deal is derived from at least four distinct streams, of which only one is the annual management fee deducted from the interest coupon.

The first is an establishment fee, typically in the order of 2 per cent of the total loan amount, payable by the borrower at drawdown. On a twenty-four-month facility, this annualises to approximately 1 per cent per annum on the outstanding balance.

The second is the annual management or portfolio fee deducted from the interest coupon, ranging from 1.5 to 2.5 per cent per annum, depending on the structure.

The third is an ongoing loan management or administration fee, often quoted separately from the portfolio fee, that may add a further 0.5 to 1.0 per cent per annum to the borrower’s all-in cost.

The fourth is a completion or exit fee, typically 1 to 1.5 per cent of the loan amount, payable by the borrower at maturity. On a two-year facility, this amounts to an additional 0.5 to 0.75 per cent per annum.

The total annualised revenue the manager extracts from the borrower across these four streams is typically 3.5 to 5.0 per cent per annum. Of this, only the second stream, perhaps 1.5 to 2.5 per cent, is what the investor might recognise as the fund’s fee. The other three are generally invisible to the investor, as they are charged to the borrower and not separately disclosed in the offer letter. This matters because the hidden charges help convert borrower risk into manager revenue.

When the full fee load is included, an investor receiving 9.4 per cent net is, under the most common fee structure, participating in a deal where the borrower’s all-in annualised cost of capital is approximately 13 to 14 per cent. The manager is capturing a spread of roughly 4-5 per cent per annum between what the borrower pays and what the investor receives. The investor therefore receives less income than the borrower pays, while having visibility on neither the borrower’s total cost nor the manager’s aggregate revenue. That spread is the difference between borrower risk pricing and investor return.

This arrangement is neither unusual nor improper. All funds management products charge fees for origination, diligence, monitoring, and administration, and the establishment, management, and completion fee structure is standard across the segment. What is distinctive in private credit is that the investor in a contributory or co-investment structure lacks direct visibility into the gross rate, the borrower-side fee schedule, or the cumulative manager revenue over the life of the deal. The borrower’s contract is with the manager, not the investor. The fee schedule appears in the information memorandum, often in language that a typical investor reads only at the outset. The net return is the figure that is marketed, repeatedly, in successive offers.

The first unknown for the investor is the borrower’s actual cost once the full fee structure is considered. The second is how that all-in borrower cost compares to the pricing and how similar the risk is in the broader market.

The mark the investor isn’t shown

For senior secured lending against income-producing commercial property at 50 per cent LVR, the appropriate market spread over BBSY in current Australian conditions is approximately 250 to 400 basis points. For senior secured non-bank residential mortgages, perhaps 200 to 300 basis points. For raw land with no income, planning-timeline risk, exit dependent on a refinance or sale into an unknown future market, no operating cash flow to service the debt, I find that the historical and global benchmark for the all-in cost that a senior lender should demand has been 600 to 900 basis points over base.

The midpoint of that raw land range, applied to a current BBSY of around 4.3 per cent, implies a benchmark all-in cost to the borrower of 11.3 to 13.3 per cent. The 13 to 14 per cent all-in cost the borrower is paying in the example is, broadly, at or slightly above this benchmark. The borrower is paying a rate consistent with how raw-land senior risk has historically been priced. This structural fact reframes the analysis. The key takeaway is not that the borrower is underpriced or the manager is discounting risk. It is that the manager is extracting the appropriate risk premium from the borrower, but only a portion is passed to the investor, with the remainder retained as fee revenue.

The investor receiving 9.4 per cent net is being compensated at approximately the rate appropriate for senior secured commercial property lending against an income-producing asset, which is a fundamentally lower-risk loan than a raw land loan. In a downside scenario, the investor still bears raw-land risk but receives income-property risk pricing in the positive scenario. The difference between those two is the additional risk premium that the borrower is paying, but the investor is not receiving, which goes to the manager across the four fee streams described above. The takeaway is clear: a lower net rate does not mean the investor is safer; it means the manager is retaining more of the risk premium.

This is the more acute version of the structural issue. The borrower is not being undercharged. The investor is being undercompensated for the risk they are assuming, with the difference accruing to the manager rather than to the party exposed to loss in an adverse outcome. The investor therefore bears the downside without receiving the full risk premium. The practical takeaway is that a lower quoted net return can mask a worse risk-adjusted position for the investor.

Why a low net rate is not a safer position

Here is the part that contradicts the textbook intuition. In broadly efficient markets, where rates and risk are connected by the discipline of repeated trading and comparable issuance, a lower rate received does signal a lower-risk position. Investment-grade corporate bondholders earn less than high-yield bondholders because the underlying credit risk is different.

In private credit co-investment markets, the discipline that makes the rate-risk relationship robust is much weaker. The investor’s net return is determined not only by the underlying credit risk but also by how much of the risk premium the manager retains in fees before passing the residual to the investor. Two investors in two different funds can be exposed to identical underlying risk to the same kind of borrower, the same kind of collateral, the same kind of exit risk and earn materially different net returns. The difference is not in the credit risk being borne. The difference lies in the fund’s fee structure.

This is the core of the misperception. An investor sees a 9.4 per cent net return and is inclined to believe this represents a safer deal than a contributory mortgage offering 12 per cent net. In reality, both products may be lending against similar risk, with the difference in net returns reflecting how much of the borrower’s all-in cost is retained by the manager rather than passed through to the investor. The investor’s lower quoted return does not mean a lower level of risk. The fee split, not the credit risk alone, drives the lower net rate.

In a downside scenario, the investor’s position in the lower-quoted deal is no safer than that of the investor in the higher-quoted deal. If the loan defaults, both investors encounter the same recovery process, workout costs, probability of capital loss, and resolution timeline. The only difference is that the investor in the lower-quoted deal has accumulated less coupon income during the performing period, leaving a smaller buffer to absorb any eventual loss and less compensation for the risk taken.

This is the structural feature of the segment that the cover-sheet presentation conceals. The net return does not compensate the investor for the credit risk they assume. Instead, it reflects the portion of the borrower’s all-in cost that the manager elects to pass through after extracting establishment, ongoing, portfolio, and completion fees. The fee architecture has partially decoupled credit risk from investor compensation, and this decoupling is not apparent in the disclosed terms. As a result, the investor may see a lower quoted return without receiving a safer exposure. The result is a return that is less closely aligned with the risk borne. The core takeaway is that the quoted net return should not be read as a safer position; it should be read as a smaller share of the same underlying risk premium.

The structural pattern

In Australian property private credit specifically, this dynamic is reinforced by several structural features.

The first is the manager’s incentive structure. The portfolio management fee is the most visible revenue stream and the one investors focus on, but it is rarely the largest contributor to the manager’s economics on a per-deal basis. Establishment fees of 2 per cent on a $20 million facility generate $400,000 of revenue at drawdown, received before the first interest coupon is paid. Completion fees generate similar amounts at exit. Ongoing loan management fees add a recurring layer. A manager with a well-developed fee architecture earns more from origination and exit than from holding the loan in the portfolio. This shifts the manager’s commercial incentive toward deal velocity, prioritising origination, drawdown, and refinancing or maturity, rather than maximising the investor’s spread on each deal. The headline product is the net return because it is what the investor receives and is most readable. The gross rate to the borrower, the establishment fee, the ongoing loan management fee and the completion fee are detailed in the information memorandum but not on the cover sheet. A retail investor or family office adviser comparing offers will tend to compare net returns directly, because that is the comparable presented to them. Two offers, one at 9.4 per cent net and one at 11.5 per cent net, will appear to differ in the credit quality of the underlying loans. They may, in fact, differ only in the manager’s fee architecture, with the higher-quoted offer having a lower total fee load and passing more of the underlying borrower payment through to the investor.

The third is the time profile of returns. A senior debt fund pays its return in coupon income, predictably, month after month. If a loss occurs, it typically materialises at maturity, when the borrower fails to repay, and the workout process begins. Three or four years of uninterrupted coupons in a 9.4 per cent product are indistinguishable from those in an 11.5 per cent product. The investor’s experience of a performing deal is identical. The difference only becomes apparent if a loss occurs, at which point the lower-yielding fund has accumulated a smaller buffer of unconsumed return to absorb the loss, while the manager has already collected the establishment, ongoing, and completion fees from the performing deals.

The fourth is the comparison set. An adviser constructing a private credit allocation often selects two or three managers from this segment, each with similar features and structures. The cross-manager comparison reinforces the perception that this is a homogeneous asset class with predictable returns. The reality that investors are systematically receiving a smaller share of the borrower’s all-in cost than the underlying credit risk would justify in a market priced for the investor’s position, and that the fee architecture has shifted the economics in the manager’s favour across the segment, is not apparent from within the segment itself.

What an investor should actually ask

Three questions cut through this structure if asked directly to the manager and answered honestly.

What is the borrower’s all-in annualised cost of capital? Not just the headline interest rate, and not just the net return to the investor, but the borrower’s effective cost, taking into account the establishment fee, the ongoing loan management fee, the portfolio fee, and any completion or exit fee, annualised over the loan term. A manager who can answer this question with a single number is one who has worked it out themselves. A manager who has to defer to the information memorandum, or who provides the components separately rather than aggregating them, is signalling either that they have not done the calculation or that they prefer the investor not to see it.

What is the manager’s aggregate revenue from this deal, expressed as a percentage of the loan amount annualised over the loan term? This is the complement to the first question. The manager’s revenue is the gap between the borrower’s all-in cost and the investor’s net return. A manager whose total annualised take is 250 basis points on a deal of this risk profile is in one position. A manager whose total annualised take is 500 basis points is in a different one. Both may be commercially defensible. The investor’s job is to know which they are participating in.

How does the all-in borrower cost compare to where you would price this same risk if you were quoting it for the first time today, with no prior relationship and no competitive pressure? This isolates whether the manager believes the deal is priced consistently with the underlying risk or is priced for competitive or relationship reasons. A manager who answers precisely is one who has thought about the spread. A manager who deflects to a discussion of the collateral, the sponsor, or the loan-to-value ratio is asking the investor not to look at the spread.

A careful reading of the information memorandum provides answers to most of these questions. The cover-sheet pitch provides none. The investor who examines the information memorandum, calculates the all-in borrower cost, and compares it to their net return has undertaken the analysis the cover sheet is designed to bypass.

The deeper point

Australian private credit has been the success story of the past decade in retail-accessible wholesale investment products. Returns have been consistent, defaults have been low, and the structures have worked. Much of that performance has been earned by managers through real work such as origination discipline, careful credit analysis, structural protection in loan documents, and active portfolio management.

However, the disclosed net return has absorbed much of the risk premium that the underlying credit risk did not justify. The four-layer fee structure, which includes an establishment fee, ongoing loan management fee, portfolio fee, and completion fee, diverts a significant portion of the risk premium paid by the borrower to manager revenue rather than to investor compensation. As a result, two products with similar underlying credit risk can present very different net returns to the investor, with the difference reflecting fee architecture rather than risk profile. This is not visible from the cover sheet, which quotes the net return, a figure that conceals the architecture rather than revealing it.

This structure is sustainable only while conditions remain favourable. The cycle in which Australian private credit has expanded during a prolonged period of low base rates, supportive property markets, and limited defaults has not been tested against the current environment. Higher base rates, tighter credit conditions, lower property valuations, revised tax settings for investor property, and increased stress in the construction sector now characterise the mid-2026 landscape. When losses begin to emerge, as they have historically in raw land senior credit, the investor’s outcome is determined by the actual loan exposure and recovery, not by the net return received previously. By that point, the manager has already collected the establishment, ongoing, and possibly completion fees on the performing deals. The investor who assumed the credit risk and received the residual net return bears the loss.

The investor who recognises the fee architecture, asks the right questions, and reads the information memorandum is better positioned than one who relies solely on the cover sheet. An investor who treats the headline net return as the main indicator of credit quality is making a fundamental error. The net return is not a measure of risk. In this market, it is an output of fee architecture that operates largely out of the investor’s view, with the underlying risk premium divided between manager and investor in proportions that are not disclosed.

Take a representative example: a $15 million senior first mortgage over a large parcel of raw land in a Victorian growth corridor, stated at 51 per cent LVR. The borrower pays interest monthly, and the facility matures in twenty-four months. The 51 per cent LVR implies a current land valuation of approximately $29.4 million, a figure that anticipates development approval well in advance.

Three further facts alter the interpretation of this LVR.

First, the same parcel sold in 2024 for about $20 million. The current valuation represents a 47 per cent increase over two years.

Second, the land lacks development approval and is not expected to obtain it until 2033.

Third, the loan matures in 2028, five years before development approval is anticipated.

Each fact is material on its own. Taken together, they reveal a transaction that differs from what the stated LVR implies: the valuation is driven by speculation rather than by the loan structure alone.

Where the 47 per cent valuation increase comes from

For raw land in a growth corridor, price consists of two elements: the value of current use (typically agricultural or rural-residential, often tens of thousands of dollars per hectare for grazing land) and a speculation premium reflecting anticipated future rezoning. On a 160-hectare parcel, the agricultural value typically ranges from $3 to $8 million. Any valuation above that range is, by definition, a speculative premium on future development potential.

In this example, the $20 million sale price in 2024 implied a speculation premium of about $15 million above agricultural value. The $29.4 million valuation in 2026 implies a speculation premium of about $24 million. This premium has increased by approximately 60 per cent over two years. During this period, no development approval has been granted; the expected approval date remains seven years away; no infrastructure milestones have advanced the land’s readiness; and the broader Australian property and credit environment has tightened.

What accounts for a 60 per cent increase in the speculation premium on the same parcel, despite no change in development status and a softening in other property valuations? The most likely explanation is that comparable sales of similarly speculative parcels have continued to rise, and the valuation method has marked the subject parcel to those comparables. This approach is professionally defensible. It is also the same methodology used for raw land in Auckland’s growth corridors before that segment declined by 37 per cent in real terms over the following three years.

Where the LVR could go

A simple sensitivity illustrates the risk. If the valuation returns to the 2024 sale price of $20 million, the effective LVR on the $15 million loan increases from the stated 51 per cent to about 75 per cent. If it falls further, to a level reflecting agricultural value plus a modest speculation premium, the effective LVR could exceed 100 per cent. These outcomes do not require a severe property downturn. They require only that the speculation premium on raw land in Victorian growth corridors adjusts to current credit conditions, prevailing rates, and an extended timeline to development.

The 51 per cent LVR is a snapshot, tied to a valuation that reflects prevailing market sentiment. In raw land senior credit, LVR is not as stable a metric as it is in residential or commercial mortgages on income-producing property. It shifts with the speculation premium, which in turn is driven by sentiment, comparable sales, and the broader interest-rate environment.

The timeline mismatch

A second structural issue is the gap between loan maturity and the development approval timeline. The loan matures in 2028, while development approval is expected in 2033. At maturity, the land will not be shovel-ready, nor will it be suitable for sale to a developer or for refinancing with a development-construction lender. The only available exits are a refinance with another speculation-stage lender or a sale to another speculator. The stated maturity does not align with the asset’s path to value.

Both scenarios depend entirely on the state of the speculation market in 2028. If the market is strong, the loan is refinanced and the investor is repaid. If it is weak, the borrower cannot refinance at the current valuation, the loan defaults or is extended, and the investor’s recovery depends on a workout against a parcel whose effective LVR is no longer 51 per cent.

This is an inherent feature of raw land senior credit with a short tenor relative to the development timeline. It is present in every such transaction in the segment. The headline LVR does not disclose this risk.

What to ask

Three questions to ask before considering a raw land senior credit offer:

What was the most recent arm’s-length sale of the underlying parcel, and at what price? This benchmarks the current valuation against an actual transaction rather than a comparable-sales model.

When is development approval expected, and how does that timeline compare to loan maturity? If approval is granted after maturity, the exit is into the speculation market rather than the development market, and the relevant credit risk is driven by sentiment rather than construction.

What is the agricultural or current-use value of the parcel, and what is the implied speculation premium? This breakdown is rarely provided in offer documents, yet it is the most useful figure for understanding the borrower’s collateral in a downside scenario.

LVR is not a constant. It is a function of valuation, which in turn depends on speculation. An investor who treats LVR as a binding measure of credit exposure is relying on the wrong metric. In raw land senior credit, the stated LVR can significantly understate how rapidly risk can change.