US Margin-Debt Growth Fell 22 Points in Four Months. The Leverage Migrated to Interactive Brokers, Charles Schwab and Robinhood — Right as the Fed Hiked

Strip out the base and the deceleration disappears. US margin debt rose from $1,415.6 billion in May to $1,480.1 billion in September — grew 4.6% while the year-over-year growth rate fell 22.3 points. The only reason the headline reads like deleveraging is that the comparison base got 22.3% heavier, because margin debt was already building fast a year ago.

The aggregate hides the migration. Between September 2025 and September 2026, US margin debt grew roughly $354 billion. Charles Schwab alone added $90.7 billion, Interactive Brokers added $27.9 billion and Robinhood added $8.2 billion — three online brokers created about 36% of all new US margin debt while sitting on just 16.7% of the stock a year earlier.

Quality matters as much as quantity. Schwab's own footnote attributes $60.7 billion of its $187.9 billion margin balance to long/short strategies on which it earns a fixed net yield rather than a floating loan rate. That is systematic, crowded, professionally risk-managed money — the part of the book with the tightest limits and the fastest de-risking reflex when a crowded trade gaps.

The timing is not neutral. The Federal Reserve raised its target range to 3.75%–4.00% on September 16, 2026 — its first increase in three years — and reconvenes October 27–28, with a hold the base case and a hike the live tail. That reprices the cost of carry on $1.48 trillion of borrowed stock, and the effect on the brokers is already visible in their own numbers.

Read the rows together and the squeeze is obvious. Robinhood's margin book grew 59% while its net interest revenue fell 9%, because a 35-basis-point fall in Interactive Brokers's margin loan yield means every new dollar of borrowed stock earns less. IBKR's answer was volume — average margin loans up 47% — and its net interest margin still fell to 1.99% from 2.16%, with pre-tax income up 14%, slower than the 19% growth in net interest income.

The right denominator changes the answer. Margin debt was 2.3% of S&P 500 market capitalization in Q2 2026, exactly the average going back to 1997, which implies an index of roughly $65 trillion. The 2026 burst was also smaller than prior ones: the April–May two-month surge of 15.9% was only the ninth-largest on record, against 19.3% in January 2000 and 20.3% in May 2007.

The most revealing document in the chain came from Seoul, not New York. Korean retail investors net bought $4.5 billion of US stocks in July 2026, roughly $840 million of it into chipmaker ADRs, and the SK Hynix ADR became the second-most net-purchased US security of the month, trading at a roughly 10% premium to the Korean line. Four of the ten most net-purchased US stocks that month were leveraged products, with a 3x daily semiconductor fund the most popular of the set.

The venue that profits from the next move is already visible. CME Group posted its highest-ever third-quarter revenue, $1.7 billion, on record average daily volume of 29.4 million contracts, with interest-rate products at 15.1 million contracts a day and equity indexes at 7.7 million. A Fed decision, a margin-call cascade and a hedge build all trade in the same contracts, and the exchange is paid on the volume and the volatility.

Read the scoreboard together and the leverage boom is good for balance sheets and bad for the spread business. Interactive Brokers is the cleanest expression: margin loans up 36% against an industry up 31.4%, accounts up 35% to 5.58 million, and a 78% pre-tax margin for an eighth straight quarter above 70%. Robinhood is the mirror image — a record $22.1 billion margin book and record $1.42 billion of revenue, but earnings per share of $0.60, down 3%, and a share price that fell on the print.

The investable conclusion is narrower than the headline. US leverage is not unwinding; it is being re-underwritten by a smaller number of brokers, at faster rates, into a smaller number of names, at a rising cost of carry. The margin-debt slowdown is a base effect wearing the costume of a warning, and the risk it conceals is compositional rather than systemic: the marginal levered dollar is more levered, more concentrated and more foreign than the average dollar.