SpaceX is set to go public this Friday, and investor excitement is palpable. The $75 billion stock offering is reportedly heavily oversubscribed, with some institutional investors committing billions of dollars to secure shares in Elon Musk's company.
From an investment standpoint, there are significant doubts: large IPOs often see post-listing declines; the company remains unprofitable; and the CEO's erratic public behavior would typically raise red flags for any other tech firm. Yet, these concerns have not dampened investor enthusiasm. Tech investors have learned one rule: no matter the business logic, never bet against Elon Musk.
However, setting aside market sentiment, a sober look at SpaceX's financial strategy reveals what investors are truly wagering on. Over the past 18 months, as Musk consolidated his corporate empire and prepared for the IPO, a new blueprint emerged with orbital data centers at its core.
True to his audacious style, Musk has proposed an aggressive plan that hinges on solving three near-impossible engineering challenges: perfecting fully reusable rockets, building a domestic U.S. chip foundry, and mass-producing satellites at an unprecedented rate.
This business model defies precise valuation. This week, two separate, rational assessments of SpaceX were published: one by financial research firm Morningstar and another by New York University finance professor Aswath Damodaran, an expert in corporate valuation. Both valuations came in far below the nearly $1.8 trillion figure touted by investment banks. Morningstar pegged the value at approximately $825 billion, while Damodaran estimated it at $1.2 trillion.
The massive gap in valuation largely stems from the company being a world-leading aerospace monopoly that is now layering on a highly speculative artificial intelligence venture. Morningstar calculated a fair share price of $63, while the IPO is priced at $135. The firm stated that the $72 difference essentially represents a call option, betting that SpaceX can build its orbital data centers with the speed and performance Musk envisions.
Both assessment reports noted that the launch services and Starlink satellite internet businesses are highly profitable and the company's most attractive segments, whereas the AI venture is fraught with uncertainty.
Navigating the Cloud Computing Frontier
First, a key question must be answered: what exactly is SpaceX's AI business? In the market analysis section of its S-1 filing, the company identifies enterprise AI as its largest opportunity. This involves building programming tools with talent absorbed from the Cursor team or advancing the Macrohard project—developing intelligent agents capable of white-collar work. SpaceX estimates the total addressable market for this sector at $22.7 trillion, with the AI infrastructure market at $2.4 trillion, and the space business market slightly below $2 trillion.
However, this plan seems at odds with the company's recent actions of selling significant computing power to AI competitors like Anthropic and Google. This aligns with the typical operating style of Musk's ventures; Space Exploration Technologies also launches satellites for Starlink's competitors. But whereas past collaborations leveraged its core industry strengths, the current move feels more reactive.
Pivoting to become a new type of compute provider might yield short-term gains, but it raises a fundamental industry question: where does the ultimate value accrue in the AI technology stack? If one cannot do both, is it better to be the compute supplier or to develop proprietary large language models?
The scaling logic in AI is clear: leading-edge labs must continuously train more powerful new models, or, as Musk suggested in his recent lawsuit against Sam Altman, extract and refine capabilities from others' models. The industry is a relentless race; those who pause fall behind. Of course, the rising performance of low-cost open-source models could disrupt this dynamic.
The orbital data center is SpaceX's proposed solution to this dilemma—by offering massive compute supply, the company aims to simultaneously be both a compute provider and a model developer.
Musk's Vision for Orbital Data Centers
This week, in a video interview, Musk explained the unique advantages SpaceX holds for deploying such data center projects. The core thesis is that only SpaceX can cheaply launch heavy payloads to low-Earth orbit while also possessing the capability for mass production of solar panels and chips. While industry experts generally believe scaling space-based data centers is at least a decade away, Musk, with many caveats, suggested the timeline could be much faster.
"This is not a forward-looking promise for us," Musk said in the video. "It's what we plan to execute and are likely to achieve: striving to reach an annualized AI compute capacity of 1 gigawatt in space by the end of next year."
Based on a maximum power supply of 150 kilowatts per satellite, achieving this goal would require manufacturing 6,666 satellites annually, or about 556 per month. This production rate is roughly double the current Starlink satellite output, which stands at about 70 per week. Although Musk stated that AI satellites have a simpler structure, the new production line is not yet built, making this target immensely challenging. Concurrently, the company's solar panel production facility is also still under construction.
This does not even address the high-profile Terra chip foundry. Musk's plan is for this factory to support the venture once it scales to terawatt-level annualized compute capacity. Chip foundries are among the most difficult modern industrial projects, typically costing tens of billions of dollars and taking a decade to build.
And the most critical question remains: what is the progress on the Starship, the essential vehicle for cheaply launching massive amounts of hardware into space?
The most recent Starship test flight showed promise but remains far from achieving rapid reusability. Initially, SpaceX may only manage to recover and reuse the rocket booster, which would significantly increase the cost of deploying space data centers. Currently, the Federal Aviation Administration is investigating an incident where a booster failed to perform a controlled re-entry as planned. SpaceX has not announced a date for the next Starship test flight, stating only that it plans to use the vehicle to launch Starlink satellites by year-end.
Perspective is needed here: NASA has a nearly $4 billion contract with SpaceX to use Starship as a lunar lander, yet the agency has not finalized a test flight mission for late 2027.
A Word of Caution for Investors
For the average public investor, buying SpaceX shares means owning a piece of a near-monopoly aerospace service provider in the West, a global communications network, and a bet on the most ambitious infrastructure project of the AI era.
All of this depends on achieving an unprecedented feat: a fully and rapidly reusable rocket. The company must also establish a high-volume AI satellite production line within 18 months—a task for which the original Starlink line took a full decade. Finally, it must build a domestic chip foundry, a project even dedicated semiconductor firms approach with trepidation. Musk is correct that, in the short term, only SpaceX has the capability to attempt these projects. But that very fact underscores the extraordinary difficulty; whether they can ultimately be realized remains an open question.
Musk previously stated he would not take SpaceX public before reaching Mars, fearing that fickle investors would lose faith along the journey. That plan is now shelved. The new blueprint unveiled ahead of the IPO presents a challenge no less daunting than a mission to Mars.
Comments