London, 26 July 2026 — EBM Weekend Read — By Brad Adams
On 12 June 2026, Space Exploration Technologies began trading on the Nasdaq under the ticker SPCX, having priced 555.6 million shares at $135 the night before. It raised roughly $75bn, rising to $85.7bn once underwriters exercised the overallotment in full — the largest initial public offering ever completed, close to triple Saudi Aramco’s 2019 record. The valuation was about $1.75 trillion. Four months earlier, the company had bought xAI in an all-stock deal that Musk was on both sides of, and the price he set for that transaction is now embedded in what public investors paid.
The argument for years was that this would all be tested at the IPO — that the moment SpaceX had to file a prospectus, outside investors would finally see how the pieces had been priced. That test has now happened, and it produced a genuine surprise. What the S-1 disclosed was not the money-printing machine the market had assumed. It was a company reporting a net loss of $4.94bn for 2025, and the reason was xAI.
Join The European Business Briefing
New subscribers this quarter are entered into a draw to win a Rolex Submariner. Join 40,000+ founders, investors and executives who read EBM every day.
SubscribeWho owns what, and what changed
Start with the pieces, because they are still not held equally.
Tesla is a public company with ordinary shareholders, quarterly earnings and fiduciary duties. Musk’s stake stood at around 13% for several years, but the Delaware Supreme Court reinstated his 2018 pay package in December 2025, and he exercised the 303.9 million options in June. He now holds roughly 20%.
SpaceX is no longer private. It carries a dual-class structure: Musk holds about 42% of the equity and roughly 93.6% of the Class B shares, which carry ten votes each against one for the Class A stock sold to the public. That gives him about 82% of the votes. Public shareholders bought economic exposure and almost no governance.
xAI has been a SpaceX segment since 2 February 2026. On the S-1 numbers it lost $6.36bn at operating level in 2025.
X, the former Twitter, sits inside xAI. Musk took it private in 2022 for $44bn, loaded it with debt, and watched its value collapse.
The ownership gradient is still the story. Musk owns roughly a fifth of the company with public accountability, and more than twice that proportion of the one where he controls four fifths of the votes. Value that moves in that direction moves toward the pocket where his stake is largest.
The X trick: from $9bn to $45bn in five months
The clearest illustration remains what happened to Twitter.
By late 2024, X was a wreck. Revenue had roughly halved under Musk, and Fidelity, one of his co-investors, had marked its stake down heavily — implying a valuation in the region of $9bn to $10bn against the $44bn he paid.
Then in March 2025 xAI acquired X in an all-stock deal valuing it at $33bn, or $45bn including debt. Overnight, an asset carried at around $9bn was worth $45bn again.
Nothing had changed at X to justify it. The number did not come from improved financials. It came from Musk deciding that xAI would pay $45bn in its own shares — shares in a private company whose value he also set. The figure landed almost exactly $1bn above the original Twitter price, which allowed Musk to tell his Twitter co-investors they had been made whole. The people made whole were his own backers. The currency was equity in another company he controls.
He barely hid it. “xAI and X’s futures are intertwined,” he wrote, announcing that he had sold one of his companies to another of his companies at a price he chose.
Tesla’s balance sheet enters the loop
Here is where public shareholders come in, and where the sequence gets uncomfortable.
In November 2025, Tesla shareholders were asked in a non-binding vote to authorise an investment of up to $5bn in xAI. It failed. About 1.06 billion shares voted in favour, but 473 million abstentions counted as “no” under Tesla’s rules, taking the against total to 1.39 billion.
On 16 January 2026, Tesla invested $2bn in xAI anyway — structured as Series E Preferred Stock under a separate framework agreement rather than under the authorisation shareholders had declined. Tesla disclosed it on 28 January and said the terms matched those agreed by independent investors in the same round.
Seventeen days later, SpaceX absorbed xAI.
To be precise about what followed, because it matters: Tesla was not left holding nothing. Its xAI position converted into an indirect stake in the merged entity, and in March the FTC cleared the conversion into a direct holding of under 1% of SpaceX. Tesla shareholders ended up owning a sliver of a company they were never asked about, having declined to invest in the one they were.
The commercial traffic runs the same way. Tesla disclosed $143m of vehicle sales to SpaceX on 1 May, and has supplied Megapack grid batteries to power xAI’s data centres. Real transactions, on terms set inside a single ecosystem.
Why doing it privately was the point
None of this is illegal on its face, and it is worth being precise about that.
Delaware has looked at Musk self-dealing before without always siding with shareholders. When Tesla bought SolarCity in 2016 — another Musk-controlled company, which shareholders called a bailout — the court ultimately found the deal fair, though other directors settled for $60m without admitting fault.
But the SolarCity fight is exactly why the structure was built this way. By completing the SpaceX–xAI merger while both companies were private, Musk could set the relative valuations, negotiate with himself, close the deal, and inform investors afterwards — without the disclosure obligations and procedural checks a public-company merger drags along.
The Tornetta case shows the friction he avoided, and how it ends. The Court of Chancery rescinded Musk’s $51.4bn Tesla pay package in 2024 precisely because the approval process was not independent or arm’s-length. That is what a public company carries. It is also, as of December 2025, a decision the Delaware Supreme Court reversed — the package was reinstated and Musk exercised it in June. Even the friction eventually gave way.
This is the same lesson EBM keeps meeting from the other direction. We have written that Rolex’s foundation ownership is its sharpest competitive weapon, because answering to no shareholder confers enormous freedom. The Musk empire is the shadow side of the same coin: when the controller also runs a public company alongside the private ones, that freedom can be exercised at the public shareholders’ expense.
The reckoning arrived, and it was in the prospectus
For most of a decade, what the market knew about SpaceX’s finances came from leaks and secondary trading. In January, Reuters reported roughly $8bn of EBITDA on $15bn to $16bn of revenue. That described a company printing money, and it was repeated everywhere, including here.
The S-1 filed on 20 May described something else. Consolidated 2025 revenue was $18.67bn, higher than estimated — but only because February’s xAI acquisition pulled a new business line into the accounts. Consolidated operating loss was $2.6bn. Net loss was $4.94bn. Accumulated losses since 2002 came to $41.3bn.
Underneath, the split is stark. Starlink generated $11.39bn of revenue and $4.42bn of operating profit, a 63% EBITDA margin, with subscribers up from 4.6 million at the end of 2024 to more than 10.3 million by March. It is a genuinely excellent business.
xAI lost $6.36bn at operating level.
So the acquisition Musk priced on both sides is what turned a profitable rocket-and-satellite company into a loss-making one, in the accounts investors were shown weeks before being asked to buy. That is not a hidden fact — it is disclosed, in the prospectus, as the rules require. It is simply a fact that only became visible once the private walls came down, which is precisely what “deferred to the IPO” always meant.
And the pension money had no choice
Here is the part that makes this everybody’s problem rather than a governance curiosity.
SpaceX floated only about 4% of its stock. On 9 June, MSCI confirmed it would apply its fast-track treatment for large IPOs, adding SPCX to its Global Standard indices from 13 June — the day after listing.
Index funds do not form a view. They buy what the index says, in the weight the index specifies, at whatever price the market sets. A tiny float plus mandatory inclusion produces structural demand that has nothing to do with whether $1.75 trillion was a sensible number for a company that lost $4.94bn.
So European pension savers whose funds track global indices bought into this. Not because a manager assessed the xAI valuation Musk set on both sides of a deal, but because a rule required it.
The price action since is its own commentary. SPCX opened near $150, closed its first day at $161, ran to an intraday high of $225.64 on 16 June, and has since fallen back through its own offer price.
The bottom line
Vertical integration is the official story, and there is something real in it. Rockets, satellites, data, compute and a social platform under one roof is a coherent industrial idea, and Musk has a record of making coherent industrial ideas work. Starlink’s numbers are not an illusion.
But strip away the mission language and a simpler mechanism is visible. Assets move between entities Musk controls, at prices Musk is uniquely positioned to set, in the direction where he owns the most. It is the same shape now visible across the AI build-out generally — Nvidia backstopping the leases on data centres that will buy its chips, Google’s AI spending pushing its free cash flow negative — where capital increasingly circulates between the same handful of balance sheets rather than arriving from outside.
The old argument was that the reckoning was postponed to the IPO. It was, and it came. What nobody quite anticipated was that the market would absorb it and buy anyway, because the AI trade rewards conviction over arithmetic and because index rules do the buying regardless.
Anyone with a global tracker in their pension is now inside the loop. They did not choose it, they were not asked, and their vote is worth a tenth of his.
Related Analysis
