SpaceX Roadshow Deep Dive: The Largest IPO in Human History, Aiming to Launch GPUs into Space
TechFlow Selected TechFlow Selected
SpaceX Roadshow Deep Dive: The Largest IPO in Human History, Aiming to Launch GPUs into Space
Only we can bring AI computing into space.
Navigating Web3 tides with focused insightsBy Xiao Bing, TechFlow Research
On June 4, SpaceX officially launched its IPO roadshow. The 62-page presentation outlines an offering of 555.6 million shares at $135 per share, raising $75 billion and targeting a valuation of $1.75 trillion. If all proceeds as planned, pricing will occur on June 11, with Nasdaq listing scheduled for June 12 under the ticker symbol SPCX.
This would be the largest IPO in human capital market history—surpassing Saudi Aramco, Alibaba, and every other precedent.
Goldman Sachs, Morgan Stanley, Bank of America, Citigroup, and JPMorgan Chase jointly serve as lead underwriters; 21 institutions participate in distribution. Elon Musk personally commits to a 366-day lock-up period, while other insiders begin phased unlocks after the Q2 2026 earnings release. Fidelity has opened subscription access to all retail accounts holding over $2,000.
The internal codename for the roadshow deck is “Project Apex”—a title fully justified by its content.
Three Pillars: Space, Connectivity, AI
In its roadshow, SpaceX defines itself as “the only company simultaneously building infrastructure across space, connectivity, and AI.” This is not marketing rhetoric. Financial data reveals starkly divergent growth trajectories, profitability profiles, and capital requirements across the three business segments—making SpaceX an exceptionally complex investment proposition.
Space: The Foundation
In 2025, SpaceX completed 165 Falcon-series launches using only eight newly manufactured boosters. Reusability has transitioned from experimental to industrial-scale production, slashing launch costs from the industry’s historical average of $18,500/kg to $2,700/kg for Falcon 9 and $1,400/kg for Falcon Heavy. Starship V3 aims to cut costs by over 99% beyond that baseline.
SpaceX delivers over 80% of all mass placed into orbit globally. That figure stood at 65% in 2023 and 45% in 2021—a level of market concentration unprecedented in any infrastructure sector.
Yet the space business itself delivers unremarkable financial performance. In 2025, it generated $4.1 billion in revenue (excluding internal satellite launches), representing just 8% year-on-year growth. More critically, Starship R&D spending reached $3 billion in 2025, dragging the space segment’s operating profit into a $657 million loss. Adjusted EBITDA declined from $1.2 billion in 2024 to $700 million in 2025.
The value of the space business lies not in its own profitability, but in its role as an ultra-low-cost deployment engine for the other two pillars. Every Starlink satellite launch—and every future orbital AI satellite deployment—rests upon the cost curves of Falcon and Starship.
Starlink: The Cash Machine
Starlink is SpaceX’s true valuation anchor.
User count grew from 2.3 million in 2023, to 4.4 million in 2024, 8.9 million in 2025, and reached 10.3 million by Q1 2026. It serves 164 countries and regions, delivering median download speeds of 225 Mbps, median latency of ~25 ms, and average uptime of 99.9%. Among all operational satellites, Starlink accounts for roughly 75%.
Financial metrics are even more compelling: Starlink generated $11.4 billion in revenue in 2025, up 50% year-on-year, with adjusted EBITDA of $7.2 billion and operating profit of $4.4 billion. This is SpaceX’s sole consistently profitable segment—and its margins continue to expand.
The roadshow deck reveals a key technical upgrade: V3 satellites. Each V3 satellite delivers 1,024 Gbps of bandwidth—more than ten times current V2 satellites. A single Starship launch can carry 60 V3 satellites, adding 61,000 Gbps of network capacity per mission—over 20× the capacity added by a Falcon 9 launch of V2 satellites.
V3 satellite deployment on Starship is slated to begin in the second half of 2026. If Starship achieves operational reusability on schedule, Starlink’s bandwidth expansion rate will leap by an order of magnitude—further widening its lead over all competitors.
Starlink Mobile (direct-to-cell satellite service) also merits attention. Approximately 650 first-generation mobile satellites are already deployed, covering ~1.9 billion people. SpaceX has partnered with around 30 mobile operators—including a 2025 agreement with American Airlines for in-flight connectivity. Second-generation mobile satellites are scheduled for deployment on Starship in 2027, delivering 5G-class speeds and voice services. In 2025, SpaceX also signed an agreement to acquire EchoStar’s U.S. and global mobile satellite spectrum licenses for $65 MHz, expected to close in November 2027.
The roadshow estimates the total addressable market (TAM) for connectivity at $1.6 trillion ($870B broadband + $740B mobile). Penetration remains extremely low given current growth trajectories.
AI: Money Pit—or Trillion-Dollar Bet?
In February 2026, SpaceX completed an all-stock acquisition of xAI, valuing the combined entity at $1.25 trillion. This transaction is the pivotal variable behind SpaceX’s $1.75 trillion IPO valuation—and the source of its greatest controversy.
The merged AI business comprises three components:
First, compute infrastructure. Colossus I and Colossus II together deliver 1 GW of computational power—marketed as the world’s largest coherent supercomputer, and the first GW-scale cluster deploying NVIDIA GB200 and GB300 chips, supported by a co-located 1 GW Tesla Megapack battery storage system.
Second, the Grok large language model. The roadshow claims Grok achieves frontier-level performance on scientific reasoning benchmarks (e.g., GPQA Diamond) and “outpaces all other leading model providers in speed.” Grok 4.3 was released in May 2026. SpaceX also signed a cooperation agreement with Cursor, granting it an option to acquire Cursor at an implied valuation of $60 billion.
Third, the X platform. It boasts ~550 million monthly active users (including Grok and X users), with ~350 million daily posts. 117 million monthly active users have engaged with Grok’s AI features. X is rolling out a new advertising platform, X Ads Manager, and plans to evolve into an “all-in-one app” integrating information, communication, media, payments, and banking.
AI monetization currently follows three parallel paths: consumer-facing (X Premium subscriptions + advertising), enterprise-facing (Grok Enterprise/API + Cursor collaboration), and compute sales (capacity agreements with cloud providers, generating $1.25 billion per month through May 2029).
But financial reality is harsh. The AI segment generated $3.2 billion in revenue in 2025—dominated by X’s ad and subscription income. Adjusted EBITDA was negative $1.2 billion; operating losses totaled $6.4 billion—consuming 61% of the company’s total capital expenditures. Morningstar estimates xAI will burn $10 billion in 2026.
The roadshow assigns a near-term TAM for AI of $3.8 trillion ($760B infrastructure + $600B consumer subscriptions + $2.4T digital advertising). Adding “AI-enabled broader opportunities,” the total TAM balloons to $26.5 trillion.
The Most Valuable Slide: Putting GPUs in Orbit
Slides 35–36 of the roadshow deck contain the highest information density—and represent SpaceX’s most distinctive strategic narrative.
The core thesis: U.S. terrestrial electricity supply can no longer keep pace with AI compute demand. In 2025, data center electricity demand reached 62 GW, against only 49 GW of available supply—a 13 GW shortfall. U.S. electricity generation grew nearly zero percent between 2008 and 2023, while China’s rose ~6% over the same period. Building ground-based data centers faces regulatory bottlenecks—including grid interconnection approvals, land-use planning, and community opposition.
SpaceX’s solution: Move AI computation into space.
Orbital AI computing satellites build directly on the Starlink V3 satellite platform. The roadshow lays out a clear evolutionary path: Starlink V3 retains core components—including inter-satellite laser links, flight computers, and attitude control—but removes the downlink antenna, large batteries, and modem, adding AI compute chips, additional solar panels, and larger radiators.
SpaceX claims orbital AI computing offers three structural advantages:
First, unlimited, clean, lower-cost solar power—distributed via the Starlink network, bypassing terrestrial grid approval hurdles. In geosynchronous orbit, satellites receive sunlight over 99% of the time, enabling uninterrupted AI training.
Second, radiation-based thermal management—lower-cost than liquid or air cooling. Data routing between orbital compute clusters and ground users leverages Starlink’s existing network.
Third, faster next-generation chip deployment. Each GPU generation delivers step-function improvements in token efficiency. Starship’s rapid payload launch cadence enables faster hardware refresh cycles than terrestrial data centers.
SpaceX estimates: launching 1 million tons of satellites annually—with each ton delivering 100 kW of compute—would add 100 GW of AI compute capacity per year, with virtually no ongoing operational cost.
On January 30, 2026, SpaceX filed an application with the FCC to deploy up to one million orbital data center satellites. The FCC accepted it for review on February 2. This is the largest data center construction proposal in human history. Orbital compute node pilot testing will begin on Starlink V3 hardware in the second half of 2026, with full-scale AI satellite deployment commencing in 2028.
The narrative’s power lies in how it repositions SpaceX—not as a rocket company or satellite internet provider, but as a *global AI infrastructure supplier*.
A phrase repeated throughout the roadshow: “Only we can do this.”
This confidence stems from a vertically integrated stack no other company can replicate: proprietary rockets (lowering launch costs) → proprietary satellites (lowering manufacturing costs) → proprietary inter-satellite network (lowering data transport costs) → proprietary AI models (direct compute consumers) → proprietary end-user platform (X, with 550M MAU). Full-stack ownership—from silicon to space to endpoint.
Google is pursuing similar ambitions. Its Project Suncatcher, announced in November 2025, plans to launch two prototype satellites in early 2027 with Planet to validate in-orbit AI workloads. But Google must rely on SpaceX for launch—and lacks its own satellite network.
Still, external observers remain measured.
Varda Space Industries estimates orbital compute currently costs ~3× more per watt than terrestrial alternatives. Musk claims cost parity within 2–3 years, but independent analysts generally project it won’t arrive until the 2030s. Cosmic radiation effects on chip reliability, vacuum-based thermal engineering, and orbital-ground latency remain unresolved technical challenges. AWS leadership publicly stated orbital data centers are “far from practical use.”
Yet even discounting this vision by 50%, SpaceX’s structural advantage holds: If any company pursues orbital computing, it will ultimately need SpaceX’s launch services—whether orbital data centers materialize in 2028 or 2035, SpaceX is the unavoidable gateway.
The Vision: Lunar and Martian Development
Slides 43 and 44 contain no revenue forecasts or timelines—just six phrases, each followed by a scene straight out of science fiction: Lunar economy, Mars-based energy production and manufacturing, point-to-point Earth travel, in-orbit manufacturing, crewed and cargo missions to Mars, asteroid mining…
The slide title reads: “We are uniquely positioned to create new trillion-dollar markets.”
Under “Lunar Economy,” SpaceX provides slightly more detail—three parallel tracks:
First, supporting NASA’s Artemis program to return humans to the Moon by the late 2020s using Starship, establishing a sustainable lunar base, and validating all systems required for long-term extraterrestrial human habitation; second, building AI-powered satellite factories on the Moon—leveraging solar energy and lunar mass drivers to launch satellites into orbit; third, scaling global AI compute capacity from GW to TW levels via this lunar manufacturing-and-launch pipeline.
Only the first track carries external validation: NASA’s Artemis contract is a real commercial award. SpaceX is the sole selected Human Landing System contractor, and landing astronauts on the Moon by the late 2020s remains plausible given current progress.
The second and third tracks remain conceptual engineering exercises. The lunar factory requires solving numerous open problems: abrasive lunar regolith damaging manufacturing equipment, precision assembly in low gravity, and engineering validation of mass drivers—each potentially requiring decades of development.
Point-to-point Earth travel (e.g., New York to Shanghai in 30 minutes via Starship) and asteroid mining lack even tentative timelines—even from SpaceX itself.
Yet these two slides answer a critical investor question: What, exactly, are investors buying at a $1.75T valuation?
Morningstar’s DCF model values SpaceX at $780B—anchored to Starlink’s predictable cash flows and space launch revenue’s stability. The nearly $1 trillion gap between $780B and $1.75T reflects a far larger proposition: *If humanity truly expands beyond Earth, SpaceX is the only infrastructure provider.*
This proposition does not require investors to believe lunar factories will be built—or that Mars colonization will happen in their lifetimes. It only requires belief in one thing: If even one of these scenarios becomes reality, only SpaceX can execute it.
Exclusivity—not certainty—is the pricing logic behind this $1 trillion premium.
TechFlow Interpretation
In TechFlow Research’s view, SpaceX—like Tesla—is a “faith stock.”
As Morningstar analyzes, Starlink alone could justify a valuation exceeding $600B. Yet the $1 trillion gap between $780B and $1.75T represents a “faith premium”—pricing in decade-long call options on orbital AI computing, lunar economy, and Mars colonization. A 94x revenue multiple is unprecedented among trillion-dollar companies.
Moreover, xAI is the largest risk variable in this IPO—and the roadshow discusses it inadequately.
The $250B all-stock acquisition in February 2026 occurred with Musk controlling both buyer and seller. This related-party transaction instantly saddled SpaceX with the entire AI segment’s losses. Pre-merger, SpaceX briefly posted $800M net income in 2024; post-merger, it reported a $4.9B net loss in 2025 and a $4.3B quarterly loss in Q1 2026. The AI segment’s annual operating loss stands at $6.4B, with projected 2026 burn reaching $10B—while Starlink’s $4.4B operating profit covers less than 70% of that deficit.
Grok’s position in the cutting-edge model race remains highly uncertain; X’s advertising recovery is still in early stages. More critically, Musk retains absolute control via dual-class voting shares—leaving public shareholders with virtually no oversight over future related-party transactions or capital allocation decisions. Morningstar explicitly labels xAI a “substantial value-destroying threat.”
Finally, SpaceX’s short-term trading logic may directly contradict its long-term investment thesis.
An extremely low 3% free float, anticipated rapid inclusion in the Nasdaq-100 (as early as July), underwriter support from 21 investment banks, and intense market enthusiasm for AI infrastructure could create severe supply-demand imbalance at launch—pushing the stock well above fundamental valuation.
Yet SpaceX’s lock-up structure is unusual: insiders may begin selling 20% of their holdings in phases after the Q2 earnings release; the first major unlock occurs in December 2026; Musk’s personal stake unlocks after 366 days—in June 2027. Coupled with quarterly earnings exposing AI’s mounting losses, a significant sell-off window may emerge from late 2026 through mid-2027.
Overall, the 62-page roadshow paints a picture of a full-stack infrastructure empire—from Earth to orbit, from rockets to AI. SpaceX’s launch capability and Starlink’s growth curve have already proven Musk’s team’s execution prowess. The question is: Where is the boundary of that execution? Within Earth’s atmosphere—or beyond?
The answer to that question will determine whether $1.75T reflects foresight—or hubris.
Disclaimer: This article reflects only TechFlow Research’s analytical perspective and does not constitute investment advice. SpaceX has not yet formally listed. Financial data presented in the prospectus is preliminary, unaudited, and subject to revision. Investors should carefully read SpaceX’s S-1 registration statement and prospectus filed with the SEC before making any investment decision, and fully understand all associated risk factors.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
@BlockFlow_News
