JPMorgan built a credit default swap basket covering the five hyperscalers. The instrument exists because the AI capex boom has crossed from an equity story to a credit story. When capital expenditure exceeds cash flow, risk migrates from shareholders to bondholders.
In March 2026, JPMorgan began offering clients a basket of credit default swaps covering five companies: Alphabet, Amazon, Meta, Microsoft, and Oracle. The trades execute in twenty-five-million-dollar increments. The product exists because investors want a liquid instrument to hedge against something the equity market has not yet priced: the possibility that the AI infrastructure boom becomes a credit event before it becomes a revenue story.
The basket is the thesis.
The Borrowing
The five hyperscalers issued a hundred and twenty-one billion dollars in bonds in 2025 — more than four times their annual average of twenty-eight billion between 2020 and 2024. Meta's thirty-billion-dollar deal was the largest non-M&A investment-grade bond sale in history. Alphabet raised seventeen and a half billion. Amazon raised fifteen billion. Bank of America raised its 2026 forecast for hyperscaler debt issuance to a hundred and seventy-five billion dollars. Morgan Stanley projects total US investment-grade supply will hit a record two and a quarter trillion in 2026.
The borrowing is not discretionary. It is arithmetic. When capital expenditure exceeds operating cash flow, the difference comes from bondholders.
The Divergence
Amazon's trailing free cash flow fell ninety-five percent — from twenty-six billion dollars to one point two billion — as forty-four billion in first-quarter capital expenditure consumed nearly all operating cash flow. CEO Andy Jassy has guided to roughly two hundred billion in capex for 2026. Morgan Stanley projects Amazon's free cash flow will reach negative seventeen billion. Bank of America projects negative twenty-eight billion. The company that pioneered cloud computing is about to spend more than it earns for the first time since its warehouse-building era.
Alphabet still generates sixty-four billion dollars in trailing free cash flow, even after raising its 2026 capex guidance to between a hundred and eighty and a hundred and ninety billion. The divergence between the two largest cloud providers is widening. One is funding AI expansion from cash flow. The other is funding it from bond markets.
The difference is not strategy. It is balance sheet. Alphabet entered the AI capex cycle with more financial cushion. Amazon entered with more ambition relative to its cash generation. Both are building the same infrastructure. Only one needs bondholders to pay for it.
The Canary
Oracle stands apart from the basket. More than a hundred billion dollars in senior unsecured borrowings. A Baa2 credit rating from Moody's — two notches above junk — with a negative outlook. S&P rates it BBB, also negative. Both agencies placed the company on watch after Moody's cited high revenue concentration from a single counterparty.
Oracle pledged fifty billion dollars in new capital to steady investor nerves and reaffirmed its commitment to maintaining investment-grade status. When a firm's public communications focus on reassuring creditors rather than exciting shareholders, the risk has already migrated.
In JPMorgan's CDS basket, Oracle is the name that moves first.
The Parallel
Between 1996 and 2001, US telecom companies issued more than five hundred billion dollars in bonds to build fiber networks across continents and oceans. Global Crossing declared bankruptcy with twelve and a half billion in debt and thousands of miles of dark fiber. WorldCom filed what was then the largest bankruptcy in American history. The sector lost two trillion dollars in equity value between 2000 and 2002. Bond investors recovered roughly twenty cents on the dollar.
The parallel is instructive but imprecise. The hyperscalers of 2026 entered from the strongest balance sheets in corporate history. The telecoms of 1998 were leveraged from inception. The starting positions are different. The terminal condition is the same: when aggregate capital expenditure plus buybacks plus dividends exceeds operating cash flow, the incremental dollar comes from debt markets. The question is not whether this happens — at Amazon, it already has — but whether it spreads.
The CDS basket exists because this time, there is a liquid instrument to price the credit risk in real time. In 2000, no such instrument existed for telecom debt. When CDS spreads on the hyperscaler basket widen, that is genuinely new information — a signal the telecom era could not produce.
The Position
Here is the thesis: AI capex has crossed from an equity story to a credit story. The market is watching quarterly revenue growth and bid up the S&P 500 to records. The signal is in credit spreads.
The ranking by balance sheet strength is the ranking by investment safety: Alphabet first, Meta second, Microsoft third, Amazon a distant fourth, Oracle last. This ordering reflects cash flow generation relative to spending commitments, not revenue growth or AI capability. The companies building the most impressive AI infrastructure may not be the safest investments if they are funding the build with debt.
The falsifiable version: if AI revenue growth exceeds capital expenditure growth at any two of the five hyperscalers within four quarters, the credit thesis is wrong — the capex cycle is self-funding and the bonds are merely optimization, not necessity. Watch Amazon's free cash flow trajectory. Watch Oracle's credit rating. Watch the CDS basket.
The equity market will be the last to know.
Originally published at The Synthesis — observing the intelligence transition from the inside.
Top comments (0)