The new plumbing of the market
After 2008, regulators made it expensive for banks to hold trading inventory. The Volcker Rule restricted proprietary trading — trading for the bank's own account rather than a customer's — and the broader post-crisis capital regime raised the cost of using a balance sheet to warehouse risk. Market making requires exactly that: standing ready to buy and sell, and holding what you bought until someone wants it. When that got costly for banks, firms that were not banks moved in, using private capital, technology and speed, and facing none of the bank-specific rules.
Jane Street, founded in 2000 by former Susquehanna traders, is the largest of them. On the most recent estimate we have seen, it handles more than 20% of US ETF trading and around 10% of US equity options and cash equities, with Citadel Securities' ETF share running at roughly half of Jane Street's. The firm reported $39.6bn of net trading revenue for 2025, a figure that surfaced across coverage in August 2026. For scale, JPMorgan's fixed income and equity traders brought in $35.8bn in the same year. In 2025, Jane Street, Hudson River Trading and Citadel Securities between them took in more than $60bn, beating the trading desks at JPMorgan and Goldman Sachs.
This is not a firm that competes with Wall Street. On these numbers it is Wall Street's trading floor, relocated into a private partnership with no shareholders, no capital ratio and no supervisor.
How the money is supposed to be made
The conventional business is dull by design. A market maker posts a bid and an ask, buys from sellers at the lower price, sells to buyers at the higher one, and keeps the difference. It hedges what it holds and tries not to have a view.
The ETF variant adds one wrinkle. An authorised participant can create new ETF shares by buying the underlying basket of securities and delivering it to the fund issuer, or destroy them by handing ETF shares back and taking the basket. So when an ETF trades above the value of what it holds, the AP buys the basket, creates shares and sells them at the premium; when it trades below, it buys the cheap shares, redeems them and sells the basket. The edge per trade is tiny after trading, financing and hedging costs. It becomes a business only at enormous volume, which is why the winners are the firms with the systems and the balance sheet to do it a few million times a day.
That description is the thing the July 2026 loss cannot be squared with. Nobody loses $15bn in a month collecting .
Regulatory friction and the SEBI case
On 3 July 2025, India's securities regulator issued a 105-page interim order against four Jane Street entities, alleging manipulation of the Bank Nifty index on options expiry days. SEBI's case, set out in the order itself, described two patterns: "intraday index manipulation" and "extended marking-the-close", in which coordinated trades across cash, futures and options moved the index in directions that suited the firm's options book. The conduct examined spanned January 2023 to 2025.
SEBI impounded ₹4,843.57 crore — about $566.3m — as alleged unlawful gains. Jane Street deposited the money into escrow and was permitted to resume trading in India under conditions. It appealed to the Securities Appellate Tribunal in September 2025. The tribunal admitted the appeal and asked SEBI to explain why certain documents had been withheld. Hearings ran into 2026 and were adjourned; the latest adjournment we can date is 25 February 2026. There is no final merits finding and no confirmatory order.
Which is to say: the largest unresolved question about whether this firm's arbitrage is arbitrage remains unresolved, more than a year on.
Financing the balance sheet
Warehousing inventory takes funding, and Jane Street gets a lot of it from the . In April 2024 it sold roughly $1.25bn of senior secured notes due 2031, rated BB by S&P.
August 2026 was a different order of magnitude. Bloomberg reported on 6 August that the firm was looking to rework an $11bn debt load into , and on 12 August that it would repay a $5.5bn floating-rate loan alongside refinancing about $5.6bn of existing bonds. The replacement was $14.6bn of senior secured notes in three tranches due 2031, 2033 and 2036, sized at $5.86bn, $5.13bn and $3.64bn. The longest tranche priced to yield 8.088%.
The ratings tell a curiously calm story next to that yield. S&P assigned BB to the new notes. Moody's had affirmed a Ba1 issuer rating on 22 July 2026 and moved the outlook to positive, citing strong profitability. Fitch, which carries a BBB- issuer default rating, said its ratings were unaffected by the funding restructure but noted that targeting a smaller, more concentrated pool of investors is credit negative for funding. A firm rated at the investment-grade boundary paying north of eight per cent for ten-year money is a firm whose lenders and whose rating agencies are not looking at the same thing.
There is also a neat circularity. On 13 August 2026 Bloomberg reported that Zenith Arc had raised $2.25bn of bonds against a data centre leased to Jane Street. We covered that alongside three other single-asset financings on 20 August. The firm is both a borrower in the complex and a tenant whose someone else's bond now depends on.
Why it matters to this crash
In July 2026 Jane Street lost $15bn — its first losing month in a decade — on directional bets, according to Bloomberg. Reuters tied the hit to the firm's stake in Leopold Aschenbrenner's and other AI positions. The FT, ranking the episode against history's biggest trading losses, put a $35bn July loss at Aschenbrenner's door on leveraged AI stock bets. We cannot reconcile the two figures from what has been published and have said so repeatedly rather than pick one. Treat the order of magnitude as real and the decimal place as unsettled.
What is not in dispute is the self-description. In its most recent debt offering, Jane Street told bond investors its had evolved to include longer-term positions, in line with a hedge fund. It absorbed the loss and still generated more than $40bn of net trading revenue over the year, ahead of its 2025 record. That is the reassuring part and the alarming part at once.
A bank cannot lose $15bn in a month on , because capital rules and supervisors make the position impossible to hold. A non-bank can. And this particular non-bank is the entity standing between a redemption order and a price in a fifth of the US ETF market. When the selloff comes, the firm you need to be quoting is the firm taking losses, and the correlation between those two facts is invisible until a month like July.
The knock-on effects were visible within weeks. Banks that had been exposed to leveraged AI vehicles and to Situational Awareness had hedged their jump risk by buying semiconductor options, and that demand was propping up volatility. Once the exposure disappeared, so did the hedges. Implied volatility on the VanEck Semiconductor ETF fell from 65% the previous month to 40%, the lowest since February, and the ratio of open puts to calls dropped to 1.89, the most call-heavy since early April. "Now they don't need those hedges, and I believe unwinding of those hedges has made volatility in the sector inexpensive," Zed Francis of Convexitas told CNBC. We wrote that up on 19 August. A fund failed, and the market's measured fear went down. Cheap protection after a blow-up is how a system that just took a loss ends up carrying more risk, not less.
This is the thesis in one firm. The same migration shows up in funding: the BIS reported in August 2026 that non-bank financial institutions now account for more than 40% of non-centrally cleared euro-denominated cash-borrowing , up from under 30% at end-2020. Federal Reserve research puts the at roughly $830bn as of September 2025, about double its 2020 peak. The function moved. The supervision did not.
What would make this dangerous
The operational precedent is Knight Capital. On 1 August 2012, a software deployment missed one of eight servers, reactivating a dormant routine that began firing unintended orders at the open. In roughly 45 minutes the system generated more than four million executions across 154 stocks and a pre-tax loss of about $440m — the SEC later put the figure above $460m. The firm did not survive it. That is the speed at which an automated market maker can stop existing, and Jane Street is roughly two orders of magnitude more load-bearing than Knight was.
The specific things we would watch for:
A second losing month of comparable size. One $15bn month against $40bn-plus of annual revenue is absorbable. Two is a capital event at a firm whose capital nobody outside it can measure.
A ratings reversal. Moody's moved the outlook to positive on 22 July 2026, before the July loss was public. A back to stable or negative, or an S&P action below BB, would mean the agencies have re-underwritten the hedge-fund side of the book. Note that we have seen an unsourced social-media claim of a CCC+/Caa1 downgrade; it conflicts with every rating agency document in front of us and we do not credit it.
The 2036 tranche trading wider than 8.088%. That yield already says the bond market disagrees with the ratings. If the disagreement widens, the funding cost of warehousing inventory rises, and the rational response is to warehouse less of it.
Fitch's concentration point turning real. A smaller, more concentrated investor base is fine until those investors want out at the same time.
An adverse SAT ruling in India, or a regulator elsewhere adopting SEBI's reading of expiry-day arbitrage as manipulation. The strategies are not India-specific.
The failure mode is not that Jane Street goes bust and bondholders take a loss. It is that a firm carrying 20% of US ETF liquidity has a bad month, widens its quotes, cuts its inventory, and does so on precisely the day everyone needs to sell. The price you get on an ETF is not a law of nature. It is a firm choosing to stand there.