How SpaceX Humiliated Wall Street

ELI5/TLDR

For 20 years the US stock market quietly shrank — companies bought back their own shares, stayed private longer, and got taken private, so there was always less stock to go around and prices floated up. That just reversed. SpaceX went public at a $1.78 trillion valuation in the biggest IPO ever, and a wave of tech giants is now selling stock instead of buying it back, all to pay for AI infrastructure. The twist: SpaceX dictated its own price, paid the banks a humiliatingly small fee, and gave public buyers shares with no votes, no right to sue, and a tiny slice of the company. The era of the stock market as a one-way machine that lifts your retirement account is over; it’s back to being a place where companies come to ask you for money.

The Full Story

The market that only ever got smaller

For roughly two decades, US equities operated under one quiet rule: there was always less stock available today than yesterday. Three doors all swung the same way. An IPO drought meant fewer new shares arrived. A buyback boom meant cash-rich firms spent enormous sums retiring their own stock. And private equity hoovered up listed companies and took them private. Less supply, the same wall of money, and prices drifted up.

“If you were an investor, this was great news. There was a lot of money looking for a home and a shrinking number of shares to put it in. And when more money chases fewer shares, prices tend to go up.”

Boyle is careful not to oversell this — earnings genuinely grew, and rates sat on the floor for most of the era. But the shrinking share count was a steady tailwind behind the longest bull market in history.

The reason the buybacks were so large tells you what these companies actually were: asset-light websites. They threw off cash and needed almost no capital to run. When they went public it was less to raise money than to let early VCs and employees cash out, and to mint a liquid stock they could pay staff in. Meta is the clean example — it raised about $16 billion at its 2012 IPO and has since spent well over $100 billion buying its own stock back (around $45 billion in 2021 alone, much of it near $330 a share shortly before the stock fell to ~$100 — buying back at the top).

The reversal, in one word: AI

As of this week, that machine has thrown itself into reverse. Goldman forecasts $225 billion of new IPO volume this year, and $675 billion once you add follow-ons and issuance from existing giants. The cause is the AI arms race. The asset-light websites have decided to become asset-heavy: data centers, millions of Nvidia chips, their own power plants, even talk of orbital infrastructure.

“The stock market has been redesigned into a giant $675 billion pawn shop where the most valuable technology companies on Earth awkwardly pass the hat around to pay their infrastructure bills.”

A useful aside on why staying private was so attractive: going public means opening the books to competitors every quarter, and a high stock price itself advertises to rivals how excited investors are about your line of business. You tip off the competition twice. The bar for what’s allowed public has also collapsed over time — per a 2002 Ritter and Welch paper, prestigious banks once wanted four years of positive earnings; by the dot-com peak, 79% of IPO firms had negative earnings at listing.

The largest IPO ever, on the company’s terms

SpaceX sold 555 million+ shares at exactly $135 each, raising ~$75 billion (up to ~$86 billion with the greenshoe over-allotment), valuing the whole company at $1.78 trillion. The tell is the size of the slice: that’s only about 4–5% of the company. Most IPOs offer ~20%. SpaceX raised a fortune while parting with almost nothing.

The deal terms reinforce who holds the cards. Dual-class shares give Musk’s class B ten votes each; the public’s class A get one. SpaceX reincorporated from Delaware to Texas in 2024, where you need a 3% stake (tens of billions) just to file a shareholder proposal. The charter forces disputes into private arbitration, waives jury trials, and bans class actions. Shareholder advocate Bruce Herbert called it unprecedented:

“The structure closes the voting door, the courthouse door, and the proposal door simultaneously.”

How Wall Street got humbled

The banks’ core justification for their fee is price discovery — set a range, call investors, build an order book, balance the company’s wish for a high price against investors wanting a deal. SpaceX inverted all of it. Matt Levine’s “Elon Markets hypothesis” captures it: Musk simply announced the price was $135 and told the banks to go sell it. No range, no negotiation.

“It’s the financial equivalent of a highly paid surgeon being asked to stand in the corner and hold a tray while the patient happily operates on himself.”

Stripped of function, they were stripped of fees. The historic US standard was 7%; big deals already discount (Facebook paid 1.1%, Uber 1.3%). For SpaceX the most powerful banks on Earth fought over the deal and settled for under 0.75% — still over half a billion dollars, but a status surrender. They responded not with resentment but, in The Economist’s borrowed Gen-Z term, by giving everyone “the ick”: Goldman and Morgan Stanley turned their lobbies into rocket-themed science fairs, BofA lit its Manhattan spire as a launching rocket. Goldman drafted the prospectus months early and took the lead-left role from Morgan Stanley’s Michael Grimes, long seen as Musk’s closest banker. (Boyle floats, then disowns, a joke theory involving the name “Grimes.”) Goldman’s CEO David Solomon had to formally deny he won the role by sliding into Musk’s DMs.

Where the money goes — and why $75 billion isn’t enough

This is the genuinely strange part. Research firm Capefar Advisors tallied SpaceX’s disclosed contractual cash commitments through 2030 and found a ~$235 billion cash gap. The largest IPO in history covers under a third of what the company has already promised to spend over four years — and the lead underwriter expects an even bigger burn than the critics do.

The first $20 billion isn’t going to space at all; it retires high-yield debt (some at 12.5%) from the former Twitter and xAI, both recently folded into SpaceX. Then the circular financing begins. Anthropic agreed to pay SpaceX ~$1.25 billion a month through May 2029 — a $45 billion commitment for compute that, per the filings, doesn’t fully exist yet. Google signed a separate $30 billion rental deal days before the IPO. And the FT’s “cursor option” footnote: an acquisition option with a $10 billion cash penalty for not doing the deal.

“Some of these firms are paying others with money none of them has yet for things that haven’t been built.”

SpaceX isn’t alone. Anthropic (~$900 billion valuation) and OpenAI have both confidentially filed — OpenAI’s filing helpfully noted it hasn’t actually decided to go public and there are things “easier as a private company.” Alphabet just raised ~$85 billion, its first stock sale in over two decades and the largest equity offering ever. Meta — the buyback poster child — now guides to $145 billion a year in capex, has done a ~$30 billion bond offering and a $27 billion Blue Owl private-credit deal, and is reportedly weighing new stock; on that news its shares fell ~7%.

The retail trap

Banks normally toss retail 5–10% of a hot IPO. SpaceX carved out 20–30% for retail. Boyle reaches for the academic literature: Kevin Rock’s 1986 winner’s-curse logic (refined by Aggarwal, 2002) says institutions are informed and retail uninformed — if a deal is great, institutions take it all and retail gets scaled back; if it’s a dud, institutions walk and retail gets the full allocation. “If you ask for 100 shares and actually get all 100, you shouldn’t be celebrating. You should be terrified.” (Here retail was 7-to-1 oversubscribed, so that signal is muddier — but the deal was clearly designed to need ~$20 billion of retail money.)

The exit mechanics are the clever bit. Brokers made it easy to buy and hard to sell. Fidelity cut its max IPO account balance from $500k to $2,000 to pull in small accounts; SoFi runs a 30-day anti-flipping ban and threatens a $50 fee for selling within 120 days. Meanwhile Musk reportedly pushed index providers for fast inclusion — NASDAQ created a rule admitting mega-IPOs to the NASDAQ 100 after just 15 trading days, FTSE Russell swept it into the Russell 1000 in 5. The S&P 500 alone refused to fast-track.

“Wall Street has in effect lured the retail public into a room, locked the doors for a couple of weeks and asked them to hold the price steady until the price-insensitive index funds are required to start buying.”

The UK got roped in too: the FCA’s new public-offer-platform regime let British retail buy without a full prospectus, via a broker called Marx Financial — which is expected to guarantee the company can operate for at least six months after the offer. A small London firm vouching to its regulator that a $1.78 trillion conglomerate is “good for the money at least until Christmas.”

Will it break the market?

Probably not. A rush of IPOs has rung the bell at cycle tops before — 1929, the late 1960s, the dot-com peak — and insiders selling all at once are usually right. But on plumbing alone the market can digest it: S&P 500 firms issued ~~$1.7 trillion of stock in the year to last September (~~$140 billion a month), so SpaceX’s $75 billion is barely two weeks of normal issuance. The real question is whether it’s a good investment. Deal size tells you nothing — Saudi Aramco was the prior record holder and has lagged; Visa’s 2008 IPO returned ~3,000%. What matters is the durability of the business and the price.

And the price is steep: over 90x trailing revenues (not profits — the company loses money at scale). Boyle’s parallel is Cisco, the picks-and-shovels star of the early internet, briefly the world’s most valuable company in March 2000. It survived, stayed dominant and profitable — and still took until December 2025, over 25 years, to close above its 2000 peak. Buy a great company at the wrong price and you can wait a quarter-century just to break even.

“Those asset-light software businesses… have become the 21st century version of the 19th century railroads. And like the railroads, they need a bottomless supply of capital to build the thing out.”

Key Takeaways

The 23-year era of public-equity contraction (IPO drought + buybacks + PE take-privates) has reversed; Goldman projects $675 billion of new equity issuance this year.
SpaceX IPO: 555M+ shares at $135, raising ~$75B (up to ~$86B with greenshoe), valuing the company at $1.78 trillion — the largest IPO ever.
It floated only ~4–5% of itself (vs the usual ~20%): maximum capital, minimum ownership surrendered.
Public buyers get class A shares (1 vote) vs Musk’s class B (10 votes); Texas incorporation + forced arbitration + class-action ban remove voting, suing, and proposal rights.
Banks accepted a fee under 0.75% (vs the historic 7%; Facebook 1.1%, Uber 1.3%) — still ~$500M, but a status surrender. Goldman led; Morgan Stanley’s Michael Grimes was sidelined.
“Elon Markets hypothesis” (Matt Levine): Musk set the $135 price himself, gutting the banks’ price-discovery role.
SpaceX has a ~$235 billion contractual cash gap through 2030 (Capefar Advisors) — the IPO covers under a third; the lead underwriter expects an even bigger burn.
First $20B repays high-yield debt (up to 12.5%) from the former Twitter and xAI, now folded into SpaceX.
Circular AI financing: Anthropic to pay SpaceX ~$45B ($1.25B/month to May 2029) and Google ~$30B, for compute/data centers not yet fully built.
Retail got an unusually large 20–30% allocation (7x oversubscribed); brokers restrict selling (Fidelity $2k cap, SoFi 30-day ban, $50 early-sale fee) while NASDAQ (15 days) and FTSE Russell (5 days) fast-track index inclusion to create forced passive-fund buying. S&P 500 refused to fast-track.
Valuation: >90x trailing revenue, company is loss-making. Cisco parallel: bought at the 2000 peak, it took >25 years (to Dec 2025) to break even despite surviving and thriving.
Market can absorb the supply ($75B ≈ two weeks of normal S&P issuance); the open question is returns, not capacity.

Claude’s Take

This is Boyle doing what he does best — taking a story everyone covered as “biggest IPO ever, rockets, Musk” and reframing it as a structural regime change in capital markets, with the deal terms as the real news. The central insight is genuinely good and underappreciated: the two-decade tailwind of a shrinking share count is reversing because the FAANG-era business model (asset-light, cash-gushing) is being deliberately abandoned for asset-heavy AI infrastructure. That’s a clean, defensible thesis.

The sourcing is solid where it matters. The Ritter and Welch (2002) and Kevin Rock (1986) papers are real and correctly characterized; the Levine “Elon Markets hypothesis,” the Capefar $235B gap, the Anthropic/Google compute contracts, the broker anti-flipping rules, and the index fast-track asymmetry are all checkable claims, not vibes. The Cisco-as-cautionary-tale comparison is the strongest analytical move — it cuts through the “but it’s a revolutionary company” defense with the one point that actually governs returns: price paid.

Where to apply the BS filter. First, the framing tilts hard toward “retail is being trapped” — but Boyle himself notes the retail book was 7x oversubscribed, which undercuts the winner’s-curse “you only get a full allocation on duds” warning he leans on. He flags this honestly, but the segment’s rhetoric runs ahead of its own caveat. Second, the $28.5 trillion TAM mockery is fair and funny, but it’s beating up a prospectus number that no serious analyst takes literally anyway. Third, several of the scariest figures (the $235B gap, the circular financing) are real but are forward commitments and contracts, not losses — a company can rationally pre-commit to spend it intends to fund through future raises, which is the entire point Boyle is making. The danger is real; the framing makes it sound more like a Ponzi than a capital-intensive build-out.

Net: this is high-quality financial journalism, dry and well-referenced, with a real structural thesis rather than just Musk dunking. Docked slightly because the retail-victim narrative is a touch overcooked relative to the facts he himself presents, and because the most alarming numbers are presented with more menace than their forward-looking nature warrants. An 8 — informative, honest about its own gaps, and genuinely changes how you read the IPO wave.