There’s a decent chance you became a SpaceX shareholder this morning.

It’s July 7, 2026. 8 a.m. You’re standing on your new rug with your coffee, scrolling the news. There’s a decent chance you became a SpaceX shareholder today. Index funds tracking the Nasdaq-100 began buying an estimated $22 to $27 billion of SpaceX stock this morning. More than Iceland’s GDP. In a single rebalance. Into a company that has been publicly traded for twenty-five days.

That’s what index funds do, right?

But the rule that made it possible is ten weeks old. And the one it replaced stood for fourteen years.

The seasoning period was the mandatory proving ground every newly listed stock had to survive before retirement money could be made to buy it. It began as a two-year wait. In 2012 it was trimmed to three months. For fourteen years after that it sat there, the last wall between unproven stocks and American retirement accounts, through every bubble and mania since. Nasdaq waived it in May. Fifteen trading days. No vote, no comment period, no regulator’s signature. The exchange that collects SpaceX’s listing fees rewrote the rulebook of its own index. How convenient.

The float minimum was eliminated outright, replaced with a multiplier that lets a five-percent float carry triple its weight. That minimum used to require at least ten percent of a company to actually be available to the public before index funds could be forced to buy it. It is the rule that would have kept SpaceX out of your retirement account entirely.

Both changes happened the same day. By the same institution. It’s almost like Nasdaq knew what they were doing.

The S&P 500 was offered the same “modernization.” They refused. Their rules still require a full year of trading and actual profits.

Musk at SpaceX? Has neither.

The last of xAI’s eleven co-founders resigned in March. Nasdaq’s new rulebook took effect in May. SpaceX went public in June at a $1.75 trillion valuation that Morningstar calls 55 percent too high, the largest initial public offering in history, three times the previous record. And this morning, while you were sipping your coffee on your new rug, the retirement accounts of millions of Americans began buying it. Automatically. By rule. On schedule. No single step required your consent. It was manufactured for you.

This is a story about a gap. On one side is the sky-high valuation: roughly $2.1 trillion at this morning’s open. On the other side is the floor, the value of the things SpaceX actually does. Launch rockets. Run Starlink. Book revenue. Morningstar’s analysts put that number at $780 billion. The distance between the ceiling and the floor is the story the market is telling itself about artificial intelligence, strapped to a division whose entire founding team walked out the door.

Depending on whose math you trust, that distance is somewhere between $200 billion and $1.3 trillion. And as of this morning, it lives inside the one kind of account with no door to walk out of. That is the longest distance between someone’s money and an exit in the history of the stock market.

Every co-founder of xAI is gone. All eleven. Not most. All.

xAI was founded in July 2023 by Elon Musk and eleven researchers pulled from DeepMind, Google, Microsoft Research, and the University of Toronto. By March 14, 2026, the last of them had resigned. The departures came in waves across eighteen months. Kyle Kosic left in mid-2024 and went straight to OpenAI. Igor Babuschkin left in August 2025 to start his own firm. Then it accelerated. Tony Wu, who led xAI’s reasoning team and was described as the most operationally central co-founder, announced his departure on a Tuesday. Jimmy Ba, who directed research and safety, resigned within twenty-four hours. Toby Pohlen followed. Zihang Dai and Guodong Zhang followed. Manuel Kroiss, who led pretraining, the man who taught Grok how to think, told associates he was done. Ross Nordeen, Musk’s right-hand operator, walked out last.

When the final resignation landed, the company merged into SpaceX. Wu exits, Musk keeps the data centers. Ba exits, Musk keeps the name. Kroiss exits, Musk keeps the training pipeline he built. Eleven departures, eighteen months, every researcher. Everything that could be bolted down stayed. Everything that could think left.

That has never happened before at this scale, to a company at this valuation, this close to a public offering. Companies lose founders and keep building. Apple lost Jobs in 1985 and shipped the Macintosh II. Google’s founders stepped back in 2019 and the company doubled. But those were operating businesses with revenue, margins, and products that worked without their creators in the building.

An AI valuation is different. It is a bet on talent compounding against compute, and the market knows it, because talent scarcity is the entire reason AI researchers command NFL salaries. Meta was offering $300 million over four years to keep its best people in their seats. That is the price tag on the dimension SpaceX just lost. When every researcher walks, the product that valuation is built on zeroes.

Four days after listing, SpaceX spent $60 billion in freshly printed stock to acquire Cursor, an AI code editor. Largest venture-backed acquisition in history. The stock they used to pay for it did not exist a week earlier. The IPO had minted roughly $740 billion in paper gains in its first four trading sessions, and SpaceX used a sliver of that paper to shop for a replacement for the team that had just walked out. Microsoft looked at the deal and passed. OpenAI approached twice and was turned down. SpaceX printed the money and closed in four days.

What did that money buy? A company whose market share had fallen from 41 percent to 26 percent in the twelve months before someone paid a record price for it. Not because the acquisition fixed the zero in the multiplication. Because the purchase looked like momentum, and momentum is what holds a ceiling up when the floor is $1.3 trillion below it. The market’s response was to erase $600 billion in SpaceX’s value over four trading sessions. Ten times the purchase price, subtracted by the same investors who were supposed to be buying the story.

The prospectus reads in stretches like a whitepaper. Fluent in the future tense, glossy and forgettable about the quarter just ended. That quarter says this: $818 million in revenue against a $2.47 billion operating loss. Three dollars and two cents burned for every dollar earned. The division spent $7.7 billion on capital expenditure in the first quarter alone, more than its entire 2025 revenue.

Grok, the consumer product, the thing the AI valuation is supposedly about, generates roughly $500 million in annualized revenue. Anthropic’s Claude generates $47 billion. In January, Anthropic cut off xAI’s access to Claude after discovering xAI’s own engineers were using it inside their coding tools. The staff of the AI company preferred the competitor’s model to their own. Reid Hoffman, who has watched a few of these collapse up close, called the venture an “unmitigated disaster” working through its third attempt at a reboot.

And the AI division’s biggest customers are those same competitors. Anthropic signed a lease on xAI’s Colossus data-center capacity worth $1.25 billion a month through 2029. Google signed for $920 million a month. Together that is $2.2 billion in monthly rent, paid by two companies that both built the product xAI could not, for the privilege of using the hardware xAI’s departed team left behind.

There is a word for what happens when a company loses a non-substitutable dimension of a multiplicative product. It does not decline proportionally. It collapses to the utility floor. Nokia had hardware, software, and ecosystem. When the ecosystem dimension went to iPhone, the stock did not lose a third of its value. It lost ninety-five percent. The hardware and the software still existed. The zero in the multiplication zeroed the product. Yahoo went from $125 billion to a $4.5 billion fire sale. BlackBerry went from $80 billion to $2 billion. In every case, the surviving dimensions were real. The business was real. The floor existed. But the market had been pricing the multiplication, and when one dimension went to zero, the repricing was not gradual. It was a cliff.

SpaceX without AI talent is a launch provider, a broadband ISP, and a data-center landlord whose tenants are its competitors. That is not worthless. That is $780 billion. But $780 billion is sixty-three percent below where index funds began buying this morning.

SpaceX earned $791 million in 2024, the year before the merger. In 2025, the combined company lost $4.94 billion. The year the AI division was absorbed, the profitable company became a money-losing one.

Crypto has a name for what happens when the builders leave right before the buyers arrive: a rug pull. In crypto it resolves in minutes, because token holders can sell the moment they smell it. Here, the largest buyers arriving this morning are structurally unable to sell. Their mandate is the index, whatever the index holds. That difference is not an accident of engineering. It took a season of rule changes to build.

The rug is set into position.

Wall Street’s principal lobby is the Securities Industry and Financial Markets Association, SIFMA. It has kept four outside lobbying firms on retainer through every quarter of 2025 and 2026, per Senate disclosure records: Crossroads Strategies, Avoq, Harbinger Strategies, and Rich Feuer Anderson. They work the House, the Senate, the SEC, the Federal Reserve, and Treasury. The securities and investment industry as a whole disclosed $51.4 million in lobbying in the first quarter of 2026 alone. Twenty-six of SIFMA’s forty-two lobbyists previously held government jobs. That is the background hum. Here is the melody:

Two things happened on March 30. The SEC issued an exemptive order easing what counts as collateral in securities lending. And Nasdaq announced the overhaul of its index methodology, the one you already know: seasoning collapsed to fifteen trading days for any newcomer in the top forty by market capitalization. Float minimum erased. Multiplier installed. The new rulebook took effect May 1. Trade coverage at the time read the changes as purpose-built for one company. Nineteen days later, that company filed its S-1. No major index has ever rewritten its eligibility rules and onboarded a company through them in the same quarter.

Two institutions were handed the same pen. S&P Dow Jones Indices, offered the same deal for the S&P 500, declined on June 4. Kept its twelve-month seasoning. Kept its profitability requirement. Looked at a first-quarter loss ratio and a five-percent float and said no. The refusal is the proof that the signature was a choice.

The customer protection rule has held since 1972. In June, it gave ground twice in nineteen days. SEC staff issued a no-action letter to SIFMA on June 10 and another on June 29, both loosening Rule 15c3-3, the wall between client assets sitting at a broker and the broker’s own bets. A no-action letter is how the machinery moves when nobody wants a vote: the industry asks in writing, the staff answers in writing, and the answer carries the force of practice without the friction of rulemaking. No hearing. No public comment. No record that you were a party to their decision.

None of this was announced to you. It was announced in filings written for the people paid to read them.

You don’t know how the rug was made. It just looks nice and it’s comfortable to stand on.

The largest initial public offering in history priced at $75 billion. Three times the previous record. SpaceX minted 555.6 million new shares at $135 against $150 billion in orders, twenty-one banks on the cover, Goldman Sachs lead-left. Thirty percent of the offering, roughly $22.5 billion, was allocated to retail investors through Robinhood, Fidelity, Schwab, SoFi, and E*Trade. Triple the typical retail slice of a mega-cap offering. No IPO in market history has handed retail that large a share of that large a deal. They were so precisely considerate to include the everyman.

That allocation did not happen by accident. Fidelity slashed its IPO eligibility from $500,000 to $2,000 for this offering alone. A 99.6 percent reduction. The brokerage that once required half a million dollars in your account to touch an IPO now required the cost of a used laptop. Why? Because SpaceX reserved thirty percent for retail, three times the industry norm, and someone had to fill the seats. In a market where Citadel Securities alone executes about forty-one percent of all retail order flow, the Financial Times’ markets desk called the structure the “biggest bagholder exercise of all time.” Two decades of IPO research says roughly the same thing in footnotes: offerings engineered around retail sentiment systematically underperform, and the sentiment, made up almost entirely by the everyman, pays for their exit.

The index mandate was not the first funnel built. Just the widest. Before the IPO, before a single share had legally changed hands, synthetic perpetual futures had been trading SpaceX on Hyperliquid and Binance since mid-May. A leveraged, borderless, twenty-four-hour derivative market that opened at $150 and spiked to $216 within hours, pricing the stock sixty percent above the offering on pure speculation. By the time the IPO arrived, $2.7 billion had already been wagered on a company that did not yet have a ticker symbol. That has never happened at this scale. When the IPO actually hit, Bybit, Bitget, and Binance cancelled their tokenized SpaceX allocations, citing a share shortage. The traders who had bought synthetic exposure were left holding receipts for a product that never shipped.

On Polymarket, thirty-three separate markets let anyone bet on the trajectory. The Tesla-SpaceX merger alone carried $800,000 in volume at 26 percent implied probability. The loop runs one direction: Polymarket odds feed CNBC segments, the segments inform the next wave of buying, the buying moves the odds. CME Group announced SpaceX single-stock futures for July 27. Every week a new instrument. Every instrument a new on-ramp. Every on-ramp a wider funnel.

The stock did what small-float stories always do. It touched $225.64 four days after listing. It hit $147.11 eleven days later. It opened near $160 on inclusion day and closed at $149, down seven percent on the day the mandatory bid arrived. The forced buying had been front-run: most of the passive flow settled at or before the open, and the price fell through it for the rest of the session. A thirty-five percent swing from peak to trough inside three weeks, on a float of three to five percent of shares outstanding. That kind of volatility on a $2 trillion company has no precedent in index history. By one analysis, index investors are now buying at more than a hundred times the company’s 2025 revenue. A valuation multiple never seen in the history of the stock market. Ten times what is considered typical.

Morningstar’s analysts, working from the S-1 itself, value the company at $780 billion, fifty-five percent below the IPO target. Their probability-weighted value for the AI segment is $170 billion, and they assign a forty-three percent chance the segment is eventually shelved altogether. Their moonshot case, the one where everything works, gets seven percent.

Governance does not soften the math. Musk holds forty-two percent of the equity and eighty-five percent of the voting power. He serves simultaneously as CEO, CTO, and chairman, and the company claims the controlled-company exemption that waives the requirement for an independent board majority. No company with this concentration of control has ever entered a major index at this valuation. AkademikerPension, a $25 billion Danish fund, blacklisted the stock outright, calling it “grossly overvalued” and describing governance failures severe enough that it couldn’t justify holding at any price above $1 trillion. The exclusion covers even its indexed products. The fund concluded that the index itself was the risk. A Dutch transport-sector pension is weighing the same exit.

The eighty-five percent also explains what the cap table looks like when someone finally gets to see it. ProPublica reported in June, from Delaware court records the outlet had to go to court to unseal, that at least a dozen investors with addresses in mainland China, Hong Kong, and Russia acquired pre-IPO stakes between 2018 and 2021. The money was routed through Cayman Islands vehicles by a middleman fund called Tomales Bay Capital. One was a Beijing venture investor whose firm partners with China’s Ministry of Science and Technology and whose portfolio includes companies later sanctioned over support for the Wagner Group. When a $50 million Chinese position drew press attention, SpaceX’s own CFO flagged the national-security review risk and the investor was quietly expelled in a dispute that ended in Delaware’s Court of Chancery. SpaceX barred Chinese buyers from the June IPO. The stakes already inside stayed inside. This is a company that builds spy satellites for the Pentagon.

Then there is the calendar. Twenty percent of insider shares unlock after the August 6 earnings report, more if the stock trades thirty percent above the offer price for five days out of ten. In total, 912 million shares come unlocked by December. That is seventeen times the entire float trading today. The people who built the company, who know its margins and its trajectory better than any analyst or index committee, will be legally permitted to sell into the same market that your index fund is legally required to buy from. The exit and the entrance share a door, and only one side gets to choose when it opens, and only they get to go through it.

What has no precedent is the completeness. Two pools were tapped simultaneously. The voluntary pool captured everyone who chose to buy: brokerage allocations, fractional shares on Robinhood and SoFi, synthetic perpetuals on crypto exchanges, prediction-market wagers. The involuntary pool captured everyone who didn’t: index inclusion, target-date funds, 401(k) defaults. Every financial mania in history left one door open. This one closed both and put the second on a timer. The involuntary channel refills on a schedule, biweekly, by payroll deferral, with no opt-out short of leaving the index entirely. Your next contribution is already on the calendar. So is the next one after that.

None of this means SpaceX is worth nothing. That is the wrong story, and it is the strawman that will be used against the right one. The launch business is real and dominant. Starlink is real and growing, ten million subscribers and counting. Morningstar’s $780 billion is not an insult. It would make SpaceX one of the most valuable companies on earth on the merits. The floor exists.

The problem is the distance between the floor and the ceiling, and who is standing in the gap.

An AI valuation is a bet multiplied across dimensions: compute, capital, distribution, and the people who turn the first three into products. Multiplication has one property addition doesn’t. A zero anywhere collapses the product. The compute is real. The capital is real. The distribution is real. The team who was responsible for keeping these other dimensions valuable are gone. And the company’s answer to the missing variables who drive other primary variables was to print $60 billion in stock and buy a code editor whose own users were already leaving. Not because the new shiny toy fixed the zero. But because the purchase has the shape of momentum, and momentum is what holds a ceiling up when the floor is $1.3 trillion below it.

That gap, if you take Morningstar’s number, would itself rank among the ten most valuable companies on earth. It is not held by hedge funds or venture capitalists who signed up for the ride. It is held by QQQ, by target-date funds, by state pensions, by every 401(k) with a growth allocation and a default setting. It is held, as of 8 a.m. this morning, by you.

The Danish fund could walk. It read the filings, named the risk, and left, even paying the tracking-error cost of excluding an index heavyweight from its portfolio. Your index fund cannot walk. Its mandate is the index, and as of this morning, the index is the story. The one protective layer that couldn’t be lobbied, no-actioned, or fast-tracked was an asset manager willing to say no. The only one that said it out loud manages Danish academics’ pensions. Not yours.

August 6 is the first unlock. The sellers will be the people who know the company best. The people who have the most observable position of the $1.3T distance. The standing bid will be the funds that bought this morning and will buy again with every paycheck cycle after. The most patient exit liquidity ever thoughtfully designed, refilled biweekly by payroll deferral. Insiders sell when they designed — earlier than you. And your fund buys because — it’s the rule. The rule they designed.

The next part is not comfortable arithmetic.

The best case is the utility repricing. SpaceX loses the AI narrative. The market prices it for what it provably does: launches, Starlink, data-center leases. Morningstar’s floor. $780 billion. That is a sixty-three percent decline from where index funds started buying this morning. In the best case, more than $1.3 trillion evaporates from the accounts of people who never asked to hold this stock in the first place.

What does sixty-three percent feel like in a retirement account? The median 401(k) balance for someone between twenty-five and thirty-four is about $14,000. After a sixty-three percent repricing across the index, that is $5,200. Three months of rent in most American cities. Gone. Not spent. Not lost to bad decisions. Deducted by a mechanism that was assembled for them while they were at work. For someone between fifty-five and sixty-four, the median is about $120,000. After repricing, that is $44,000. Retirement delayed by a decade. Maybe canceled. These are people who did what they were told: contribute to your 401(k), pick a target-date fund, don’t try to beat the market. They followed the instructions, and the instructions led here. The results are not mistakes. They’re meant to look like natural market forces. These are designs. By the same people who gave you the instructions.

That is the best case. It gets worse.

SpaceX is not the only company whose valuation is loaded with AI narrative. NVIDIA, Microsoft, Google, Meta, Amazon. These are not fringe positions. They are the top six holdings of every major index fund in the country. Their combined AI premium, the portion of their valuation that depends on the market believing artificial intelligence will justify the multiple, runs somewhere between five and eight trillion dollars. If the narrative cracks for one, every analyst who wrote “AI revenue will justify this price” has to rerun the model simultaneously. Not one stock repricing. The story repricing. The top of every index repricing at the same time.

And index funds cannot pick and choose. When redemptions accelerate, when people see their balances dropping and panic-switch to bonds or cash, the fund has to sell its holdings to cover the outflows. It sells the falling stocks at falling prices. The selling pushes prices lower. Lower prices trigger more redemptions. More redemptions force more selling. This is not a theory. It is the reflexive loop that turned a crack in the housing market into a global financial crisis in 2008. The only difference is that in 2008, the toxic asset was a subprime mortgage buried inside a collateralized debt obligation. This time, the toxic asset is the AI narrative, and it is sitting at the top of every retirement account in the country, in plain sight.

The contagion does not stay in equities. Banks hold corporate bonds issued by the same companies whose stock is falling. AI companies borrowed heavily to build data centers, and those debt instruments reprice when the equity collapses. If the government backstops the failures, more Treasury issuance drives yields up and bond prices down. Banks holding bonds as safe assets take mark-to-market losses. This is how Silicon Valley Bank collapsed in 2023. One bank, one duration mismatch, one weekend. Now scale that across the sector, simultaneously, with the same trigger.

When banks take losses, they tighten lending. Companies that depend on credit lines cannot roll their debt. Data-center construction halts. Chip orders get cancelled. The supply chain that was built to serve AI demand, the construction workers, the electricians, the fabrication plants, the logistics networks, sees contracts evaporate. This is not tech workers losing their jobs. This is the real economy falling on its ass because the financial economy told a story that turned out to be too expensive to be true.

And then the freeze. This is not your grandpa’s bank run where people line up at the building. Your money is already digital. When the system locks, your card declines at the grocery store. Your Venmo balance shows zero. Your bank app says “temporarily unavailable.” The FDIC insurance fund holds roughly $128 billion. That does not cover multiple large-bank failures simultaneously. People do not sit quietly through that. In 1933, they lined up around the block. In 2026, they break down the door. History is very clear about what a population does when it discovers the floor was never there.

If the dollar weakens materially, the damage stops being domestic. The dollar is the world’s reserve currency because international markets trust US bonds and US institutions. A crack in that trust is not a theoretical risk for America’s adversaries. It is a strategic opening they have been positioning for since 2015. China’s yuan internationalization, digital yuan infrastructure, and record gold purchases are not hedges. They are the contingency plan. Saudi Arabia has already signaled willingness to price oil in yuan. If the dollar weakens and that signal becomes policy, global demand for dollars drops, which weakens the dollar further. Another reflexive loop with no exit. US imports become more expensive. Inflation spikes. The Fed is trapped: raise rates to defend the currency and deepen the recession, or cut rates to stimulate and watch the dollar fall. There is no good move. That is the textbook definition of a dominated position.

And beneath all of this sits a coincidence that becomes harder to ignore the longer you look at it. The same people positioned at the center of this structure, the ones building the sky-high AI valuations, running the companies, lobbying for the rule changes, are also, separately but consistently, building bunkers. Elaborate, multi-million-dollar, off-grid survival infrastructure. Musk in Texas. Thiel in New Zealand. Zuckerberg in Kauai. That is not a hobby. That is a revealed preference. People do not pour fortunes into doomsday architecture because they believe the system they are profiting from is going to hold. They do it because they have seen the math from their side of the table, and they have decided that the appropriate response is a concrete room with a generator and a dedicated water supply.

It’s July 7, 2027. 8 a.m. No coffee. The stores ran out. No news. Cell service hasn’t worked for months. Everything has been pulled from underneath you. The rug is all you have left.

By now you are probably asking the obvious question. The people who arranged this, the rule-writers, the underwriters, the sellers — they’re not stupid. They can read a lockup calendar and a loss ratio as well as anyone. They already know how the math resolves. They have seen the floor. They have seen the gap. They have seen the calendar. They coordinated and built them for Christ sakes.

So why build it?

If the old financial system loses credibility; you have the opportunity to offer a new one. If you have one already designed, the collapse is not your worst case. It’s your debut. If people lose faith in the dollar, and you already control the alternative rails, the panic is not your problem. It is your customer acquisition strategy. And if the population trying to organize a response has spent the last decade being algorithmically surveilled and sorted by some of the same folks that designed the rug pull, the resistance is not a threat; it’s catalogued, rich-data set that makes actionable consequences a click of a button.

This is the ‘what’ and ‘how.’ The ‘why’ is next.