Whether a lender loses money was decided when the loan was written. How much it loses is being decided now, on site - and that is a construction question almost nobody holding the loans is equipped to answer.
We have been on the lender side of a project takeover. Here is what actually happens. The loan documents describe an orderly step-in: replace the builder, complete the works, sell the stock. What you inherit instead is a half-built site with incomplete records, unknown defects, certification gaps, subcontractors who are owed money and have already moved their crews, and a quantity surveyor's certificate stating - accurately - that the building is 60 per cent built. What no certificate states is whether anyone can afford to finish it. Every replacement builder prices that uncertainty at 30 to 40 per cent above the old contract, guarantees nothing it did not build, and wants an indemnity for the rest. And the lender is slow. It assesses, commissions reports, seeks approvals - a process that takes months, because lenders are conservative precisely when speed is worth the most. All the while, interest capitalises, subcontractors take other jobs, and sunset dates creep closer.
Keep that scene in mind, because 43 lenders - most of them private credit - are about to live it simultaneously.
On 25 August, Bathla Group - among western Sydney's largest builder-developers of lower-cost housing - entered voluntary administration owing about $3.4 billion, with 2,000 homes part-built and 13,000 to 15,000 more in the pipeline. Within days two managers had suspended redemptions across funds holding Bathla loans, a listed mortgage trust had halted trading, and a fourth had capped withdrawals at a fund with no exposure at all. The commentary has settled into a familiar groove: too much debt, liquidity mismatch, valuations held at par, a regulator that warned about all of it last year. All true, all now consensus, and none of it answers the only question that decides the size of the losses.
Our claim in this issue is simple. Whether a lender loses money was decided when the loan was written - that was the credit question, and for much of this market it has already been answered. How much it loses is being decided now, on site - and that is a construction question almost nobody holding the loans is equipped to answer.
For decades a project stood on three legs: the developer took sales risk, the builder took cost risk under a fixed-price contract carried on its own balance sheet, and the lender relied on both with step-in rights over each. One party's failure did not take down the others - the building got finished. After 2021, costs rose 20 to 30 per cent, builders on fixed prices absorbed it, and the ones without balance sheets failed. Survivors stopped signing fixed prices; banks would not lend without one; so developers brought the builder inside - in-house arms, JVs, captive builders - and private credit funded the integrated result: more of the cost, fewer presales, interest capitalising rather than being paid. The two hedges that kept cost risk, sales risk and lender risk apart were gone. Nobody repriced the loans to reflect it.
Bathla is this model at scale: developer and builder inside one group of 542 entities, a thin-margin volume machine that serviced its debt only while it grew.
The reason integration is lethal to lenders is not just that a cost blowout and a sales shortfall now land on the same equity. It is that the lender's own check quietly stopped checking anything that matters. The QS certifies work in place each month - a physical check. It does not see the commercial position underneath: what the trades were actually let for, which variations are unpriced, which time claims are live, whether subcontractors have been paid. On an arm's-length job those are the builder's problems, watched by a developer with its own money at stake. In an integrated group there is no second set of eyes, and the builder's problem is the project's problem. Sponsor-level diligence - the group as a whole, not project by project - is the discipline that was skipped, and with 542 entities it is easy to see why: no single lender could see the whole. When the group fails, the builder fails in the same moment, and the lender inherits the scene in our opening paragraph.
The lender's own check quietly stopped checking anything that matters. The QS certifies work in place. It does not see the commercial position underneath.
The sector's favourite comfort metric is a loan-to-value ratio quoted against net realisable value - the forecast proceeds of a completed, sold-out project. Here is what it is worth on a half-built site, in stylised but honest numbers.
Take a project underwritten at $100 million NRV with a $65 million facility - “conservative” 65 per cent. It fails at 60 per cent complete with $52 million drawn. Remaining build cost was $24 million at the old contract; a replacement builder pricing blind adds 35 per cent, call it $32 million. Six to nine months of standstill plus a twelve-month completion adds roughly $9 million of capitalised interest at private credit rates. Meanwhile the market the stock settles into has fallen - settlement cuts become the comps - so the $100 million book realises perhaps $90 million gross, and after selling costs, GST and defect friction, call it $80 million net.
The lender now has two options. Fund to completion: total money out roughly $93 million against $80 million back - an effective LVR of 116 per cent, and a loss even for the senior. Or sell the site today: a buyer pays completed value less completion cost, less its own margin and risk discount - roughly $30 million - which is about 60 cents in the dollar on the amount drawn. That is the real menu behind “65 per cent LVR”: complete at 116 per cent, or exit at 60 cents. Junior and mezzanine positions in that stack are not impaired; they are gone.
The numbers move with the assumptions, but the direction does not. And note where the 10 per cent price fall comes from - not a market crash, but the discounts a distressed completion generates itself, as forced settlements become the comps for everything left to sell. Strip it out and run the numbers at flat prices: completion still loses money, and the sell-now exit still sits near 77 cents in the dollar. A failed builder, on its own, consumes the whole of a “conservative” 65 per cent buffer. A falling market is what turns a bad outcome into a terrible one.
One more piece of arithmetic, from the administration itself. Teneo's opening request was $20 million to keep sites moving for five weeks, and the administrators have since put the cost of simply supporting construction across roughly 45 active sites at $1 million to $1.3 million a week - payroll and critical subcontractors: care and maintenance, not progress. That request tells you there is no cash buffer anywhere in a $3.4 billion structure, and it previews the burn every stalled project in this cycle will run while its lender deliberates.
A presale is unrealised revenue: a promise to buy in two or three years at today's price, with 10 per cent down. Every buyer must be re-approved for finance at completion, and their bank values the finished apartment against recent sales, not the contract. On an $800,000 contract with an 80 per cent loan, a $720,000 valuation leaves the buyer $64,000 short in cash. The developer's remedies - keep the deposit, sue for the shortfall - take a year-plus and recover from people who by definition did not have the money, and past the sunset date the buyer walks owing nothing. A developer whose loan falls due at completion cannot wait, so it cuts prices to force settlements through - and each cut becomes a registered comparable that lowers the valuation of the unit next door and the unsold stock its own lender is secured against. In our experience, bank-funded projects carry roughly half their debt in presales; private credit projects borrowed more against fewer. Both survive a 10 per cent price fall on paper. The privately funded one fails on time, not price.
Issue 4 argued the May Budget would force a yield adjustment through investor-grade property. Bathla's management blames those changes for the collapse, which is awkward: on paper the Budget favoured them. New builds kept negative gearing and can elect the old CGT treatment; the policy was designed to push investors toward exactly what Bathla sold. It hurt them anyway, through transition dynamics rather than tax positions. A new build is only new once - every resale buyer holds the less favourable regime, so first buyers discount today's price for the impaired exit. New-build valuations anchor to established comps, which the policy pushed down, so settlement valuations came in under contract prices. Grandfathering locked investors into existing holdings, choking the equity-recycling that funds off-the-plan purchases. And the months between announcement and legislative clarity were an air pocket a leveraged volume model could not fly through. The adjustment we described transmitted through comps and exits, not through anyone's tax return. Design intent does not govern the transition; transition dynamics do.
Broadly, apartment and townhouse projects in this market are being delivered through one of four structures, and the structure decides who breaks first.
Separate developer, separate builder on a genuine fixed price, lender holding its own agreement with the builder. Failure mode: the builder loses its margin to escalation - most acutely on 2024 and early-2025 tenders, which carried contingency of 3 to 5 per cent of the contract against costs now escalating at 5 to 7 per cent a year - on a two-year build, the escalation eats the entire buffer several times over - and fails; the project survives with a new builder, slower and more expensive than its lender expects. Painful but contained, because it only works where the builder has a balance sheet, and below tier one almost none do.
Separate company, own licence, fixed price on paper - but one client. It cannot push back on price, claims or payment terms without losing the relationship that keeps it alive. To a lender it looks arm's-length and gets funded like it; it behaves like in-house. When money tightens, the developer cuts it loose to preserve cash. We would expect this to happen more often than not.
The builder holds equity, so it is protecting its own position when it prices claims, and its balance sheet is exposed on both sides of the deal. Typically funded non-bank senior with mezzanine behind it. Failure mode: escalation eats margin and equity at once, the partners cannot agree who funds the shortfall, the lender enforces.
Developer owns the builder; no risk transfer at all. This one splits purely on balance sheet. A group that can fund its own overruns is among the most resilient structures in the market. A thin group on stacked debt with few presales is the most fragile thing in it - any one of an overrun, a delay, a settlement failure or a gated lender at renewal takes the whole group, and there is nobody left under contract to finish. Bathla is the thin version.
The rule across all four: the delivery model decides who breaks first inside the relationship; the capital stack decides how fast the sponsor follows. Senior debt plus real sponsor equity mostly survives in any model. Anything stacked at three or four different rates is the most aggressive structure in the market and goes first, whatever it calls itself.
Commentary is cheap this month, so here are ours - specific enough to be marked right or wrong, and we will score them publicly as the cycle runs. For transparency: Northcliffe advises clients in this market, so we have an interest in the situations we describe - which is precisely why the calls belong on the record, marked honestly.
More redemption gates before Christmas, including at funds with no Bathla exposure - precautionary redemptions do not read exposure lists.
The early administrator sales will be the good marks. Funds will sell distressed positions quickly at moderate haircuts rather than work them, because suspending distributions to preserve capital triggers the run it is meant to avoid. The desperate sellers arrive later and worse - part-complete construction positions clearing at deep discounts to their book values, and junior positions largely wiped out. The first receivership - over 65 completed townhouses in Kellyville, appointed on 31 August - is exactly where we would expect the early marks to come from: finished stock, not half-built.
More than one manager exits the industry within twelve months - wound up, replaced by their trustee, or forced to sell their book - driven by the liquidity structure alone, without material credit losses. A gate protects the fund, not the manager: a gated manager cannot write new loans, so the fee income stops; platforms and ratings houses walk; and the fund's own financiers are not bound by the gate. The fund runs off while the business behind it disappears.
Captive builders fail first - cut loose by developers choosing insolvency later over insolvency now. Shedding the builder sheds its accrued losses and buys months; it does not finish the buildings. Developers who can fund a replacement survive the trade; the rest follow their builder down - stacked-finance sponsors leading, as facilities mature into gated lenders. The bulk of it by mid-2027. Structure subcontractors are already failing.
Second wave through 2027-28: sponsors solvent on paper but unable to refinance, and arm's-length projects losing builders off 2024-25 tenders. The next failure is currently performing - because the kill mechanism is the maturity date, not the arrears list.
Whilst the calls above focus on the Bathla situation, there will be some flow on to other markets in NSW and nationally. The rest of this issue is about South-East Queensland - where we often do business, and where the same trade runs on a steeper cost curve.
From late 2026 the Queensland government becomes the largest buyer of trades in the state - Games, hospitals, transport - and pays what it takes; the projected worker shortfall peaks around 35,000 in 2027-28, with escalation near 7 per cent a year through decade's end. Queensland's project trust accounts explain why months of unpaid subcontractors is a Sydney pattern, not a Queensland one - but a trust only protects money that goes in, and in an integrated group the developer's problem and the builder's problem are the same problem. Costs will not fall while private volumes do. When builders fail, their tradespeople do not get cheaper - they leave for the public pipeline or another state, or re-form without the loss-making contracts that made them cheap. There may be pockets of relief in trades the government does not buy, but the structure and services trades that drive a residential build will be bid by hospitals and rail for the rest of the decade. By 2029 we expect completions well short of what Queensland's ten-year-high approvals imply, vacancy under 1 per cent, rents and completed-stock values rising, and institutional money replacing retail private credit.
A reader may notice an apparent tension in all this: we expect prices to fall and failures to spread, and we also expect a shortage with rising rents and values. Both hold, because they land on different stock at different times - and one causes the other. The near-term falls are specific: distressed part-complete product and development land in affected corridors, driven by forced settlements becoming the comps. The shortage arrives later and lands on completed dwellings and the established market, as today's failed pipeline becomes tomorrow's missing completions. The second wave of 2027-28 does not need falling prices at all - those sponsors fail on financing, not price, as facilities mature into gated lenders and escalation eats margins; a developer can collapse into a rising market if it cannot refinance. And the causal link runs one way: while construction costs rise faster than prices, replacement cost sits above market value, nothing new starts, and what is already built becomes scarce. The distress is not in tension with the shortage. The distress is the mechanism of the shortage.
The distress is not in tension with the shortage. The distress is the mechanism of the shortage.
If SEQ apartment values rise 20 to 30 per cent over three years, and there is genuine depth of buyers able to borrow and settle at those prices, the gap closes from the revenue side and 2029 is a boom. Both conditions must hold. Higher prices without buyer depth just slow the rate of sale - and for a developer paying 12 per cent while it waits, a slower rate of sale is the same problem as a lower price.
For investors in private credit, the questions have not changed; the willingness to ask them has. What is the LVR actually measured against, and dated when? Is interest paid in cash or capitalised? Do redemption terms bear any relationship to asset duration? How concentrated is the book? And the question this cycle has added: who exactly is the builder, what balance sheet stands behind them, and what happens if the sponsor and the builder turn out to be the same pocket?
For everyone else: over the next three years there is more money in finishing projects than in starting them. The market is oversupplied with origination and structuring skill - which is partly how it got here - and desperately short of the skill that decides recoveries: establishing the true commercial position of a half-built project in days rather than months, and getting a builder back on site against a realistic budget. The value in every workout of this cycle will be won or lost in that handover window.
Will Bathla's lenders lose money? For most of them that was settled at origination, when integrated developer-builders were funded on arm's-length terms without anyone repricing for the risk transfer that had quietly disappeared. The open question is how much, and it will not be answered by the loan documents, the LVR or the QS certificate. It will be answered by what a replacement builder charges to finish a building it did not start, how many months the lender spends deciding, and what the forced settlements do to the comps for everything left to sell.
Three things follow. The distress is not confined to the funds that lent to Bathla - liquidity structures fail on withdrawal behaviour, not exposure lists. The structure of the project, not its LVR, tells you who breaks first, and the capital stack tells you how fast. And in South-East Queensland the same failure lands on a public works pipeline that will keep trades expensive for the rest of the decade - which is why the near-term distress and the later shortage are one process, not two. The long game is being on the right side of the handover window.
Four things that will tell us early whether the calls above are right.
Teneo asked for $20 million to hold roughly 45 active sites for five weeks, and has since put the running cost of supporting construction at $1 million to $1.3 million a week. If a second tranche is sought before those five weeks are out, the burn is running ahead of the administrators' own estimate and the holding-cost arithmetic for every stalled project in this cycle is worse than it looks. If the money lasts and sites are handed to purchasers or completers inside the window, the holding cost is manageable and the losses concentrate in the completion premium instead.
The prices the first administrator and lender sales clear at, against the book values they were held at. Part-complete positions clearing inside a 20 per cent haircut say the early marks are the good ones, as we expect. Anything clearing beyond 40 per cent this early means the desperate sellers have arrived ahead of schedule and the junior write-offs come forward with them.
Late 2026 is when the public pipeline starts buying trades in volume. Watch structure and services tender pricing on SEQ residential jobs against the 7 per cent escalation we have assumed. Prints above that bring 2027-28 arm's-length builder failures forward; prints materially below it, or a visible easing in structure-trade availability, would be the first evidence that private volume falling is offsetting public volume rising - and the first thing that could soften our view.
The number that decides recoveries. Every takeover in this cycle will price a replacement contract against the old one, and the premium has been 30 to 40 per cent. If competition for completion work pulls it toward 20 per cent, the complete-at-116 case improves and more lenders fund through rather than sell. If it holds or widens, more sites are sold part-built at the 60-cent exit, and each one resets the comps for the next.
The Long Game is published by Northcliffe Advisory. This issue reflects our judgement as at 5 September 2026, based on publicly reported information about an administration that is days old - figures will change. Figures on the Bathla Group administration, including debt, part-built and pipeline dwelling counts, entity numbers and the administrator's funding request, are drawn from the administrator's public statements and press reporting to 5 September 2026; lender redemption and trading announcements from the funds' own disclosures; Queensland construction workforce and escalation projections from published state government and industry forecasts. The loan-to-value, presale and holding-cost worked examples are stylised illustrations built on the stated assumptions, not modelling of any actual project or fund; actual outcomes will differ. The delivery-model descriptions and the calls in this issue are Northcliffe Advisory's own judgement and are labelled as such. Northcliffe Advisory advises clients active in this market. This is general commentary, not financial, tax, or investment advice. Readers should consider its appropriateness to their own objectives, financial situation and needs, and seek advice, before acting.
No pitch. Just a conversation about your situation.
enquiries@northcliffeadvisory.comExpert commentary on macro and markets. Delivered when it matters.