DEAD RECKONING (1)
A three-part dispatch from someone who has seen this film before
The Man Who Knows
Author’s note: This piece was completed before Jamie Dimon’s annual shareholder letter was published on 6 April 2026, in which he warned that private credit losses will be “higher than expected,” that the market lacks “great transparency or rigorous valuation ‘marks’ of their loans,” and that some credit providers will prove to have done “a far worse job than others.” In the preceeding days, Blue Owl Capital gated two funds after receiving $5.4 billion in redemption requests, KKR restricted withdrawals from its retail BDC, and the Congressional Research Service briefed Congress on private credit redemption risks. The argument below was written without knowledge of these developments. The reader may draw their own conclusions from the convergence.
I was working in equity capital markets in the City of London in the years before the 2008 financial crisis. In 2006 and into 2007, I wrote about what I was seeing; a system under increasing internal stress that official commentary was either unable or unwilling to acknowledge. What I observed then was not primarily about bad mortgages or reckless borrowers. It was about something more instructive and more sinister. It was about who knew what, and when.
During those years I was reading a blog called Calculated Risk, written by Bill McBride. His co-author was Doris Dungey, a mortgage industry insider with twenty years of experience who wrote under the pseudonym Tanta. She had processed loan files. She understood, from direct professional knowledge, what the documentation said and what it concealed. She wrote with a clarity and a scepticism that institutional commentary could not match and she remains for me the benchmark for what serious, evidence-based financial warning looks like. She died of ovarian cancer in November 2008, aged 47. I will return to her in Parts Two and Three.
History has since confirmed the sequence. The firms with the deepest knowledge of the underlying credit market began positioning against it, quietly, while continuing to distribute the same product to clients. Goldman Sachs, according to documents later reviewed by the United States Senate Permanent Subcommittee on Investigations, made a firm-wide decision in December 2006 to take a negative stance on the US mortgage market. Over the following months, Goldman packaged, sold, and simultaneously shorted 47 synthetic CDOs with an aggregate face value of $66 billion. The most notorious of these was Abacus 2007-AC1, completed in April 2007, in which Goldman helped a hedge fund select mortgage bonds the fund believed most likely to fail, then sold the long side of the exposure to clients including IKB Deutsche Industriebank, a German lender that lost approximately $150 million on the transaction. Goldman eventually settled with the SEC in 2010 for $550 million, acknowledging that its marketing materials had contained “incomplete information.”
Bear Stearns, meanwhile, was running two in-house hedge funds with leverage ratios ranging from roughly ten to one up to twenty-five to one, against a portfolio of mortgage-backed CDOs. The funds were marketed to investors as conservative, income-oriented vehicles. When they gated redemptions in June 2007, the combined investor losses eventually reached $1.8 billion. Bear Stearns as a firm collapsed nine months later. JPMorgan Chase initially agreed to acquire the firm for $2 per share, a figure later raised to $10 following shareholder revolt, against a 52-week high of $133.
I am not recounting this history for nostalgia. I am recounting it because in the first quarter of 2026, I watched a sequence of events that caused a familiar feeling to return.
* * *
The Signal
In early March 2026, JPMorgan Chase quietly reduced the collateral value it was prepared to assign to software loans held by private credit funds. Insiders described the move to the financial press as preemptive, driven by market valuations rather than actual loan losses. The bank was, in the language of risk management, getting ahead of the problem.
Three weeks later, JPMorgan launched a basket of credit default swaps covering five of the world’s largest technology companies: Alphabet, Amazon, Meta, Microsoft, and Oracle. The instrument trades in $25 million increments, with $5 million allocated to each company’s CDS contract. It allows institutional investors to take either long or short positions on the creditworthiness of these firms.
JPMorgan is one of the principal underwriters of bond issuance for these same companies, and the connection matters. The five hyperscalers, as they are known, issued $121 billion in bonds in 2025 alone, against an annual average of $28 billion between 2020 and 2024. JPMorgan has been at or near the front of the majority of those transactions. It now offers clients a structured instrument to express a view on the creditworthiness of the very companies it has been helping to borrow.
Goldman Sachs is separately offering hedge funds total return swaps allowing them to take positions against corporate loan portfolios without holding the underlying debt. Citadel Securities began facilitating similar bond baskets in November 2025.
None of this is illegal. None of it is necessarily improper. Markets require two sides, and hedging instruments serve legitimate purposes. But the sequencing matters, and the sequencing here is striking. The firm that originates and distributes marks the collateral down, then creates the hedging instrument. I have seen this film before.
* * *
The Plumbing Beneath the Signal
The JPMorgan CDS basket has attracted attention as a technology story, a new derivatives product for the AI era. Focusing on the hyperscaler basket alone, however, misreads where the structural risk actually sits.
The more consequential development is what JPMorgan’s collateral markdown reveals about the private credit market, and the chain of vehicles through which that market’s risk has been distributed.
Private credit has grown from roughly $1 trillion a decade ago to a global market estimated at $3.5 trillion in assets under management, according to the Alternative Credit Council. Estimates vary by source and definition: Morgan Stanley places the US market alone at approximately $3 trillion and projects it will reach $5 trillion by 2029, while Preqin estimates global direct lending assets at approximately $979 billion as of mid-2025. The variation reflects differing definitions of what constitutes private credit. This expansion occurred primarily because post-GFC banking regulation, principally Basel III, restricted banks’ ability to lend to middle-market and leveraged borrowers. Private credit funds, which are not subject to the same capital requirements, filled that gap. The arrangement suited everyone: borrowers got capital, investors got yield, and banks got fees from back-leverage arrangements that put their balance sheets to work without the regulatory overhead.
The numbers associated with those back-leverage arrangements deserve attention. According to figures cited by Moody’s, Wall Street banks had provided approximately $300 billion in financing to private credit funds by mid-2025, a figure that JPMorgan itself disclosed had tripled from roughly $50 billion in 2018 to $160 billion in 2025 for its own book alone. These are secured lending facilities, collateralised against the private credit funds’ loan portfolios. When JPMorgan marks down the value of software loans as collateral, the mathematical consequence runs through the entire chain.
Software companies account for roughly 25 per cent of all private credit loans. The majority of those loans were originated in 2021 and 2022, at peak valuations and aggressive terms, to finance private equity buyouts of software businesses. JPMorgan estimated in early 2026 that between $40 billion and $150 billion in CLO holdings could face repricing pressure from AI-driven disruption to software business models. The range of that estimate is itself informative: it speaks to the opacity of the underlying exposure.
* * *
The Retail Dimension
The pre-GFC story has one element that rarely receives sufficient emphasis: how far down the risk distribution chain the exposure eventually reached. It was not investment banks alone that lost money. It was municipal pension funds in Washington State, German regional lenders, and insurance companies in the Midwest, all of whom had been sold instruments they could not fully evaluate, rated by agencies whose models proved catastrophically wrong.
In the current episode, the distribution chain runs through a class of vehicle called the non-traded Business Development Company, or BDC. These are closed-end funds that have grown from essentially zero in 2021 to more than $200 billion in assets, marketed primarily to high-net-worth individuals and retail investors seeking yield above what conventional fixed income provides. They typically offer quarterly redemption windows with gate provisions, which limit how much capital can be withdrawn in any single period.
The redemption data for Q1 2026, as shown in Financial Times charts shared by Mohamed El-Erian, makes uncomfortable reading. Apollo’s non-traded BDC recorded redemption requests equivalent to 11.2 per cent of net asset value. Morgan Stanley’s reached 10.9 per cent. Several funds show unmet redemptions; requests that exceeded the gate threshold and could not be honoured in full. This is not yet a crisis. But it rhymes with a pattern that serious analysts will recognise: retail and high-net-worth investors moving towards the exit at roughly the point that institutional actors are reducing their own exposure.
* * *
What This Is Not
I want to be clear about what I am and am not arguing.
Bank balance sheets today are materially stronger than they were in 2007. Post-GFC capital requirements, stress testing, and living will frameworks mean that the major banks are better equipped to absorb losses. The Federal Reserve’s 2025 stress tests found that under severe recession scenarios, large banks remained above minimum capital requirements even accounting for NBFI exposures. The Bank of England’s December 2025 Financial Stability Report confirms that UK banks are well capitalised.
The hyperscalers themselves are not subprime borrowers. Alphabet, Amazon, and Meta generate extraordinary operating cash flows. Their debt is investment grade for good reason.
The crisis of 2008 was not solely a banking crisis. It was a crisis of opacity, leverage, and misallocated trust: the systematic packaging and distribution of risk to parties who did not fully understand what they were holding. The risk today has migrated away from bank balance sheets and into a private credit market that, in the words of the Bank of England’s own Financial Policy Committee in December 2025, “has not been tested through a broad-based macroeconomic stress at its current size.”
That sentence, from one of the world’s most cautious institutional voices, is the one worth dwelling on.
The question is not whether today is 2007. It is whether we are again watching the people who understand the system most clearly begin, quietly, to hedge against it. And if we are, what that tells us about what they think comes next.
* * *
References: Part One
1. United States Senate Permanent Subcommittee on Investigations, “Wall Street and the Financial Crisis: Anatomy of a Financial Collapse,” April 2011. Goldman Sachs case study, including December 2006 firm-wide decision, 47 synthetic CDOs, $66 billion aggregate face value.
2. Securities and Exchange Commission, Litigation Release No. 21592, “Goldman, Sachs & Co. and Fabrice Tourre,” 15 July 2010. $550 million settlement, Abacus 2007-AC1.
3. Securities and Exchange Commission, Litigation Release No. 20625, “Ralph R. Cioffi and Matthew M. Tannin,” 19 June 2008. Bear Stearns hedge fund losses of approximately $1.8 billion.
4. Bloomberg, “JPMorgan Offers Clients a New CDS Basket to Hedge AI Debt Risk,” 23 March 2026. CDS basket structure, $25 million increments, five hyperscalers.
5. Mellon Investments Corporation, “Record-Breaking AI-Related Debt Issuance in 2025,” December 2025. $121 billion bond issuance, $28 billion annual average 2020-2024.
6. Alternative Credit Council, “Financing the Economy 2024.” Global private credit AUM estimate.
7. Moody’s, cited in multiple financial press reports, mid-2025. $300 billion bank financing to private credit funds.
8. Financial Times, charts shared by Mohamed El-Erian, Q1 2026. BDC redemption data including Apollo (11.2% of NAV) and Morgan Stanley (10.9%).
9. Bank of England, Financial Stability Report, December 2025. Financial Policy Committee language on private markets not being tested at current size.
10. Federal Reserve, 2025 stress test results. Bank capital adequacy under severe scenarios.
11. Doris Dungey (Tanta), Calculated Risk blog, December 2006 to September 2008. See also obituary, The New York Times, 1 December 2008; The Washington Post, 4 December 2008. Federal Reserve, “Understanding the Securitization of Subprime Mortgage Credit,” 2008, citing Dungey’s work. ---
This document is analytical commentary. It does not constitute financial advice, a financial promotion, or a recommendation to buy, sell, or hold any security, financial instrument, or investment product. The author is not a regulated financial adviser. The author holds no positions, long or short, in any of the securities, instruments, or companies discussed.