SpaceX’s earnings reveal a company selling internet, compute and sci-fi ambition—not just rockets

SpaceX’s first quarterly results as a public company show that the business is now driven far more by satellite internet and AI compute than by rockets. The filings and earnings call indicate that spaceflight still matters, but it accounts for a relatively small share of revenue while Starlink and data-center leasing have become the company’s main growth engines.

That shift matters because it changes how investors, competitors and regulators should understand Elon Musk’s most valuable company: less as a pure launch operator and more as a hybrid telecom-and-compute platform with enormous capital demands and a growing AI footprint.

What SpaceX’s numbers actually show

SpaceX’s latest quarterly report suggests the company’s revenue mix is much broader than its name implies. Rockets remain central to its identity, but they are no longer the main economic story. Instead, satellite connectivity appears to be the clearest profit center, while compute leasing has emerged as one of the fastest-growing and most expensive parts of the business.

According to the company’s first quarterly earnings statement as a public company, space-related activity brought in just over 10% of revenue in the quarter and did not reach $1 billion. By contrast, Starlink generated $4.2 billion and was the only major segment that did not post an operating loss.

The numbers underline a basic truth about the business: launching rockets has become only one part of a far larger operation that now includes internet services, infrastructure for AI companies and an expanding set of side bets tied to Musk’s broader ecosystem.

Why Starlink now looks like the core business

Starlink is the most mature and commercially reliable part of SpaceX’s portfolio. It generated the bulk of the company’s revenue in the quarter, and it was the only division that did not lose money from operations. That is a significant milestone for a business whose public image is still dominated by launches, spacecraft and Mars rhetoric.

The satellite internet service also gives SpaceX a product that can be sold repeatedly, unlike launch contracts, which tend to be cyclical and dependent on customer demand. That makes Starlink resemble a telecom business more than a space company, particularly as SpaceX explores services that could eventually compete with the biggest U.S. carriers.

How does SpaceX want to compete with phone carriers?

SpaceX is aiming to extend Starlink beyond home broadband and remote connectivity. On the earnings call, company leaders described plans for a phone service that would go up against AT&T, Verizon and T-Mobile, signaling an ambition to move into an even larger slice of the consumer communications market.

That push would make Starlink not just a satellite provider but a broader connectivity platform, one that could bundle internet, voice and mobile access in areas where terrestrial networks are weak or where SpaceX can undercut incumbents with space-based coverage.

Gwynne Shotwell, SpaceX’s president and chief operating officer, used the earnings call to sketch out a future in which the company’s connectivity business reaches well beyond broadband, positioning Starlink as a potential challenger to the major U.S. wireless carriers.

How did AI become such a big part of SpaceX?

AI became a major part of the business because SpaceX built infrastructure for xAI and then discovered that the project was not as straightforward as expected. The company’s Colossus 1 data center in Memphis was originally meant to support Grok, Musk’s in-house chatbot and model family. When the setup proved difficult to run efficiently, the facility was repurposed, and SpaceX began selling compute capacity to outside customers as well.

That pivot turned a technical problem into a revenue opportunity. Rather than using the Memphis facility only for xAI’s own training workloads, SpaceX is now leasing data-center capacity in a market where AI firms urgently need chips, power and space. The result is a new business line that looks a lot like the so-called neocloud model used by companies such as CoreWeave and Nebius.

What problems did the Memphis facility run into?

The biggest issues were operational. The data center reportedly faced latency problems that made in-house model training difficult, and its hardware mix included both newer and older chips, creating bottlenecks. Those complications made the original xAI-only setup less efficient and helped push the company toward renting capacity to other buyers.

In practical terms, that means the facility was less a pure internal AI engine and more an expensive infrastructure asset looking for utilization. In the AI business, unused capacity is costly, and finding outside tenants can help offset that burden.

SpaceX is becoming a compute landlord

The most surprising shift in the earnings picture is that SpaceX now appears to be a serious player in the market for rented compute. The company has signed deals with Google, Anthropic, Reflection AI and Cursor, adding another layer to its business beyond launch services and satellite internet.

That move matters because data-center capacity is a commodity business with thin margins and enormous capex requirements. The more supply comes online, the harder it becomes for providers to charge premium rates. SpaceX is therefore entering a market where technical execution, energy costs, chip access and construction speed all determine whether the model can scale profitably.

Bloomberg reported that analyst Alexander Potter expects neocloud spending to reach $65 billion next year, an increase of $17 billion from his prior estimate. That level of demand reflects just how fast AI infrastructure spending is expanding — and how much money is chasing it.

What did SpaceX say about revenue growth?

SpaceX executives said the company is on track to reach a $100 billion annualized revenue run rate, or ARR, once the contribution from Cursor is included. CFO Bret Johnsen told investors that the business is moving toward that milestone, while Musk went further and said hitting $100 billion ARR in December was “not a question mark.”

Those remarks were striking, not because the growth target is impossible, but because they show how SpaceX now talks about itself as a platform business with multiple high-scale revenue streams rather than as a company defined by rockets alone.

On the call, SpaceX’s finance chief framed the company’s AI and compute contracts as a path toward an annualized revenue run rate of $100 billion, while Musk argued that the target could be reached by December and might even be exceeded.

Why compute is a risky business, even for SpaceX

Revenue is not the same as profit, and the economics of data centers are unforgiving. Compute infrastructure requires constant investment in chips, cooling, power and real estate. Hardware ages quickly, customer demand can shift fast, and prices fall as more capacity comes online.

That makes SpaceX’s compute strategy very different from its rocket business, which benefits from scarcity and highly specialized demand. In AI infrastructure, the commodity nature of compute means providers often compete on price, not just performance. That dynamic can compress margins even when top-line growth looks impressive.

It also introduces a new kind of operational complexity. Unlike launch contracts, which are episodic and high-profile, data-center leases depend on uninterrupted uptime, stable power delivery and successful hardware refreshes. If those assumptions fail, revenue projections can become much less reliable.

• Data centers require large upfront capital outlays.
• Chip cycles make hardware obsolete quickly.
• Construction delays can hurt utilization and returns.
• AI customers often shop for lower-cost capacity as supply expands.

How did the space business become the smallest piece of SpaceX?

The answer is that demand for launches has not kept pace with the company’s broader ambitions. SpaceX remains a dominant launch provider, but it is also heavily reliant on its own missions, which means it is still one of its own biggest customers. That is not the sign of a mature external market for launch services.

The report suggests that the aerospace side of the company failed to break the $1 billion mark in quarterly revenue and represented only a little more than one-tenth of total sales. That is a far cry from the image of SpaceX as a launch company first and everything else second.

The mismatch between branding and business reality is especially sharp because the company’s public narrative still revolves around Mars, colonization and deep-space exploration. Yet the money is increasingly coming from telecom and infrastructure deals on Earth.

Space data centers: vision, skepticism and spectacle

Musk has said he took SpaceX public partly because he wanted to build data centers in space. The idea is not new in science fiction circles, but the practical obstacles are enormous. Space-based computing would need to solve problems of launch cost, orbital deployment, radiation, thermal management, networking and maintenance at a scale no company has demonstrated.

SpaceX has already proposed an orbital data center to the Federal Communications Commission, describing a system that could involve as many as 1 million satellites. But the application appears thin on specifics, including details such as satellite size and rollout timing. That lack of detail has led many observers to treat the proposal more as strategic theater than an executable industrial plan.

What is Musk’s Moon and orbital plan?

The broad concept is that SpaceX would eventually scale from ground-based compute into orbit and beyond, with a future in which a Musk-controlled chip business called Terafab produces vast quantities of chips, Optimus robots handle labor and a mass accelerator could one day be built on the Moon.

That sounds more like a speculative narrative than a concrete engineering roadmap. Even so, the rhetoric serves a purpose: it keeps investors and followers focused on a grand future in which today’s expensive infrastructure bets look like stepping stones rather than distractions.

Critics say that is exactly the problem. If the terrestrial data-center business is still struggling with latency, bottlenecks and capex, then the jump to space-based compute looks less like inevitable progress and more like a branding exercise for a still-unproven business model.

The political and financial backdrop

SpaceX’s compute business does not exist in a vacuum. Musk’s political relationships, his influence in Washington and his broader strategy of tying companies together all shape how the business is perceived and how it may be valued. If outside buyers see strategic advantage in staying close to Musk, that could support premium pricing for compute services.

At the same time, his other companies remain entangled with SpaceX’s economics. Tesla stands to benefit from SpaceX purchases of Megapack storage systems, and the aerospace company has also been buying Cybertrucks, giving Musk yet another internal loop of demand across his empire.

That interconnectedness may help explain why SpaceX’s business picture is so unusual. It is not just a space company, a telecom company or an AI landlord. It is all of those things at once, with customer relationships that often circle back to Musk-controlled firms.

Why do insider lockups matter now?

Because timing matters for valuation and morale. The company’s insider lockup period begins expiring on August 6, opening the door for early investors and insiders to sell shares. If significant selling follows, the stock could face additional pressure, especially if investors become uneasy about the distance between the company’s branding and its actual revenue mix.

That makes the timing of the earnings narrative important. Big talk about Moon infrastructure, orbital compute and trillion-dollar futures may help keep attention on the upside just as liquidity constraints start to ease for insiders.

What does this mean for investors and competitors?

For investors, the main lesson is that SpaceX is no longer a simple bet on launch demand. It is a multi-business infrastructure company with exposure to telecom, AI compute and capital-intensive speculation about the future of space industry economics. That creates both optionality and risk.

For competitors, the message is that Musk’s companies can now pressure several markets at once. Starlink threatens traditional telecom and broadband players. The compute business puts SpaceX in contention with neocloud providers. And the company’s sheer scale means it can use cross-company relationships to strengthen its position in hardware, storage and AI infrastructure.

The challenge is that each of those businesses demands a different skill set. Launch operations are not the same as running data centers, and neither is the same as competing in consumer telecom. SpaceX may be broadening its opportunity set, but it is also stretching itself across industries with very different economics.

The bottom line

SpaceX’s latest earnings make one thing clear: the company is becoming less about spaceflight and more about infrastructure. Starlink is the most reliable source of revenue, compute leasing is the newest growth engine and rockets are now only one slice of a much larger, riskier and more capital-intensive portfolio.

Musk may still talk about Moon-based accelerators, orbital data centers and future megastructures, but the numbers point to a more grounded reality. SpaceX is increasingly a telecom and AI infrastructure company that happens to launch rockets — not a rocket company that happens to have a side business.

Whether that transformation makes the company more valuable or merely more complicated will depend on whether its compute, connectivity and launch businesses can keep growing without overwhelming one another. For now, the earnings report suggests that the company’s true center of gravity is firmly on Earth, even if its ambitions remain pointed at space.

Key figures at a glance

• Starlink revenue: $4.2 billion
• Space-related revenue share: just over 10%
• AI spending in the quarter: $15.8 billion
• Expected neocloud spend next year: $65 billion
• Target annualized revenue run rate: $100 billion