The SpaceX IPO Is a Finance Story. The Real Story Is Wireless Infrastructure.
SpaceX went public today. The financial press will spend the next week on the valuation — $1.77 trillion, the largest IPO in US history, shares priced at $135, Elon Musk retaining 82% voting control. That story will be well covered.
Here is the story that won’t be.
SpaceX is, first and foremost, a wireless infrastructure company. It has done more to transform how humanity builds and shares communications networks than any organization since the early commercial mobile carriers of the 1990s. The IPO is an endorsement of that bet. To understand what SpaceX is actually worth — and what it is actually doing — you have to start with physics, not finance.
Three transformations are already complete. A fourth is underway.
Part 1: The Launch Cost Collapse
Everything SpaceX has built in communications rests on a single underlying achievement: making it dramatically cheaper to put things in orbit.
In the Space Shuttle era, getting a kilogram to low Earth orbit cost roughly $54,000. Commercial launch was dominated by a handful of incumbents. Access to space was a privilege of governments and large institutions. That wasn’t a policy choice — it was the economics of expendable rockets.
Falcon 9 changed the math. SpaceX landed its first orbital-class booster in December 2015, and the company has since reflown individual boosters more than 300 times. Today, a dedicated Falcon 9 launch costs about $74 million, or roughly $3,200 per kilogram. SpaceX estimates its internal cost at a fraction of that. The company now accounts for more than 90% of global commercial orbital launches — flying roughly every two to three days.
Starship, if it meets its design goals, points toward something even more dramatic: a fully reusable vehicle designed to carry 100 to 150 metric tons to LEO, potentially at sub-$100 per kilogram. That would be another order-of-magnitude reduction.
Why does this matter for wireless? The entire LEO broadband industry is a direct derivative of the launch cost curve. Lower launch cost means more satellites. More satellites means denser constellations. Denser constellations means more spectrum reuse and more capacity. You cannot build Starlink at 2010 launch prices. The communications infrastructure and the launch infrastructure are the same investment.
Part 2: Satellite Broadband at Scale
Starlink is now 61% of SpaceX’s revenue — $11.4 billion in 2025, with the only profitable margins in the company at around 39% operating margin. Those numbers matter, but they’re not the interesting part.
The interesting part is what Starlink actually delivers and to whom.
Geosynchronous satellites, the kind that have served direct-to-home TV and internet for decades, orbit at roughly 35,000 kilometers. At that distance, the round-trip signal delay — the latency — is over 600 milliseconds. That’s long enough to make video calls awkward and real-time applications nearly unusable. Starlink satellites orbit at roughly 550 kilometers. Round-trip latency is 20 to 40 milliseconds, comparable to a mid-quality terrestrial connection.
For a rural household with no cable, no fiber, and no real alternative, that difference is the difference between a functional internet connection and a hobbled one. Starlink has crossed 10 million subscribers. For a meaningful fraction of those customers, it is not a premium upgrade. It is the only viable option.
The FCC authorized an additional 7,500 Gen2 satellites in January 2026, bringing the total authorized constellation to 15,000. In April, the FCC overhauled its spectrum sharing rules in a way estimated to deliver roughly a sevenfold capacity boost. A related authorization now allows Starlink to connect directly to standard mobile handsets — no dish required — eliminating cellular dead zones for partnered carriers.
The policy question this creates is not financial. It’s governance. Starlink is now communications infrastructure for remote communities on every continent. It is the backup link for disaster response, the primary internet for researchers in Antarctica, the emergency connection for communities that terrestrial networks don’t reach. It is a private company’s proprietary constellation. The IPO makes that tension visible in a new way, because it makes the ownership visible.
Part 3: The Spectrum Story No One Is Telling
This is the part the financial press won’t cover, because it requires understanding how spectrum governance actually works.
Spectrum — the range of radio frequencies used for wireless communication — is a shared resource. Different users have to coordinate access, or they interfere with each other. For satellites, the relevant coordination framework has historically been a set of technical limits called Equivalent Power Flux Density, or EPFD, limits. These limits constrain how much power a non-geostationary satellite constellation can dump into a given area, to protect the GEO operators — Viasat, DirecTV, Intelsat, and others — whose customers were already there.
The EPFD framework was designed in the 1990s. It assumed a small number of NGSO entrants competing with established GEO operators. It did not anticipate a single operator deploying thousands of satellites.
In April 2026, the FCC voted unanimously to replace the EPFD framework with performance-based spectrum sharing. Under the new rules, NGSO operators like SpaceX can deploy denser constellations, as long as real-world interference stays below measured performance thresholds. The FCC formally recognized Adaptive Coding and Modulation — a technique that allows ground terminals to compensate for interference in real time — as a legitimate mitigation strategy. GEO operators objected. The FCC sided with the performance-based approach.
This matters for three reasons. First, it removes a regulatory ceiling on how dense Starlink can become. Second, it sets a precedent: every future NGSO constellation — Amazon’s Project Kuiper, OneWeb, and dozens of smaller operators — will now share spectrum under this framework. Third, it represents a genuine rewriting of international-grade spectrum governance, driven primarily by one company’s ability to outscale the existing rules.
Meanwhile, the FCC separately approved SpaceX’s acquisition of approximately 65 MHz of nationwide mid-band spectrum from EchoStar. This spectrum — harmonized with global 3GPP standards — is what gives Starlink Mobile a credible path to operating like a terrestrial mobile carrier in the spectrum it occupies.
SpaceX went public having effectively lobbied a rewrite of international spectrum governance to accommodate its constellation. The IPO is partly a bet that those regulatory outcomes are durable.
Part 4: What’s Coming Next — Orbital AI Compute
The fourth transformation is the most speculative, and the financial press is actually covering this one, which means it’s worth being precise about what is announced versus what is projected.
In February 2026, SpaceX acquired xAI — Musk’s AI company — for $250 billion. The combined entity was reorganized into separate business units and is now part of the IPO prospectus.
SpaceX has announced a satellite platform called AI1: an orbital compute node with 150 kilowatts of peak compute capacity, a 70-meter solar wingspan, and liquid radiators for heat dissipation into the vacuum of space. Two prototype launches are planned for early 2027. The prospectus lists early compute customers — Google and Anthropic are named — at contract values that, if accurate, are substantial.
The engineering logic is straightforward. Space offers two things that are increasingly scarce for terrestrial data centers: effectively unlimited solar power and a near-perfect heat sink. No land acquisition. No grid approvals. No cooling buildout. Starship can carry multiple AI1 satellites per launch.
The open question — and this is genuinely open — is the transmission economics. Consumer broadband and orbital AI compute are different problems. Moving petabytes of AI training data through a satellite link, at the latencies and throughputs currently achievable, is not the same as serving video to a rural household. The link-budget math for orbital compute at scale is not yet demonstrated. The prospectus projects; the physics will decide.
What is clear is the integration: every AI1 satellite is also a Starlink node. The constellation and the compute fabric are the same infrastructure. That is a genuinely novel architecture.
What the IPO Moment Actually Means
SpaceX is going public as a company that has collapsed the cost of access to orbit by roughly 95%, built the world’s largest satellite broadband network, forced a rewrite of international spectrum governance, and begun deploying orbital compute infrastructure. The $1.77 trillion valuation is an attempt to price all four of those things simultaneously — including the fourth, which is early.
For engineers and researchers, this is what it looks like when the physics constraints are removed faster than the regulatory and governance structures can adapt. The interesting open problems are in the places where SpaceX has outrun the frameworks: spectrum coexistence at mega-constellation scale, link-budget design for orbital compute, interference characterization for NGSO-to-NGSO sharing as multiple large constellations compete for the same orbital bands.
For citizens, something simpler and more significant: a company that operates communications infrastructure for underserved communities on every continent, that has reshaped how nations think about spectrum access, is now partially answerable to public shareholders. That is new. What it means in practice is still being worked out.
The finance story will dominate the next few days. The infrastructure story will matter longer.
AI Notice: The development of this post was accelerated with the help of Anthropic Claude and Google Gemini.