SpaceX did the big thing first. It sold 555.6 million shares at $135 and raised $75 billion, an all-primary deal big enough to reset the IPO record by itself. [E1] The tape then did the louder thing: $150 open, $160.95 close, up 19%, about $2.1 trillion of equity value, with orders reportedly above $250 billion. [E2]
That is the easy paragraph. The harder one is what buyers actually bought. Public holders own low-vote Class A stock, while Musk's high-vote structure leaves him with 85.1% voting power; Reuters' index math puts the free float around 3% to 4%. [E3][E4] This is public capital with private-company control. The steering wheel did not come with the shares.
S&P noticed. It declined to waive the normal gates, because SpaceX meets none of the fast-entry tests that matter here: a year of public trading, GAAP profitability and at least 10% float. [E4] Nasdaq is looser; its May FAQ created a top-40 fast-entry path after 15 trading days and removed the old minimum float requirement, while capping low-float weights. [E5] So the next argument is not just valuation. It is which index owns the plumbing.
Now the better story: SpaceX is selling compute capacity like a landlord sells floors. Google agreed to pay $920 million a month from October 2026 through June 2029 for about 110,000 Nvidia GPUs and related hardware, with delivery and termination clauses that matter. [E6] Reuters also reported an Anthropic arrangement at $1.25 billion a month through May 2029, but Musk later said the deal was only a six-month lease with a 90-day mutual off-ramp. [E7][E8] Annualize the headline and you get a huge number. Discount it and you still get the point.
xAI put the dream in the official wrapper: Anthropic had expressed interest in "multiple gigawatts of orbital AI compute capacity." [E9] Translation: the rocket company wants public-market capital for a stack that starts with Starlink, runs through government launch and ends somewhere between a data center and a sovereign utility.
The existing machine is not imaginary. Reuters' IPO charts put 2025 revenue at $18.67 billion, Starlink at about 60% of it, users around 10.3 million, satellites around 9,600 and launches at more than two a week. [E10] Washington Technology, reading the S-1, says about one-fifth of sales came from the US government and describes Starshield work including target tracking, reconnaissance and missile warning. [E11] That is why this is a macro story. Public investors are financing communications, defense access and AI compute under one concentrated vote.
The bull case is real infrastructure scarcity. The bear case is that some of the new revenue is cancellable, the margins are not public yet, and the customers are enormous enough to become the risk. Google and Anthropic are not cute logos in a pitch deck. They are customer concentration with better lawyers.
Traders are mostly arguing supply first. On X, the cleanest version of the bull case was that "near term, flows may matter more than fundamentals"; [E12] the weekend was full of low-float index math and June 24 lockup chatter. [E13] So the call is SPCX at or above $150 by June 24: the float is too tight for valuation gravity to work quickly, but S&P’s no means the biggest automatic buyer is not coming yet.
That is the weird bargain: buy a public stock, fund quasi-sovereign infrastructure, wait for index committees, and hope the landlord lets you see the lease.