You're Not Buying Part Of SpaceX, You're Refinancing It

Funny thing is, nobody believes SpaceX is an "investment." The only people buying it are (a) delusional or (b) playing a greater-fool theory with the hopes they can offload it on some bigger sucker. Nonetheless, it feels prudent to analyze it as a real investment.

It is not an investment, it is a liquidity event for insiders that the public is funding, whether directly or indirectly through mutual funds and ETFs. Let's go through SpaceX's own registration statement (Amendment No. 2, filed June 3, 2026) and its June 5 free writing prospectus.

$135 for $7.85 of book value

SpaceX is selling 555,555,555 Class A shares at $135.00 to generate about $75 billion gross (which instantly becomes $55 billion, but more on that in a second), by far the largest IPO raise in history. The IPO never states a valuation, but ~13.1 billion pro forma shares multiplied by $135 = $1.8 trillion. This number is not real; unlike, say, Apple or Google, where every stockholder really could sell their stock and someone would come along and buy it for roughly the same price, thereby raising the full value of the company back to where it was, nothing like that could possibly happen for SpaceX. SpaceX's $1.8 trillion "valuation" or "market cap" is manufactured by its tiny public float representing just over 4% of the actual shares.

As the dilution tables on page 71 of the Prospectus lay out. IPO subscribes are paying $135 per share for shares with a pro forma net tangible book value after the offering of $7.85 per share, an immediate $127.15 per share (94%) dilution. By the company's own arithmetic, those new suckers/investors will contribute 48% of all the capital ever put into SpaceX in exchange for just 4.2% of the shares.

The worst terms in IPO history, and what it would take to outrun them

While 94% dilution may seem shockingly high, it isn't far off the historical precedents of other landmark mega-IPOs:

The capital-for-equity swap has precedent too. Alibaba's new investors put up 56% of lifetime capital for 5% of shares, and Snowflake's put up 63.6% for 10.1%. The structure isn't novel, but that's the most concerning part for a would-be buy-and-hold SpaceX investor: we have a track record of how these deals turn out, and it depends entirely on one thing.

The companies that made extreme dilution irrelevant did it by rapidly converting hypergrowth into actual earnings. Google went from $3.2 billion of revenue at IPO to $21.8 billion in five years, and it was profitable every single year, so its multiple compressed gracefully while the stock compounded roughly 25% annually. Facebook grew revenue to $27.6 billion and net income to $10.2 billion by year five. The shareholders of companies that scaled revenue without reaching profitability became bagholders. Snowflake grew revenue sixfold in five years and still lost money for first-day IPO buyers; Rivian grew revenue a hundredfold and compounded at roughly negative 26% annually. The market's rule is brutally consistent: reach (or visibly approach) real net income within five years, or your multiple collapses no matter how big your story is.

Let's see how that applies to SpaceX. It starts out grossly overvalued even for a growth company, $1.8 trillion is roughly 95 times trailing revenue ($18.7 billion in FY 2025). It's earnings are in shambles: it lost $4.9 billion last year and $4.3 billion in the first quarter of 2026 alone.

While the company claimed profitability in 2024, the reality is different. FY 2024's $791 million of net income included a $659 million non-cash tax benefit from releasing a valuation allowance, an accounting entry that was fully reversed in 2025. Operating income tells the honest story: $466 million in 2024, negative $2.6 billion in 2025, and negative $1.9 billion in Q1 2026. SpaceX has never produced a genuinely profitable year in the periods presented by its own IPO.

At ~$1.8 trillion, even a generous 30x earnings multiple requires roughly $60 billion of annual net income just to hold the IPO price flat, a ~$65 billion upward swing from the $4.9 billion it lost last year. To actually beat an index fund over five years, it would require materially more. SpaceX's ~95x trailing revenue doesn't look like Google's IPO (~7x sales) or even Facebook's (~20x), it looks like Snowflake's revenue bolted onto Rivian's capital intensity.

And where would $60 billion of earnings come from anyway? Every disclosed trend points the wrong way: net losses are widening ($4.9B in 2025; $4.3B in Q1 2026 alone), Starlink is the SpaceX's best segment but its average revenue per unit is down 33% and falling as they have to discount heavily to get new customers, customer launch revenue is down 42%, and AI advertising (i.e., "X," the go-to site for Nazis) is down 23%. Nobody wants to pay for xAI's crappy "MechaHitler" CSAM-generator, so xAI is already leasing GPUs out to competitors like Anthropic and Google.

The clearest demand signal of all is that SpaceX's AI segment, the destination for most of your capital, has so little internal use for its GPUs that it's subletting them to competitors like Anthropic and Google, both of whom can cancel with 90 days' notice if the trend to cram AI into everything slows down. Companies with real AI demand hoard compute. Companies with crappy "MechaHitler" CSAM-generating AI that nobody wants to pay for lease it out and call it a cloud business.

The first $20 billion instantly vanishes

In March 2026, SpaceX drew down a $20 billion bridge loan, largely to cover up his prior mistakes, like buying Twitter. Doing so triggered a $1.5 billion loss on debt extinguishment and over $1.1 billion of cash prepayment premiums, pushing quarterly interest expense up 48.5% year over year. Why would any company eat $2.6 billion in avoidable losses? Simple: to yank money out of a company they knew wasn't going anywhere. In the same quarter, the company spent $4.35 billion buying back stock from employees and existing shareholders.

The bridge loan's credit agreement requires a 100% mandatory sweep of net IPO proceeds to repay it within six months. The first ~$20 billion of the money investors hand over at $135 per share goes directly to retiring the debt that funded other people's exits and Musk's bad judgment.

Cumulatively, this is a pattern rather than an isolated incident. Roughly $6.7 billion of company cash went to share repurchases from 2023 through Q1 2026, the overwhelming majority of it in the final quarters before the IPO. Total identifiable cash flowing to insider-related parties over that period is about $8 billion.

The insider buffet

The related-party section of this S-1 reads like a private chat group with wiring instructions.

The board's relationship with Valor Equity Partners is particularly concerning. Antonio Gracias, founder and CEO of Valor Equity Partners, sits on SpaceX's board and beneficially owns 7.3% of the Class A shares. SpaceX (through its xAI subsidiaries) has signed three equipment lease deals with Valor entities totaling $20.2 billion in undiscounted future payments. These are accounted for as financing obligations (debt, essentially), and the closest benchmark the filing offers for what such financing costs is its disclosed 22.6% weighted-average finance-lease discount rate, against a 4.4% blended rate on the company's ordinary equipment financing book. A board member's firm is providing hardware financing at an implied cost in the neighborhood of a credit card's, but the filing discloses no fairness opinion, no competing bids, and no independent committee process. The cash is already moving: Valor lease payments ran $1.9 billion in just the four months from January through April 2026, roughly $479 million a month, accelerating into the offering.

Transactions with Tesla represent another massive outflow. SpaceX and its xAI subsidiaries have purchased roughly $1 billion of goods and services from Tesla (about 20% beneficially owned by Mr. Musk) since 2023, including Megapack batteries, Cybertrucks, and licensing. In April 2026 alone, while the SpaceX IPO was being marketed, the company purchased $269 million of Megapacks. The S-1 calls these arm's-length transactions but offers no benchmark.

The business investors would actually own

Strip out the narrative and SpaceX is three distinct businesses: one is good and getting worse, one is being cannibalized, and one is a furnace.

• Starlink (Connectivity) is the strong performer, reporting $11.4 billion of FY 2025 revenue and $4.4 billion of operating income. But the unit economics are visibly decaying. ARPU has fallen from $99/month in 2023 to $66/month in Q1 2026, down 23% year over year and accelerating, exactly as the company says it expects as growth shifts to lower-income markets. Q1 2026 subscribers grew ~105%, but revenue grew only 31.6%. Starlink has run out of potential subscribers willing to pay premium rates and is now starting to compete with other forms of internet access, a battle it cannot win, not profitably.
• Launch (Space) is being eaten by its sibling. Customer launches collapsed from 43 in FY 2025 to 7 in Q1 2026, with 82.5% of the quarter's Falcon launches carrying SpaceX's own Starlink satellites instead of paying customers. Launch Services revenue fell 41.7% year over year. Meanwhile, Starship has produced zero customer revenue twelve flight tests in and burns over $900 million of R&D per quarter. The segment lost $662 million in Q1 alone.
• AI (xAI/X) is even worse: the segment lost $6.4 billion in FY 2025 and $2.5 billion in Q1 2026. Capex exploded from $463 million in 2023 to $12.7 billion in 2025, reaching $7.7 billion in Q1 2026 alone, a ~$31 billion annualized pace. This can't be excused as growth into the lucrative AI business because, as shown by the Anthropic and Google deals, nobody wants to pay to use Grok. X offers no help either, as the advertising business underneath it shrank 22.6% year over year.

The math here is simple: the segments that make money generate a few billion a year of operating income combined that the AI segment incinerates at a ~$10 billion annual run-rate (Q1's $2.5 billion loss, annualized) and rising, with no hope of recouping any of it at a profit. The $75 billion raise doesn't fix that equation. It just buys more time.

The two contracts holding up the AI story

The bull pitch is that SpaceX has become an AI cloud provider, with Anthropic paying $1.25 billion a month and Google paying $920 million a month. These are real contracts representing about $75 billion combined if both run full term, but the termination clauses tell a different story.

The Anthropic agreement is non-terminable for only three months, after which either party can walk on 90 days' notice. Google's deal can be terminated by either party on 90 days' notice after December 31, 2026. Google can terminate even earlier, or pro-rata cut its fees, if SpaceX fails to deliver access to ~110,000 GPUs by a hard September 30, 2026 deadline. Neither the S-1 nor the FWP tells us whether SpaceX even owns that hardware yet or what it will cost to procure.

The AI segment's only material third-party revenue is two contracts with two sophisticated counterparties who build their own data centers for a living, both cancellable on a quarter's notice, and who are rapidly developing AI that real customers actually consider using. Even if both run to the end, those contracts cover roughly 22% of the segment's likely cumulative five-year cost. A revenue stream your tenant can cancel by Christmas is not a foundation for a $1.8 trillion valuation.

Furthermore, the Google contract, signed June 5, isn't in the registration statement at all. It arrived via free writing prospectus two days after Amendment No. 2 was filed. While legal, omitting this from the main registration statement raises serious questions about disclosure hygiene for the largest IPO in history.

Investors will own 4.2% of the equity and 0% of the say

Post-offering, Mr. Musk controls 82.4% of the voting power through 10-vote Class B shares. SpaceX will be a "controlled company" under Nasdaq rules, formally exempt from independent director majorities and independent compensation committees. The same governance environment that produced the Valor leases, the Tesla purchases, and the milestone mega-grant will continue, except now with public investors' money in it. There is no activist coming, no proxy fight to win, no circumstance under which this company is anything other than what Musk (or his creditors) want it to be.

And we already know who isn't planning on holding long-term: the lock-up is front-loaded. Up to ~911 million insider shares unlock on the second trading day after the first earnings release, and another ~456 million unlock if the stock holds 30% above the offer price into that date. The insiders aren't planning on buying-and-holding, so why should anyone else?

The bottom line

Every buy-and-hold investment is built upon a good price and a likelihood of improving economics for the company. SpaceX's own filing shows deteriorating unit economics in its best segment, a structural loss-maker absorbing all the capital, a 95x-trailing-revenue price tag with 94% day-one dilution, and a governance structure deliberately built so that public shareholders can never do anything about any of it.

SpaceX may well keep launching rockets and building data centers, but the question is whether this security, at this price, with these insiders, under this structure makes sense for anyone to invest in. The prospectus makes clear the answer is "no."

The author is not your financial advisor, and nothing here is investment advice.