When Repayment Moves Into the Future

Future money

The half-built building

Lendy was a British property lender funded by ordinary people investing small amounts online. It started with bridging loans, which are short loans against a building that already exists. Somebody owns a property, needs cash quickly, borrows against it, and repays when they sell or refinance. If it goes wrong, there is a building to sell.

Lendy then moved heavily into development finance, which is lending to build something that does not exist yet. The company explained the difference on its own help pages with unusual clarity. For a bridging loan, the loan is measured against the current value of the security when the loan is made. For a development loan, it is measured against the value the site will have once the project is finished.

Read that twice. The security for the second loan is a building that has not been built.

Development lending is a normal and useful business, and plenty of firms do it well. But it comes with a feature that bridging does not have. A finished building can be sold. A half-finished building is worth less than the land plus the money already spent on it, and turning it into something saleable requires more money to arrive. That money usually arrives in stages, as the project passes milestones.

So the whole structure rests on a question nobody at Lendy appears to have asked loudly enough. What happens if the next money does not come?

It did not come. The administrators' report set out the sequence. From early 2017 the number of problem loans rose, investors lost confidence, and money coming onto the platform fell from £15m in the first quarter of 2017 to £5m in the fourth. The decline meant Lendy could not fully finance many of the development loans it had already committed to, which left borrowers unable to build out their schemes, which produced part-built sites worth far less than their projected finished value, which produced losses, which reduced confidence further.

Notice what is not in that chain. It does not begin with bad borrowers. It begins with money that could leave, attached to buildings that could not be finished without it. When Lendy entered administration in May 2019 after action by the Financial Conduct Authority, it had 25 live development loans with £116m outstanding against sites given a projected finished value of £265m, many of them only partly built and worth considerably less than those projections implied.

Now hold Lendy next to two firms doing similar lending in the same country under the same regulator in the same years.

LendInvest wrote bridging and development finance too. It closed its retail investment platform and rebuilt its funding around institutional mandates, bank facilities, listed bonds and annual securitization. By its most recent financial year it reported over £9bn lent since inception and record annual originations of £1.44bn. Same assets. Different money behind them.

Assetz Capital is the more revealing case, because the company said the quiet part out loud. In December 2022, facing rising interest rates that made bank savings accounts more attractive, Assetz saw modest outflows from its retail accounts and decided to close the retail platform permanently and fund entirely from institutions. The detail that matters is what it did on the way out. The company confirmed its access accounts would remain gated to ensure all loans committed, but not yet fully funded through retail investor capital, are concluded as planned. It was already roughly 80% funded by institutional money, so it could afford to lose the rest.

In plain terms, Assetz would not let today's investors take their money out while yesterday's promises still needed it. That is the exact problem that destroyed Lendy, recognized in advance and closed off.

So the first version of the pattern looks like this. If repayment depends on more money arriving later, the thing to understand is not the borrower's credit. It is whether the money can stop coming, and who gets to decide that.

Future price

The other kind of future

Now change the setting completely.

A taxi medallion is a license to operate a cab. Cities cap the number issued, which makes the license itself scarce, which makes it worth something independent of the taxi bolted to it. In New York, medallion prices climbed to roughly $1.3m. In Chicago they went from about $50,000 in 2003 to roughly $370,000 by 2013, according to Medallion Financial's own annual report.

Lenders would advance up to 90% of a medallion's value, and this was widely regarded as low risk because the market functioned as the equivalent of a regulated monopoly, with limited supply and almost no competitive threat. Banks and credit unions built substantial businesses on it, and for years the loans performed. The US credit union regulator notes that until 2014, values were generally stable or rising and delinquency rates were low.

Here is the part that is easy to miss.

A loan of 90% against a $200,000 medallion could plausibly be serviced from what a taxi earns. The driver works, the fares come in, the debt amortizes. The medallion is there as protection if things go wrong.

A loan of 90% against a $1.3m medallion is a different animal wearing the same collar. The taxi did not start earning six times as much. Nothing about the daily business changed. What changed is that the loan could no longer plausibly be repaid out of what the cab earned. It could only be repaid by the medallion still being worth something like $1.3m when the time came to sell or refinance it.

The loan-to-value ratio stayed roughly constant the whole way up, which is precisely why nobody noticed. A stable ratio against a number that has risen sixfold is not a stable risk. It is a completely different loan with the same label.

When ride-hailing apps arrived, the scarcity that the whole structure rested on evaporated. Chicago medallions that had touched $370,000 were worth about $25,000 by the end of 2019, or $19,500 after the costs of selling them. Medallion Financial's own filing puts the weighted average loan-to-value across its medallion book at approximately 190% at that date, meaning the debts were nearly twice what the collateral could fetch. The company had stopped writing new medallion loans in July 2015.

Some lenders did not survive it. Melrose Credit Union held about $1.5bn of assets and watched delinquent loans rise fivefold in two years to $670m before being wound up. Montauk Credit Union was seized by state regulators.

These were credit unions and banks, funded by insured deposits. That is the most stable money any lender can have. Depositors do not get to demand their savings back on a whim, the government guarantees them, and there is no fund manager who can decide overnight to stop buying. The funding was not the problem, and having good funding did not help.

Repayment had moved into the future here too. But the missing piece was not more money. It was a price.

Same pattern, bigger scale

The same mistake at a hundred times the size

Countrywide Financial was the largest mortgage lender in America. Among the products it pushed hardest in the middle of the 2000s was the pay-option mortgage, which let borrowers choose each month how much to pay, including an amount so small that the shortfall was added to the loan. The balance owed could grow rather than shrink.

A mortgage like that is not really repaid by the borrower's income. It is repaid when the borrower eventually refinances or sells, and both of those depend on the house being worth enough at the time. If house prices keep rising, the arrangement works and everyone is pleased. If prices stop rising, there is nothing to refinance into.

The volume was not small. Countrywide's own monthly disclosures show it funding $8.6bn of pay-option loans in July 2005 alone, and $95bn across the full year. What matters more is how much it kept. The pay-option mortgages Countrywide held on its own books grew from roughly $4.7bn at the end of 2004 to $32.7bn at the end of 2006. Angelo Mozilo, who ran the company, wrote in an email later surfaced in litigation that they had no way, with any reasonable certainty, to assess the real risk of holding these loans.

He wrote that while the position was growing sevenfold.

Countrywide had funding that most lenders can only envy. It owned a federally chartered savings bank with deposits and access to government-sponsored borrowing, and it had a standby credit line from forty banks. In August 2007 it announced it would move nearly all its lending into the bank so it could rely less on capital markets, and it drew the entire $11.5bn line at the same time. None of it worked. The company was sold to Bank of America the following year for a fraction of what it had been worth.

A house in California and a taxi license in Chicago have nothing whatsoever in common. The mistake underneath them was the same one. In both cases the loan had stopped depending on what the asset earned and started depending on what somebody would pay for it later, and in both cases the paperwork carried on looking familiar while that happened.

What durable funding can and cannot do

Locking the doors does not fix the room

At this point a sensible reader reaches for an obvious solution. If the danger is money leaving, stop letting it leave.

Behringer Harvard REIT I is the reason that answer is incomplete. It was a property fund that raised roughly $2.9bn from ordinary American investors between 2003 and 2008 and was never listed on any exchange, so there was no market to sell into. In March 2009 it suspended almost all redemptions. In December 2012 it suspended them entirely, including for death and disability, along with the distributions it had been paying.

Investors were locked in about as completely as it is possible to be locked in. Nobody could run, and nobody did.

Shares originally sold at $10 were revalued to $4.25 in 2010, then to $4.01 at the end of 2012, while such trades as occurred in the informal secondary market ran well below that. When the vehicle finally listed in 2015 under a new name, investors received roughly a third of what they had put in, before counting the distributions they had collected along the way. Analysis by Securities Litigation and Consulting Group, a firm that works on non-traded fund disputes and therefore has its own view of them, puts investor principal losses in the region of $1.4bn, and makes the point that American commercial property overall had recovered strongly during the same period.

The lock held for six years and the value went anyway. Preventing a run is not the same thing as preventing a loss. There is a further possibility that our cases suggest without proving, which is that when nobody can force the issue early, a weak position can simply sit there getting bigger for longer.

What history shows

What history actually shows

Two versions of one change, and they need different questions.

Sometimes repayment starts depending on money that has not arrived yet. The building needs another draw. The receivable needs a sale that has not been made. The installation needs its next milestone payment. Lendy and Greensill both sat here. The question that matters is whether the transaction can still be completed if the next money stops, and Assetz answered it by refusing to let its investors leave until existing commitments were seen through.

Sometimes repayment starts depending on a price that has not been realized yet. The taxi medallion, the Californian house. Here the funding question is beside the point. Insured deposits did not save the medallion lenders and a bank charter did not save Countrywide. The question is whether the borrower could repay if the asset never became worth more than it is today.

In every case, the wrapper stayed still while the substance moved. Lendy's development loans paid roughly the same monthly rate as its bridging loans. Greensill's prospective receivables were governed by the same contractual definition as real ones. Medallion lenders held their lending ratios constant while the denominator rose sixfold. The price of the risk did not move, because from the outside the transaction still looked like the transaction it used to be.

Counterexample

The case that limits all of this

None of the above says that moving repayment into the future is a mistake. Mosaic is the reason to be careful.

Mosaic began in 2010 as a solar crowdfunding platform letting small investors put money into commercial solar projects. In 2014 it changed nearly everything at once. It moved to lending directly to homeowners for rooftop solar, over terms of twenty years and more, funded not by its crowd but by warehouse facilities from banks and institutional capital. New borrower, new duration, new collateral, new money behind it.

It worked for eleven years. Mosaic became one of the largest residential solar lenders in the United States, financing billions of dollars of installations. When it eventually filed for bankruptcy in June 2025, the cause was not the transition it had made in 2014. It was that borrowing costs rose, residential solar installations fell sharply, and the margin between what it paid for money and what it charged for it disappeared.

Doing the transition properly bought Mosaic a decade. It did not buy immunity, and the funding structure that carried it through the change was the same one that eventually became too expensive. What we are describing is a signal worth stopping for, not a law that determines outcomes.

One more caution, on the other side. Shopify and Square both appear here as examples of a structure that works, and both have operated and grown for years through a pandemic and a sharp rise in interest rates. Neither publishes the loss rates on these products in a form we could check. Continued operation at scale is real evidence and it is not the same as proven credit performance, and we are not claiming the second.

What follows

What this predicts

If the pattern holds, four things follow.