The 135-Day Rule JPMorgan Just Gave Up to Win AI Wealth

Here is a rule J.P. Morgan made up for itself: it will not lend money against stock in a company that recently went public until the company has been public for 135 days. That is not a government rule. Federal margin rules, which exist to discourage lenders from treating brand-new stock as perfectly good collateral, generally let a broker-dealer lend against an IPO's shares after 30 days. Goldman Sachs, according to the Financial Times, typically doesn't wait longer than that. J.P. Morgan waited 135 days, four and a half times as long, because it liked being — or appearing to be — the conservative lender. And now the bank is quietly abandoning that position, while reportedly telling the press that nothing about its policy has changed.

The product is securities-based lending, and its underlying economics could not be older. A rich person who wants money without selling stock borrows against the stock instead: you pledge the shares, the bank hands you cash. You keep the shares, you keep the upside, and — the part that matters for rich people — you never "sell," so you never trigger the capital gains tax. This is the famous "buy, borrow, die" strategy, the mechanism by which wealthy families finance their lives and pass appreciated assets to their heirs, and J.P. Morgan's own private bank markets it as a "potentially tax-efficient" way to get cash without selling.

The reason any waiting period exists is that a stock that went public last month is the worst collateral in the financing business. It has no established price. It trades only a thin slice of its float, and most of its shares are held by employees who are contractually locked from selling. The Fed's Regulation U, for that matter, caps loans used to buy or carry "margin stock" at 50 percent of the collateral — and margin stock means exchange-listed or Nasdaq-traded securities, so private shares sit outside that regime entirely (and a personal-spending loan is not "purpose credit" anyway). For this kind of money, the bank sets its own terms; it is its own appraiser, its own underwriter, and its own risk committee. The 135-day rule was the bank imposing on itself a discipline the law mostly declines to impose. The number is oddly specific; the function is not: wait until the market has actually traded the stock for a while before letting a client borrow against it.

J.P. Morgan's pitch to SpaceX shareholders makes its pessimism explicit. The private bank offers a revolving line of credit secured by private-company shares — one year, at least $2 million, interest-only monthly payments at a variable rate of SOFR plus two percentage points, no application or origination fees — and the loan has to be covered at least ten times over by the pledged stock. That is a dollar lent for every ten dollars of SpaceX put up, an advance rate that is the bank's way of saying it genuinely doesn't know what the shares are worth. The product is designed to carry a client "through the SpaceX IPO and subsequent lock-up periods," with a refinance into cheaper financing once the lock-up lifts. The point is to let the owner borrow instead of selling.

Which raises the obvious question: why would a careful bank take risky collateral, price it like it is radioactive, and still compete to make the loan? Because the loan is the front door to the client. IPO time is short, and the relationship is long: the listing produces a tax bill, a brokerage account for the proceeds, a trust, an estate plan, a mortgage, and thirty more years of salary and stock compensation. The bank that lends against the trapped stock first gets to hold all of that. The FT frames the current push as a race for a specific pile. SpaceX went public in June in the biggest IPO ever. Next comes Anthropic — the maker of Claude — last valued around $965 billion and expected to float as soon as October at a valuation of $2 trillion or more, with engineers at the top AI labs earning millions to tens of millions in company stock. J.P. Morgan was one of two dozen banks on the SpaceX deal, and Goldman Sachs and Morgan Stanley lead the AI floats, so its edge has to be something else — and lending is the something. J.P. Morgan took $75 million from its role on the SpaceX listing; Morgan Stanley, which administers employee equity plans, collected more than $74 billion in net new assets in the second quarter from IPO wealth, including SpaceX. The market has been pricing the banks for this pipeline: analysts forecast Morgan Stanley, J.P. Morgan, Bank of America and Citigroup earning a combined $74 billion this year, up 26%, and J.P. Morgan's shares sit near record highs. A chunk of that valuation is an option on AI companies continuing to go public and minting new millionaires.

Now the part worth thinking about, because it is the reason the seasoning rule existed in the first place — and the easing does not remove the risk; it moves it earlier. The ten-times coverage protects the bank from price, up to a point: lend a dollar against ten dollars of stock and the loan scrapes through even a 90% collapse in the collateral, on paper. What it does not protect against is the two ways this kind of collateral actually dies. First, the shares can be trapped: if the client's stock is locked up, the bank cannot sell it to collect, so a margin call against a locked position is mostly a negotiation between a rich person and the person who lent them money. Second, "listed" is not the same as "liquid." A stock that has traded for 30 days is technically usable as collateral; whether it is genuinely liquid depends on how much of it you can sell without moving the price. SpaceX is the demonstration. It priced its record IPO at $135 a share in June, popped 19% in its first day to close at $160.95, and then spent the summer fading — it touched about $134 in mid-August, below its IPO price — before settling near $141 by late August, still more than 10 percent below the first-day close. That happened while the supply side swelled: on August 6, around 912 million shares became eligible for sale — more than the entire public float had been, so the unlock more than doubled the shares available to trade — with another 12.9 billion shares set to unlock by mid-2027. Musk alone holds about 42% of the company and can't sell for a year. That is the difference between listed and liquid, and it is exactly what a seasoning rule was meant to make you wait out.

So read the bank's official statement correctly. J.P. Morgan said its "overall policy remains unchanged" and that it assesses transactions on a "case-by-case, client-by-client basis," exceeding regulatory requirements by weighing market liquidity. This is, in a literal sense, true: a policy built out of client-by-client exceptions is never "changed"; it is merely applied. But that is the frame. The 135-day rule was never really the bank's protection — the haircut was, and the relationship was, and both of those are intact. What changed is that J.P. Morgan has decided that being the last lender to the newly public rich is a luxury it can no longer afford. The conservative position was costing it the fastest-growing pile of new money in finance, so now it competes on speed: 135 days, then 30, then — for the right client — zero, lending against stock before the company has traded a single public share.

For an ordinary investor, two things follow. First, if you own the big banks, you are now explicitly holding an option on the AI-IPO machine continuing to produce new millionaires — the machine whose newest poster child spent the summer giving back its entire first-day pop as the float swelled. Second, the mechanics are universal: whenever you borrow against your own holdings, the whole game is in the collateral terms — the haircut, the seasoning, who can sell and when. The reason J.P. Morgan can be so strict on price (ten to one) and so suddenly loose on time (135 days down to zero) is that it is the same box: the bank gets its safety from the gap between what it lends and what the collateral is worth, not from waiting. It has simply decided the wait was costing more than it was worth.