SpaceX IPO: What Changed in the Updated S-1
Twelve days ago I published a breakdown of SpaceX's original S-1. The conclusion was straightforward, take a look at the filing before you buy into the hype, because the numbers show a business far more complex and loss-making than the headlines suggest.
The S-1 has since been updated with four material disclosures that weren't in the original filing, here is what changed and whether it moves the needle on the original thesis.
The Anthropic Contract
The most significant new disclosure is a Cloud Services Agreement signed in May 2026 between SpaceX and Anthropic.
Anthropic agreed to pay SpaceX $1.25B per month for access to the Colossus I and II data centres through to May 2029, totaling a run rate of $15B annually and $45B over the full term.
SpaceX generated $18.7B in total revenue FY2025, a single customer is now paying $15B annually for compute capacity alone. The S-1 acknowledges the competitive issue directly, Anthropic is simultaneously a customer of SpaceX's compute infrastructure and a direct competitor to xAI's Grok model with both chasing big enterprise customers.
The 90-day termination clause is the key caveat, meaning the $45B in potential contracted revenue is not guaranteed. If the Anthropic relationship deteriorates commercially, competitively, or because Anthropic builds its own compute infrastructure, the contract terminates with a quarter's notice.
The Colossus data centre was built in 120 days, SpaceX has since compressed its build time to 66 days, dramatically faster than traditional operators constrained by land acquisition, power procurement, chip availability and cooling infrastructure.
The Q1 2026 Numbers
The updated S-1 includes Q1 2026 financials that weren't available in the original filing.
In 2024, before the xAI merger, SpaceX was profitable with $791M in net income. The merger turned a profitable launch and connectivity business into a heavy loss-maker overnight. SpaceX's AI business ran a $2.5B deficit in Q1 2026, contributing to a net loss of $4.27B on $4.69B in revenue. The Anthropic contract helps bridge this gap, but it does not fix the underlying profitability on its own.
Total capital expenditure in Q1 2026 exceeded $10B, $7.7B of that went to the AI infrastructure segment and Starlink received $1.3B.
Operating cash flow stayed positive at $1.0B in Q1, while investing outflows reached $16.7B. The company bridged the difference with financing, $7.1B raised in Q1 alone, on top of $29.1B of total principal debt. The $75B IPO raise reads less like ambition and more like necessity. The accumulated deficit of $41.3B disclosed in the original S-1 has not changed.
The Cursor Partnership
SpaceX entered into a compute-and-option agreement with Cursor in April 2026, giving it the right to acquire Cursor for $60B in newly issued stock following the IPO. The deal has a $10B floor, if SpaceX completes the acquisition the $10B counts toward the purchase price, if it walks away it owes Cursor a $10B breakup fee. The $60B acquisition would be financed in newly issued public shares, diluting every Class A shareholder who bought at IPO to fund it.
The Dilution Warning
Buried in the risk factors of the amended filing is an interesting sentence that should be read carefully: "We may issue a significant amount of equity in connection with future transactions."
Class A shares (the ones you get via the IPO), carry one vote each. Musk's Class B shares carry 10 votes each, there are also Class C shares with no voting rights currently used for executive compensation but available to fund acquisitions without diluting Musk's voting power. He can issue Class C shares to acquire companies, increasing the total share count and reducing your ownership, without his own voting power moving by a single vote. Additionally, the S-1 states "Mr. Musk will have the power to control the outcome of matters requiring shareholder approval, including election of all our directors."
The pattern is already established, in 2024/25 SpaceX raised $31.9B through equity issuance, with the S-1 flagging this will continue. The language has been widely read as preparation for a potential Tesla combination, Musk owns both companies and the structure gives him the mechanism to execute it without shareholder approval.
Morningstar's Assessment
Against the backdrop, Morningstar published an independent fair value assessment, putting it at $780B, less than half the $1.75T ask. Only Starlink is profitable and xAI is projected to burn $10B in 2026 alone. Morningstar's analyst wrote that xAI poses a "material threat of value destruction" and that its competitive position relative to OpenAI and Anthropic leaves its "economic moat indeterminate."
My View
The Anthropic contract is significant with $15B per year from a single customer, transforming the AI segment from a pure cost centre into a revenue-generating asset. SpaceX building data centres at record speed validates the AI infrastructure play in a way that wasn't clear in the original filing.
The Q1 2026 numbers make the picture harder not easier, with a $4.3B net loss in a single quarter, $10B in capex spending, $29.1B debt, $60B in ongoing acquisition deals and future dilution warnings for a company that does not yet have a path to profitability.
The Anthropic contract only requires 90 day notice to cancel, Grok is not considered a leading AI model by the analysts who cover the sector and the valuation at $1.75T is approximately twice what Morningstar thinks the business is worth.
The original conclusion holds. The business being sold on June 12th is larger, more complex, more indebted, and more interesting than the original filing suggested. Forced buying from index funds will support the price in the near term, the real price discovery happens after the lockup expires, index funds have the correct weighting and after the market has seen one or two quarters of actual financials. The roadshow started on June 8th and the listing is June 12th.
Ticker Thoughts is independent analysis. No position held in $SPCX at time of publication.