The Landlord, The Exit, and The Ghost of Houston

The Anthropic deal was not a business story. It was the final piece of a very carefully constructed IPO.

By E. J.

Disclaimer: All analysis here is speculation and speculation only. Written for fun and intellectual curiosity. Not financial advice. Not an accusation of anything. Just someone connecting dots and asking questions when there is nothing better to do.

Part 1: The Landlord

On May 7, Anthropic announced it would rent the entire compute capacity of Colossus 1, the Memphis data center built by xAI, from SpaceX.

220,000 Nvidia GPUs. 300 megawatts of power. It is certainly something much needed for Anthropic at this moment. Its users (including the very important me), are complaining about reaching limit with one line of instructions. Before this moment, I thought I'll have to wait for the company to go public to solve the issue. But here it emerges. No warning, no sign.

This deal is also good news for Musk. Some estimates put the deal at $3 to $4 billion in annual revenue for SpaceX, though the contract economics have not been publicly disclosed. If those estimates are close, the deal gives SpaceX something IPO investors love to see: recurring infrastructure revenue.

Silicon Valley called it a pragmatic match. Musk has compute Grok can't fill. Anthropic has users and not enough compute. Simple supply and demand. Two adults making a sensible arrangement.

well, Y'all know me. Let's slow down.

Three things happened in the same week. Each one, on its own, is a business story. Together, they are something else entirely.

One: Anthropic signed the compute deal with SpaceX.

Two: Musk announced xAI would be dissolved as a separate company and absorbed into SpaceX, now renamed SpaceXAI.

Three: SpaceX's roadshow is scheduled to begin the week of June 8.

OK I lied, there is a Three Point Five: Musk is seeing Altman in Oakland.

The timing is not a coincidence. It is the central theme. But before we go deeper, please join me to congratulate the newly upgraded landlord.

Part 2: The Numbers That Disappeared

If you read the previous piece on Cursor, you already know what xAI's numbers looked like. Let me remind you anyway.

xAI ran a $6.4 billion operating loss in 2025. It consumed 61% of SpaceX's total capital expenditure that year. The consolidated entity — SpaceX plus xAI — posted a $4.9 billion overall loss after the merger. Grok generated less than $1 billion in annualized revenue. Musk himself publicly admitted xAI "wasn't built right the first time." Several co-founders had already left.

This was the number that would have greeted IPO investors in June. But now, the number became part of a larger consolidated story.

xAI is no longer a separate company with a visible burn rate. It is now a department. Its losses are absorbed into a larger balance sheet. Its estimated $1 billion monthly cash consumption could become harder for casual investors to isolate within the larger consolidated presentation. The departed co-founders quickly downgraded to departed department leaders. Sort of senior director level at most. And the same hardware that was burning money to train Grok? It is now a revenue-generating cloud infrastructure asset with Anthropic as its anchor tenant.

The liability became an asset. The cost center became a business line. The people issue became remote past stories. The problem became the solution, the pitch.

All before the S-1 goes public.

Musk is indeed, the genius of our time, the magician brings magic.

Part 3: 94 Times

Let's talk about the valuation. SpaceX is targeting $1.75 trillion.

Starlink is the only business that makes money. The consolidated entity posted $18.7 billion in revenue. xAI alone burned $12 billion of it. The launch business added $4.1 billion in revenue. What it cost is not broken out. And xAI's burn adds to that opacity. How much of its $12.7 billion CapEx bleeds into shared infrastructure is anyone's guess.

Disclosure: well, I'm not an accountant and I spent about 10 min on the S1. So feel free to correct me if more numbers are available to construct a more detailed picture. my point is simple: the entire revenue of the consolidated entity last year was only $6 billion higher than what xAI burned as an independent company. Just keep that in mind while you continue to read.

At $1.75 trillion, investors will be paying approximately 94 times revenue. Roughly 220 times EBITDA, if we use $8B EBITDA proxy. (Go back to last paragraph if you needed)

For context: Nvidia at peak AI euphoria traded at around 30 to 40 times revenue. Apple trades at 8 times. Google at 6 times.

To justify $1.75 trillion on fundamentals, investors need to believe SpaceX will grow revenue tenfold, lower spending, dominate AI infrastructure, win orbital data centers, hold its defense contract monopoly, loved globally, etc. All need to happen simultaneously, all on schedule.

That is not a valuation. That is a prayer.

A prayer works when everyone in the room needs it to come true.

Which brings us to Part 4.

Part 4: The Exit

Private-market trackers estimate SpaceX has raised approximately $12 billion across 31 funding rounds from over 240 investors. Again, to put that into context, Google went public after 3 rounds. Nvidia after 2. Apple after 1. When these companies went public, they only have limited number (comparing to 240) investors that had vested interests, or, skin in the game.

The cap table is hard to verify from the outside, but the rumors are not subtle. Private-market trackers put Alphabet's 2015 investment at roughly $900 million and its current stake somewhere around 6 to 7.5%, potentially worth more than $60 billion at a $1.75 trillion valuation. Fidelity's Contrafund alone held approximately $5.8 billion in SpaceX shares as of December 2025. Founders Fund, in since 2008, may be sitting on one of the great venture markups of the last two decades. Sequoia. Andreessen Horowitz. And, after the xAI merger, Nvidia and the Qatar Investment Authority.

These are not people waiting to see how the story ends.

They already know how their story ends. They need the IPO to clear successfully so they can close the chapter.

Now look at the five senior underwriters: Goldman Sachs, Morgan Stanley, JPMorgan, Bank of America, Citigroup. These same institutions have been facilitating trades in SpaceX private shares for years. These institutions and their wealth-management networks may have extensive relationships with pre-IPO SpaceX holders. But of course, their work ethics can be trusted completely.

Their job, officially, is to price the offering fairly and protect incoming investors. Their actual incentive is to break that Champagne tower with joy.

Ohh, how much I loved Leo in The Wolf of Wall Street.

Part 5: The Thirty Percent

Here is where it gets interesting.

SpaceX CFO Bret Johnsen told the 21-bank syndicate something that has never been said at this scale before: "Retail is going to be a critical part of this and a bigger part than any IPO in history."

The plan is to allocate up to 30% of shares to retail investors. The industry standard is 5 to 10%. On a $75 billion raise, that is $22.5 billion. It will be the largest retail tranche ever assembled for a single offering.

For context, Saudi Aramco has held the title of largest IPO to date. The entire IPO in 2019 raised $25.6 billion. Aramco allocated 0.5%, or $128 million to retail. SpaceX's retail investors are getting $22.5 billion.

SpaceX is enabling average Joes to ride its uprising and become billionaires. Shame on Aramco, and Google, Apple, Nvidia... (the list virtually covers most of the tech industry)

On June 11, SpaceX plans to host 1,500 retail investors at a major event spanning the US, UK, EU, Australia, Canada, Japan, and South Korea.

Very generous.

Institutional investors usually ask hard questions. If they have no skin in the game, or they don't need the IPO to clear, they'd hire analysts. They'd stress-test S-1s. They'd notice when a $6.4 billion annual loss disappears into a merger one month before a roadshow. They'd notice when the anchor customer for the new cloud business was signed suspiciously close to pricing. They'd notice when the underwriters have existing positions. They'd write articles about it, or they'd quietly withdraw.

Retail investors see the headline.

SpaceX. Rockets. AI. Anthropic chose us over Amazon, Microsoft, and Google. Elon.

Retail is being invited to the party, so that they can hold the door open so everyone else can leave.

Part 6: The Kill Switch Nobody Is Talking About

One more thing about the Anthropic deal that did not make most headlines.

In a post on X, Musk wrote that SpaceX "reserves the right to reclaim the compute" if Anthropic's AI "engages in actions that harm humanity."

This clause was not in the formal press release. It is unclear whether it appears in the contract. But if enforceable, think about what it actually means.

Anthropic, one of the three leading AI labs in the world, is now partly dependent on infrastructure controlled by a direct competitor. If that dependency deepens — and 220,000 GPUs and 300 megawatts coming online tends to create dependency — pulling the plug becomes a civilizational lever, not just a business decision.

Musk is simultaneously suing OpenAI in federal court. He may now have infrastructure leverage, however informal, over Anthropic. And if SpaceX proceeds with one of the largest retail allocations in IPO history, any future regulatory action against the company could become politically harder, because retail investors would be exposed at unprecedented scale.

Who decides what harms humanity?

Part 7: Too Big To Question

SpaceX is real. Starlink is real. The launches are real. The defense contracts are real. This is not a company made of air.

Yet there is a difference between a real company and a correctly priced one. And there is a difference between a correctly priced company and one where the people setting the price have every incentive to set it high and every mechanism to ensure retail absorbs the risk.

The Anthropic deal was the final narrative piece. The xAI absorption cleaned the balance sheet. The 30% retail allocation ensures the demand. The 21-bank syndicate, many of whom have been trading SpaceX private shares for years, ensures the infrastructure. The June 11 retail event ensures the FOMO.

By the time the public S-1 lands, the story is already written.

A machine that runs on belief needs people who believe. And right now, 1,500 retail investors are being invited to a very exciting event in June, across six countries, to hear about rockets and AI and the future of humanity.

If you want to understand how this kind of machine works, someone wrote the manual. It is called Too Big to Fail.

I hope everyone read the footnotes and small prints.

Conclusion: The Ghost

There is a particular kind of financial architecture that only becomes visible in retrospect. The information was always there. In the S-1. In the timing. In the cap table. In the conflict disclosures buried in footnotes.

It becomes visible in retrospect because in the moment, the story is too good to question. The numbers are too big to doubt. The name is too famous to scrutinize.

History has a long memory for these moments.

I have no idea how this ends. My bet, like the Cursor piece, is that he pulls it off. At least partially. June will come. The roadshow will be spectacular. The retail event will trend globally.

What happens in month seven is a different article.

Subscribe if you want to read it.

Previous in this series: Cursor Has No Moat. SpaceX Has No Cash. Perfect Match.

And if you haven't read Tesla's Most Expensive Illusion, the pattern starts there.

Sources

Fortune, May 7 2026 — Anthropic/SpaceX compute deal analysis, revenue estimates, kill switch clause

CNBC, May 6 2026 — Anthropic SpaceX Colossus 1 announcement

Fortune, May 8 2026 — Anthropic growth figures, xAI dissolution announcement

Tom's Hardware, May 7 2026 — GPU count, compute specifications

Reuters, April 2026 — SpaceX S-1 excerpts, xAI operating loss $6.4B, consolidated net loss $4.9B, consolidated revenue $18.7B, Starlink operating income $4.42B, total capex $20.74B, xAI capex $12.7B

Reuters / CNBC, April 1 2026 — Project Apex, 21-bank syndicate, retail allocation plan, CFO Bret Johnsen quote

Alphabet / SEC 10-Q filing, Q3 2015 — $900 million SpaceX investment confirmed in quarterly filing

Investing.com / AlphaPilot, December 2025 — Alphabet 6.11% stake, current estimated value

FilingFrog, April 2026 — Fidelity Contrafund SpaceX holdings $5.8 billion as of December 2025

Wikipedia / CNBC, November 2019 — Saudi Aramco IPO prospectus, retail allocation 0.5%, total raise $25.6B

TechMarketBriefs — S-1 financial analysis, cash position, capex breakdown

Outlook Business / IndexBox, April 2026 — Retail allocation details, June 11 investor event