The Leveraged ETF Crash Insurance No One Wants to Own

The Bloomberg headline today says banks are offloading risk from leveraged ETFs with exotic "crash puts". That's technically true, in the same way a casino can say it has offloaded risk when it sells its insurance policy to someone else. The risk didn't disappear. It just found a new landlord.

The products promise two or three times the daily return of an underlying stock, and retail investors have lined up to buy them - particularly the ones tied to AI-linked semiconductors like SK Hynix and Samsung. But the investor holding the ETF is not the counterparty to this trade. As one Bloomberg Opinion piece put it, the retail buyer is the raw material.

The mechanism behind the product runs through banks. To deliver that 2x or 3x daily return, ETF issuers enter total return swaps with dealer banks - Barclays, Citigroup, Goldman Sachs, and Bank of America are among the most active. The bank promises the leveraged return, and the ETF pays a financing spread above SOFR. In exchange, the bank has to hedge its exposure, typically by trading the underlying stock, futures, or options. The swap is marked to market daily.

Here's where the plumbing gets interesting. The bank has a structural problem called gap risk. If a stock drops more than 50% in a single session, a 2x ETF can lose more than its total net assets. The ETF's collateral - usually cash or Treasuries - isn't enough to cover the bank's loss. That's the scenario the bank can't hedge with normal stock or futures hedging, because by the time the market opens and the dealer can act, the damage is already done.

So the bank buys insurance. The product of choice is what's known as a crash put, or cliquet - a chain of deeply out-of-the-money put options that reset to the current price at regular intervals. Each "leg" of the cliquet pays out if the stock falls below a certain threshold over the period. If the threshold is met, the next leg resets and the protection continues. If the stock never crashes, the buyer of the crash put loses the entire premium.

The cost of this insurance has more than tripled for the hottest names over the past three months, as one Bloomberg Opinion piece noted. Goldman Sachs pitched a trade in May called "Expensive Crash Cliquet" - literally naming the product for what it is. The yields offered to investors willing to take the other side, selling this tail protection and using leverage to amplify returns, ranged from 14.2% to 20%.

Now let's follow that risk transfer, because this is where the mainstream narrative gets thin. Who's buying Goldman Sachs' expensive crash cliquets? Hedge funds. Yield-seeking asset managers. Carry-hungry institutions that are now collectively short a book of correlated AI gap risk for up to one year.

Natasha Sibley at Janus Henderson, whose diversified alternatives team is in the trade, told Bloomberg: "I have never seen this level of demand in this product." Ramon Verastegui at Kairos Investment Advisors called them "very efficient, back-to-back risk-transfer tools" and noted that banks are actively trying to develop the crash-put market to hedge their leveraged ETF book.

Understanding what I understand about spreads and economics, the question isn't whether the bank has hedged its gap risk. It has. The question is whether the market has absorbed enough of this concentrated tail exposure to remain stable when the event actually occurs. These cliquets are over-the-counter products. There's no central clearinghouse, no transparent pricing, no way to know how much of this risk is concentrated in a handful of institutions. The over-the-counter nature makes it impossible to quantify their growth.

And the events are not purely hypothetical. On July 14, Lucid Group sank as much as 57% intraday before recovering some of the loss. The GraniteShares 2x Long LCID Daily ETF lost approximately 92% of its value and was shut down a day later - the first termination of a U.S.-listed single-stock leveraged ETF. That 57% intraday move is exactly the kind of gap the cliquet is designed to cover. The worst session on record for SK Hynix, one of the biggest leveraged ETF names, was a 15.4% drop on July 13. South Korea's 30% daily price limit means a single session crash of 50% isn't possible there - but a multi-day limit-down move still triggers the cliquet because it calculates on a close-to-close basis.

Seventy-three leveraged and inverse ETFs have already closed in 2026, according to Morningstar data through July 23. That's more than three times the pace of recent years. And the closure rate is accelerating, not decelerating.

So here's the chain: Retail investors buy leveraged ETFs for amplified upside. Banks provide the leverage through total return swaps. Banks hedge their gap risk by buying crash puts. Hedge funds and yield-hunters sell those crash puts for 14% to 20% in premium income. If nothing crashes, everyone collects a check. If something does crash - a 50% single-day move in an AI-linked stock - the hedge fund selling the protection doesn't just lose its premium. It gets hit with a convex payout it likely hasn't fully hedged, because these are deeply out-of-the-money strikes that no one expects to print.

This is the same structural pattern we saw with the Treasury basis trade. Bloomberg's Simon White documented hedge fund exposure to banks at roughly $4.5 trillion, up from about $2 trillion just a few years earlier. The average gross exposure of U.S. hedge funds has almost doubled since 2022. The leverage train has been expanding in both directions - retail chasing leveraged upside through ETFs, institutions selling tail protection to finance it. There's nowhere to hide, as White put it.

The SPY put-to-call volume ratio sits at 1.28 today, and put-to-call open interest is at 1.95. Implied volatility is 13.15%. The market is pricing in a calm summer. VIX is low. Complacency is high. That's exactly the regime where crash put premiums are juiciest - and exactly the regime that tends to end abruptly.

The conditional chain is straightforward. If AI-linked stocks keep trending higher and the leveraged ETF book keeps expanding, cliquet costs keep rising, more hedge funds sell protection, and the carry keeps flowing. But if one of these names gapes down - and they're volatile enough that this is a matter of when, not whether - the institution on the wrong side of the cliquet faces a loss that's not linear. It's convex. And if multiple institutions are exposed to correlated AI name gap risk simultaneously, you don't get one blow-up. You get a cascade.

Banks didn't offload the risk. They outsourced it. And the people wearing it right now are collecting 20% yields and calling it a good trade.

The views expressed here are my own and do not constitute investment advice.