The $75 Billion Bill: Who Funds the SpaceX IPO

On June 12, 2026, SpaceX begins trading on Nasdaq under the ticker SPCX, priced at $135 a share. It will sell about 555.6 million shares and raise close to $75 billion at a $1.75 trillion valuation — the largest IPO in history, more than twice the previous record, Saudi Aramco’s 2019 listing.

It comes down to one number: someone has to hand SpaceX $75 billion. To buy that much new stock and stay fully invested, the rest of the market has to sell an equal amount of what it already owns. The rest of this piece traces where that selling comes from and what it moves.

The raise is a bill to the rest of the market.

An IPO of new shares is a cash transfer to the company. SpaceX collects ~$75 billion from institutions (~$52.5 billion) and retail (~$22.5 billion, via the record 30% carve-out through Robinhood, Fidelity and Schwab). Index funds buy none of the raise itself — they can’t hold SpaceX until it’s added to an index, which begins after the listing.

That cash has to be funded, from spare cash or by selling something else; a fully invested book has only the second lever.

That selling doesn’t net out, though: every sale has a buyer, so the market’s holdings don’t actually shrink. What leaves the system is $75 billion of cash, into SpaceX — and the scramble to rebuild it is what pressures prices, on whatever gets sold first.

The closest cousins can’t pay.

So what gets sold? The instinct is to fund a new holding by trimming its closest cousins — and SpaceX has two sets of them, one for each frontier business. Both come up short.

On the space side, the public pure-plays — Rocket Lab, AST SpaceMobile, Intuitive Machines, Planet — have had enormous runs, but they still trade thin: even the most active turns over only a billion or two a day. On the AI side, its truest peers — the model builders, OpenAI and Anthropic and xAI itself — are private. The public proxies are the neocloud names: CoreWeave, Nebius, IREN, Applied Digital, all recent listings with thin float:

The whole pure-play complex — these names plus smaller peers like Planet (PL) and Applied Digital (APLD) — trades around $11 billion a day, roughly a third of one day’s NVIDIA (~$32 billion). That is no pool to source tens of billions from: the truest peers can’t be sold, and the rest would gap violently and still fall short.

So the bill falls on the giants.

It lands where the depth is: the mega-cap AI names and large semis — not the closest substitutes, but the only ones that can give up tens of billions in days. Routed there, the ~$51 billion that has to be sold (sized below) goes down comfortably:

Those are the largest landing spots; the rest of the $51 billion spreads across the wider complex. NVIDIA takes the biggest single amount — about $9 billion — and barely notices, trading around $32 billion a day. The tight spots, even here, are the semiconductor ADRs: Taiwan Semi and ASML, where the selling approaches a full day’s volume.

So the dollars come out of the giants — the names least like SpaceX, because they are the only ones deep enough to sell.

The index rotation is the visible tip.

Set against the funding rotation, the index buying that arrives with the IPO is small. The trackers buy SpaceX into each benchmark and sell pro-rata to fund it — the legible, rule-bound part. At the IPO it’s the tip; the funding rotation is the iceberg.

It also arrives in stages, and only the first is near the IPO:

In those first weeks, the index funds buy only about $14 billion, because just ~4% of SpaceX floats at the IPO — Vanguard’s VTI around the fifth trading day, the Nasdaq-100 around the fifteenth. Their buying grows over the next one to three years as lock-ups expire, and would grow by another ~$166 billion if SpaceX ever enters the S&P 500. But the S&P is shut: on June 4 its committee kept the profitability and seasoning gates in place, so that money waits until SpaceX turns a GAAP profit — years away, since the early-2026 xAI merger turned a profitable rocket-and-Starlink business into a $4.9 billion loss.

Why it isn’t a clean $75 billion hit.

Two things soften the blow. First, not all of it is funded by selling — retail often pays from new deposits, and some institutional money is fresh inflow — which is why the base case is ~$51 billion of selling, not $75 billion. Fund none of it from cash and it is the full $75 billion: every figure in the giants table about 50% larger.

Second, the cash doesn’t vanish; SpaceX spends it back into the system. About $20 billion repays its bridge loan, returning to the banks that made it. The remaining ~$55 billion funds capex over the next year or two — much of it AI compute, which means Nvidia chips. So some of the money raised by selling Nvidia comes back to Nvidia as revenue. The net effect is a transient drain on liquidity and a rotation in who owns what.

What it adds up to.

The real footprint at the IPO is the ~$51–75 billion the market sells to fund the raise, not the ~$14 billion the index funds buy alongside it. The cousins can’t carry that, so it comes out of the liquid AI mega-caps and semis — and since SpaceX spends the cash back, it’s a rotation, not a drain.

Method notes & sources

• Raise = 555.6M shares × $135 = ~$75B of new equity. Buyer mix ~70% institutions / ~30% retail. Index funds buy $0 at the IPO (they cannot hold SpaceX until it joins an index).
• Selling-to-fund = raise net of the share funded by cash/new money. Base case: retail 50% cash, institutions 25% cash → ~$51B sold; fixed-pie (no new cash) → $75B (scales every figure ~1.5×).
• Flow of funds: a primary issue is, in aggregate, a cash transfer to the issuer, not a net sale of other shares (every sale has a buyer); the price pressure comes from investors rebuilding cash. SpaceX recycles the cash: ~$20B bridge repayment + ~$55B capex (much of it AI compute), so the drain is transient.
• Cousins: SpaceX’s closest public pure-plays — space (Rocket Lab, AST SpaceMobile, Intuitive Machines, Planet) and AI compute/data centres (CoreWeave, Nebius, IREN, Applied Digital). Combined they trade ~$11B/day, about a third of NVIDIA alone; the truest AI peers (OpenAI, Anthropic, xAI) are private and unsellable. This shows the cousins cannot be the funding pool — only the natural-but-unavailable one.
• Landing basket: deep, liquid AI mega-caps and large semis (where the bulk of the selling can actually be absorbed). $ sold per name = basket weight × dollars sold; days of volume = that $ / avg daily $ volume.
• Average daily $ volumes are measured from IBKR daily bars (swing project), a ~3-month average through 2026-05-13 (e.g. NVDA ~$32B, TSM ~$4.9B, ASML ~$2.4B; cousins ~$0.3–2.8B); PL and APLD are estimates (not in the data set).
• Index rotation (secondary): Nasdaq-100 weights on listed cap, capped at lesser of listed cap or 3× float (FAQ eff. 2026-05-01; 10% float minimum removed; fast entry ~15 trading days). CRSP total-market adds large IPOs on the 5th trading day (its Apr-2026 float-cap test lets SpaceX qualify), so VTI is the first index buyer; other families float-adjusted. Calibrated to Tesla (Dec 2020): model 1.65% / $70.9B vs actual 1.69% / $72.7B. Outputs: Stage 1 ~$14B (Nasdaq ~$7.5B); Stage 2 mature non-S&P ~$105B (~$90B post-entry); Stage 3 S&P latent ~$166B, gated by the 2026-06-04 S&P DJI decision (GAAP profit + 12-mo seasoning + IWF ≥ 0.10 all kept).
• Sources: SpaceX S-1 (2026-05-20) and S-1/A amendments (2026-06-01, 2026-06-03); Nasdaq-100 methodology FAQ (eff. 2026-05-01); CRSP US Total Market methodology (Apr 2026); S&P DJI MegaCap consultation results (2026-06-04); peer volume from IBKR daily bars (3-mo avg through 2026-05-13).

This article does not constitute investment advice. For the filing that opened this story, see The Stars Our Destination: SpaceX Files for IPO.