Insurance on the Future

In March 2026, JPMorgan offered its clients a new product. It was not a fund, or a bond, or a way to invest in the artificial-intelligence boom. It was a way to bet against it — a custom basket of credit-default swaps on the debt of five of the largest hyperscalers in the world: Alphabet, Amazon, Meta, Microsoft and Oracle, traded in $25 million increments, equally weighted across the five. For a fee, an investor could now buy, in a single liquid instrument, insurance that pays out if the companies leading the AI revolution struggle to repay the hundreds of billions of dollars they are borrowing to build it.

Sit with the symmetry of that for a moment. JPMorgan is one of the principal banks underwriting the very debt the basket is designed to protect against. It helps the hyperscalers sell their bonds to investors, and now it helps those same investors bet that the bonds will sour. The bank is, in effect, selling both the cruise tickets and the lifeboats — and the brisk demand for lifeboats is the story.

Because the JPMorgan basket is not an isolated curiosity. It is the most visible symptom of a market-wide scramble that has been building quietly beneath the euphoria of the AI equity trade: the rapid, explosive construction of the machinery to short the AI boom's debt. The stock market is pricing the dream. The credit-derivatives market is pricing, and selling, the nightmare — and the volume tells you which one the smart money is paying more attention to.

The tenfold surge

The numbers describing this build-out are the kind that make credit veterans uneasy, because they have seen the shape before.

At Bank of America, the monthly notional volume of credit-default-swap trading on the hyperscalers is up roughly tenfold since the start of 2025. Industry data from the DTCC shows that single-name CDS volumes on a small cluster of US technology groups have jumped about 90% since early September alone. Most striking of all, single-name credit-default swaps on these tech giants have gone from near-zero a year ago to among the most actively traded contracts in the entire derivatives market. An instrument that essentially did not exist for these companies twelve months ago is now one of the busiest corners of credit. That is not the signature of complacency. It is the signature of a large and growing number of sophisticated investors quietly buying protection.

The prices of that protection have moved too. The cost of insuring Oracle's debt has widened dramatically — its five-year CDS blowing out by hundreds of basis points over the past year to record levels. CoreWeave, the debt-laden AI cloud company, trades at credit-default-swap spreads of around 640 basis points, the pricing of a company the market treats as genuinely speculative. A brand-new CDS market materialized for Meta specifically after it sold roughly $30 billion of bonds tied to a single AI data-center project. Wherever the AI debt has flowed — and it has flowed to the tune of more than $120 billion of hyperscaler issuance in the past year, with $250–300 billion more expected in 2026 — a market to bet against it has sprung up directly behind it, like a shadow.

The 2008 rhyme

Anyone who lived through the financial crisis will feel a cold flicker of recognition here, because the sequence is not merely similar to 2007–08. It is structurally identical.

In the years before the mortgage crisis, the thing that turned a housing downturn into a global catastrophe was not the bad loans themselves — it was the vast superstructure of derivatives built on top of them, and specifically the creation of liquid, standardized ways to bet against mortgage debt. The ABX index, the synthetic CDOs, the single-name and basket credit-default swaps on mortgage exposure: these instruments are what allowed the crisis to propagate, to leverage, and to inflict losses many times larger than the underlying mortgages themselves. They also, crucially, were the mechanism by which the few who saw it coming — the Michael Burrys of The Big Short — were able to position for the collapse. The existence of a deep market to short something is simultaneously a sign of financial maturity and the precondition for a cascade.

That exact superstructure is now being assembled, in plain sight, on top of AI debt. The JPMorgan five-name basket is, functionally, an ABX index for the hyperscalers — a standardized, liquid way to express the trade "AI capex disappoints and the debt struggles." The tenfold surge in CDS volume is the market gaining the depth required for that trade to be put on at scale. None of this means a crisis is imminent or inevitable; most of this debt is investment-grade, the hyperscalers are among the most profitable companies on earth, and a deep CDS market can simply be prudent risk management rather than a doom signal. But it is worth being precise about what has been built, because the same people now reassuring equity investors that the AI boom is sound are, with their other hand, manufacturing and selling the instruments that would let the boom's failure spread fast and far.

What the banks can smell

The most revealing detail is not the volume or the prices. It is who is driving the activity, and why.

According to the reporting, Wall Street banks are trading more of these derivatives in part because they have to in order to keep doing business with the hyperscalers at all — the sheer scale of the debt means dealers must hedge their own exposure, and clients demand liquid ways to manage theirs. In other words, the surge is not coming from fringe doomsayers. It is coming from the core of the financial system: the banks underwriting the bonds, the institutions holding them, the dealers intermediating them. These are the participants with the most complete view of the AI build-out's financing — the term sheets, the covenants, the cash-flow projections, the customer concentrations — and what they are doing, collectively, is paying up for insurance and building ever-more-liquid ways to short the thing they are simultaneously selling.

This is the credit market's version of the tell that runs through every story of this era — the CoreWeave founders selling their stock, the Strategy preferred forcing a bitcoin sale, the Palantir insiders cashing out, the Oracle bondholders suing, DoubleLine's Robert Cohen putting the odds of an AI debt bubble at 100%. In each case, the people closest to the asset, with the best information and the most at stake, are quietly positioning for disappointment while the public is urged to buy the dream. The credit-derivatives boom is the same signal written in the most sophisticated instruments the market has: a tenfold increase, in a single year, in the demand to be protected against the failure of the companies the equity market is pricing for perfection.

None of the individual pieces is alarming on its own. Hedging is healthy; liquid markets are good; investment-grade debt usually gets repaid; and the AI revolution may well prove to be exactly the world-changing, cash-generating force its champions promise, in which case all this insurance expires worthless and the worriers look foolish, as worriers often do at the top and the bottom alike. But step back and look at the aggregate, because the aggregate is the message. An enormous wave of debt has financed the AI boom. A market to bet against that debt has gone from nothing to one of the busiest in finance in twelve months. The bank arranging the bonds will now sell you a basket to short five of the issuers at once. And the demand for that basket — for insurance on the future everyone else is buying stock to celebrate — is exploding.

When the same hands that build the boom start, so eagerly, to build the hedges against it, they are telling you something they will never say out loud. The equity market is selling you the future. The credit market is quietly, busily, profitably insuring against it. One of them is going to be wrong, and the one writing the insurance policies usually reads the fine print first.