Muddy Waters Capital
Carson Block (founder, CIO)
Famed activist short seller expressing the AI-bubble risk indirectly — rather than shorting big-tech equities directly (which Block calls a 'widowmaker'), it bets on AI-driven job-loss feeding through to corporate credit, via shorts on investment-grade and high-yield credit ETFs.
- Not open — Not an open-ended vehicle — analysis only, not an available allocation.
Short positions
Apr 2026: bets against HYG as an AI-meltdown / credit-stress hedge
Apr 2026: paired credit-ETF short on AI-driven job-loss / labor-market fears
Recent moves
As of Nov 2025 Block said it was 'not the time' to short big tech directly; by Apr 2026 he was hedging an AI meltdown through credit-ETF shorts (HYG, LQD) on AI-job-loss fears, and was rethinking a planned India long-short fund 'in light of our house view on AI' and its labor-market impact.
Our take
Included as an iconic short with a real, sourced AI-relevant thesis — but note the framing: Muddy Waters' AI bet is a credit trade, not an equity short of Nvidia/Palantir. That's actually the more sophisticated expression (if AI capex/job-loss cracks, it shows up in spreads before it shows up cleanly in mega-cap equity), and it sidesteps the 'widowmaker' problem of shorting names that keep ripping.
Block quantifies the macro thesis: AI displacing roughly 15% of US knowledge jobs, about 15 million positions across tech, law and finance, with lost 401(k) contributions as the transmission channel into equities.
That is falsifiable; the expression is where we are skeptical. Carry: shorting HYG or LQD means paying coupon plus borrow every day the thesis has not arrived, on a view with no arrival date you control — unlike an activist short, where your own report is the catalyst. Instrument mismatch: LQD is long-duration IG, so Treasury yields dominate its price, not spreads. The recession Block describes would plausibly widen spreads AND collapse yields, and the rates leg could offset or swamp the spread leg. CDX, single-name CDS or payer swaptions express 'spreads widen' cleanly; LQD does not. HYG is the better leg (shorter duration, spread-dominated) but is also the crowded consensus hedge. In fairness to Block, he links the two books — calling SoFi 'probably ground zero' for an AI hit to its core borrowers — so the separation we draw is ours, not his.
Thesis
Muddy Waters runs two businesses that look like one: a forensic single-name activist-short shop — Sino-Forest and NQ Mobile vindicated, Luckin amplified from an anonymously-sourced report later admitted as fraud, AppLovin and SoFi still open and contested — and a top-down macro credit hedge shorting HYG and LQD on the view that AI-driven job losses widen spreads. The first is a capacity-constrained skill; the second is a directional macro opinion expressed in instruments the firm has no research edge in. Judging the fund means separating them.
Concentrated activist short selling: forensic accounting and channel work on a small number of names, published as public reports that are themselves the catalyst. Long/short mandate across equities and credit, globally. Layered on since 2025-26: a macro overlay shorting investment-grade (LQD) and high-yield (HYG) credit ETFs as an AI-disruption hedge, plus reported bullishness on mining and a rethought India long-short vehicle.
Assessment
- Documented forensic edge: Sino-Forest and NQ Mobile were vindicated by bankruptcy and resignations, not just price declines (Luckin was a third-party report MW distributed).
- Self-catalysing model — the published report is the trigger, so the fund isn't dependent on the market independently discovering the fraud on its own timeline.
- Refusing to short Nvidia/mega-cap AI directly is the right call: shorting a momentum leader with unlimited loss and index inflows is the classic widowmaker.
- Capacity discipline — reported AUM in the hundreds of millions (figure disputed, see unavailable) suits a strategy whose alpha comes from concentrated small/mid-cap campaigns.
- The AI-credit macro trade sits far outside the firm's demonstrated circle of competence — forensic accounting on one company is not the skill of forecasting economy-wide credit spreads.
- Shorting LQD is largely a Treasury-duration short, not a spread short. In an AI-driven recession, falling yields could offset the spread widening the thesis depends on.
- Negative carry with no fixed catalyst: you pay HYG/LQD coupon plus borrow indefinitely while waiting for a macro regime change you cannot trigger yourself.
- Extreme key-person concentration — Block personally is the research brand, the media distribution and the campaign credibility. No visible succession.
- Recent campaigns are contested on the accounting but no longer on the tape: AppLovin rose for a year then broke (−23.2% in Jul 2026, ~−46% off its Dec-2025 high) and SoFi is −37.8% YTD. Unrealised, and the claims remain disputed.
Record
Returns are not publicly disclosed. The most-cited figure — roughly 19% annualised net of a 2.5% management and 30% performance fee — comes from Institutional Investor reporting circa 2020-21 covering the fund's first ~5 years from its 2016 launch. That is a stale, self-selected vintage: it captures a window in which several campaigns (Luckin) resolved spectacularly, and says nothing about 2021-2026. The widely repeated 2020 annual sub-figure could not be re-verified at check time, so we drop it. No audited or independently verified track record is available. Attribution matters as much as level: an activist-short book's returns are lumpy and campaign-driven — a handful of resolutions carry the compounding — so a single-digit number of wins likely explains most of the record, which makes statistical inference about 'skill' weak even where the individual calls were clearly correct.
Risks & fit
- Short-squeeze and unlimited-loss exposure on concentrated single-name campaigns; a wrong or early call compounds against the book.
- Litigation risk is structural: a German market-manipulation inquiry over the 2016 Ströer report; SoFi said in March 2026 it would explore legal action over MW's report.
- Key-person: no evident continuity if Block steps back; the strategy is inseparable from his personal credibility.
- Macro overlay could bleed carry for years if AI capex sustains growth and spreads stay tight, dragging on an otherwise idiosyncratic book.
- Crowding — HYG/LQD shorts are the consensus recession hedge, so the payoff compresses if the scenario is widely anticipated.
The credit thesis dies if, over 18-24 months, AI adoption lifts corporate margins without aggregate layoffs feeding through to loan delinquencies, and IG/HY option-adjusted spreads stay at or below post-2021 lows while HYG/LQD holders collect coupon — making the short a pure carry bleed. The single-name franchise is falsified differently: a run of campaigns where the target rebuts the report and the stock makes new highs a year later (the AppLovin shape, and the open question on SoFi) would signal the forensic edge has decayed as disclosure improves and rival shorts crowd the same screens.
Analytically interesting to anyone studying how activist short selling actually generates returns, whether a research edge transfers to macro expression, and how negative-carry hedges behave when the catalyst has no date. Note this is a private fund whose returns are undisclosed and whose positioning is known from interviews and press rather than filings — a case study in strategy design, not an accessible or externally evaluable product.
Reported at 2.5% management and 30% performance — above standard 2-and-20. Defensible only if the alpha is genuinely idiosyncratic; the same load on a directional credit-ETF short is an expensive wrapper on a two-click view.