The billionaire loophole hiding inside the S&P 500

The mechanics are simple enough that a private banker can explain them over a single lunch. A founder holds a large, concentrated, often illiquid-by-optics stake in the company they built. Selling it would mean tax, usually at 20% or more in the US on long-term capital gains, and it would mean signalling to the market that the person who knows the business best is cashing out.

So instead they borrow against it. Banks love this business because the collateral is public, marked to market every second the exchange is open, and the borrower is, by definition, extremely rich. The loan-to-value ratios are conservative, often 10% to 20% of the stock's value, which sounds safe until you remember that the whole arrangement assumes the stock is worth roughly what it was worth last Tuesday.

This is not new in structure. What has changed is scale and repetition. Oracle's share price move this year, driven by a market suddenly convinced that Oracle is an AI infrastructure company rather than a database vendor, has taken Ellison's paper wealth into territory where a fifth of his holding, pledged as collateral, represents tens of billions of dollars in potential exposure sitting on the balance sheets of a small number of banks.

Morgan Stanley and Goldman Sachs are known to compete hard for this kind of lending precisely because it is sticky, profitable, and largely invisible outside SEC filings that most people never read. A margin call on a position that size does not behave like a margin call on a retail trading account. It behaves like a systemic event, because the person facing it can move the very stock being called against just by choosing whether to sell.

The comparison to 2008 is tempting and wrong in the way that matters. Mortgage-backed shadow banking multiplied risk through slicing and reselling, so that a bad loan in Ohio ended up inside a pension fund in Norway, and nobody could trace the exposure. Founder-collateral lending does the opposite: it concentrates risk into single, identifiable names, at single, identifiable banks, against single, identifiable stocks.

That sounds safer. It is not, because concentration means correlation. If Oracle's valuation is being carried by a market-wide AI re-rating, and that re-rating cracks, Ellison's collateral falls at the same moment as the exposure his lenders are holding, at the same moment as every other AI-adjacent founder pledge in the system. The 2008 crisis spread risk thin enough that it took months to find where it lived.

This version would tell you exactly where it lives on day one, and give you no time to do anything about it. , - The reason this has scaled now, rather than a decade ago, is the same reason retail bond investors are suddenly paying attention to yields again. Borrowing against equity only makes sense as an alternative to selling when the cost of that borrowing is manageable relative to the appreciation of the collateral.

For years it was a trivially good trade, banks lent near-free money against stock that kept compounding at 15 to 20% a year, and founders got liquidity without dilution or tax. The environment has shifted. Ten-year Treasury yields have pushed past 5% for the first time since 2007, the highest in almost two decades, and that repricing has rippled into the cost of every form of borrowing that references it, share-backed lending included.

Banks are not walking away from these facilities, the fees are too good, but the spreads are widening and the covenants are tightening, quietly, in ways that rarely make it into a press release. That matters more than most people writing about bond yields this year have bothered to note. Nobody has been asking what happens to founder lending books when the risk-free rate stops being close to zero.

A private banker structuring a $2bn facility against a single stock in 2021 was pricing in an environment where volatility was low and stayed low. That world is gone. Treasury volatility just posted its sharpest jump in more than a year, and the very shorted, high-beta stocks that used to be dismissed as market froth are having one of their best years on record, which tells you the correlation between confidence and reality has come unstuck at exactly the moment collateral quality starts to matter again. , - The people who win from this arrangement are, obviously, the people doing it.

Ellison gets liquidity without triggering a taxable event and without ceding a share of control at Oracle's board table. Musk got to buy Twitter without becoming a forced seller of Tesla stock into a falling market, which would have compounded the very problem the sale was meant to solve. Politically ambitious billionaires get to fund campaigns, foundations, and pet causes off paper wealth that the IRS cannot touch until it is realised, if it ever is.

Wealth advisers describe this candidly to clients rather than dressing it up. One senior banker interviewed for a Bloomberg piece on the great wealth transfer put it plainly: the job now is building enough trust with a client that they let you structure around the tax event entirely, rather than simply manage what is left after one. That is the entire pitch.

Not investment performance. Tax and liquidity engineering, delivered through a loan document instead of a trade. The losers are harder to name because the arrangement is designed so there are none, until there are. Regulators do not currently require aggregated disclosure of share-pledge lending in the way they require disclosure of, say, systemically important bank exposures.

The SEC gets individual filings when a founder pledges shares, but no regulator anywhere is adding those filings up across founders, across banks, across sectors, to ask what the concentrated total looks like in a falling market. The Bank of England's Financial Policy Committee stress-tests UK banks against property crashes and rate shocks.

Nobody stress-tests a scenario where five AI-adjacent founders get simultaneous margin calls because the sector that made their paper wealth also made their collateral melt at the same time. That asymmetry, obvious once stated, is the entire argument. A shadow banking system does not need slicing, tranching or acronyms to be dangerous. It only needs collateral that is correlated, lenders who are concentrated, and a regulator who has not yet noticed the categories being used to describe it are twenty years out of date.

Founder share pledges tick all three boxes and currently sit almost entirely outside systemic risk frameworks built for mortgages and interbank lending, not billionaire liquidity management. , - What would prove this wrong is a market that keeps rising indefinitely, in which case none of this ever gets tested and the whole apparatus remains a clever tax and liquidity trick with no downside anyone need worry about.

That is not a fringe possibility, equities have shrugged off yield spikes before and JPMorgan's own strategists are on record arguing stocks can keep climbing even as the bar for earnings rises with borrowing costs. But betting the stability of a sliver of the credit system on stocks never falling is precisely the kind of assumption that regulators are paid to distrust, and precisely the one currently going unexamined.

The founders themselves are rational actors responding to a tax code and a lending market that reward exactly this behaviour, and there is no reason to expect any of them to stop voluntarily. The question worth watching is not whether Ellison or Musk get margin called this year. It is whether any regulator, in London or Washington, starts asking banks to report aggregate share-pledge exposure the way they already report mortgage books and leveraged loan exposure.

Until that happens, the collateral sitting behind a meaningful slice of the world's largest personal fortunes will keep being priced, quietly, as if the market only moves in one direction.