The $1.4T IPO problem isn't demand. It's the Russell rebalance.

What happened

The Economist ran the numbers this week on a question that has been quietly nagging at every late-stage equity holder in the Valley: if Anthropic ($350B last mark), OpenAI ($500B+ on secondary), and SpaceX (~$550B) all listed in the same 18-month window, can the public market actually digest them?

The headline figure is roughly $1.4 trillion of float looking for a home. For scale: that's larger than the combined market cap of Berkshire Hathaway and JPMorgan. It's roughly 3% of the S&P 500, packed into three tickers that would arrive correlated to each other on AI capex, GPU supply, and U.S. industrial policy. The Economist's framing — can the stockmarket *swallow* this — is the wrong verb. Markets clear at a price. The interesting question is what gets sold to make room, and on whose schedule.

The mechanical answer is that FTSE Russell's annual reconstitution and S&P's quarterly committee meetings, not the IPO bookrunners, become the binding constraint on liquidity.

Why it matters

Here is the part the generalist financial press keeps missing. About 54% of U.S. equity AUM is now passive (ICI, 2025). When a name the size of OpenAI hits the Russell 1000 at the June reconstitution, every index-tracking fund on earth has to buy it — not because they like the story, but because their tracking-error mandate gives them no choice. That's the *forced* leg of demand. It's also the leg that arrives on a calendar everyone can see, which is why active managers front-run it every year and have done since the 1990s.

Three things break when the inclusion candidate is this big and this correlated to existing holdings:

1. The funding trade. To buy $40B of OpenAI at inclusion, passive funds sell pro-rata across every other holding. Nvidia, Microsoft, Meta — the names most exposed to the same AI capex cycle — get sold to fund a position in… more AI capex exposure. The index does not know it is rotating you from one AI bet into a more concentrated AI bet; it is just rebalancing weights.

2. The lock-up cliff. Standard 180-day lock-ups mean the float available at IPO is a small fraction of fully-diluted shares. Index providers have started using free-float-adjusted methodologies precisely to avoid this — but the adjustment lags. The result is the well-documented "inclusion pop and fade": the stock rips into the rebalance, then bleeds for 6–12 months as employee and VC supply unlocks into a market that already bought what it had to buy. Snowflake, Rivian, and Coinbase all traced this curve.

3. The VC distribution waterfall. Sequoia, a16z, Founders Fund and the sovereign LPs behind them don't get cash at IPO. They get shares, distributed in-kind to LPs after lock-up. Those LPs — pensions, endowments, family offices — then sell on their own timeline, which is mostly "as soon as compliance lets us." The $1.4T isn't a one-day absorption event; it's a 24-month distribution overhang that the VC accounting calendar, not the market, dictates.

The HN thread on the Economist piece (448 points at time of writing) is doing the usual thing of arguing about whether OpenAI is worth $500B. That's the boring question. The interesting comments are from people who lived through the 2021 SPAC unwind, pointing out that the absorbable supply of "willing" capital — i.e., capital not forced in by indexing — is much thinner than headline AUM suggests. Hedge funds are not going to take the other side of a forced buy at any price retail will tolerate.

What this means for your stack

If you hold pre-IPO equity at any of the big three, or at the second tier of AI labs that will trade off these comps (Mistral, xAI, Cohere, Perplexity), three practical things follow.

Model your liquidity to the Russell calendar, not the S-1 calendar. The IPO is when the clock starts. The actual price-discovery event for your shares is the next reconstitution after lock-up expiry. For a June IPO with a 180-day lock-up, that's the *following* June — a 12-month wait, and a known event that everyone will trade around. Plan tax events accordingly. If you can, ladder your 10b5-1 sales to start before the reconstitution, not after.

Reassess concentration in your 401(k). If you work at one of these companies and you also hold a target-date fund or an S&P 500 index, you are about to be massively overweight a single thesis. The default "diversified" portfolio of 2027 will have more AI-correlated exposure than the dot-com peak portfolio did in software in March 2000. This is a position to size deliberately, not by default.

For founders raising now: the comps are about to break. Your Series C is being marked against secondary trades in OpenAI and Anthropic at prices that assume public-market exit math. The moment those companies actually print public quotes — and trade down post-lockup, as the mechanics above predict — every late-stage round in the AI vertical gets re-marked. If you're raising, close before the first of the three lists. If you're a buyer of secondaries, wait.

Looking ahead

The consensus take is that the IPO window opens in 2026 and these three are the headline acts. The under-discussed mechanic is that they will likely list in sequence, not parallel, precisely *because* the bookrunners know the absorption problem is real. Watch for SpaceX first (most operational cash flow, cleanest story), Anthropic second (B-corp structure is the only governance complication left to resolve), OpenAI last (the for-profit/non-profit conversion is still litigated). Each listing teaches the market about the next one, and each rebalance window after each listing is a tradable event. The story isn't whether the market can swallow $1.4T. It's that the swallowing happens on FTSE Russell's calendar, and you can read that calendar a year out.