Eight weeks after the largest IPO in history, SpaceX shares are back roughly where the story began. In between came euphoria, a brief $2.6 trillion valuation, a collapse of more than half from the peak, the first public quarterly report and one important realization: valuing SpaceX today does not mean valuing a rocket company. It means valuing a Starlink cash engine, a loss-making space infrastructure business and a capital-hungry AI operation at the same time.

Some IPOs are simply about raising capital. Others turn the capital market itself into an experiment. SpaceX belongs to the second category. Trading began on June 12 at $150, fifteen dollars above the offering price. The stock closed its first session at $160.95. Four days later it briefly traded at $225.64. By early August, it had fallen below $105. Today, on August 10, SpaceX is once again trading around $133 — almost exactly where the investment banks originally priced the company.¹²

I find these eight weeks more interesting than the IPO itself. Not because a young stock falling from $225 to $105 is somehow extraordinary. New listings do violent things. What matters is that the market has been forced, for the first time in public, to decide what SpaceX actually is. A space company? A satellite internet provider? A telecom operator? An AI data-center platform? A defense contractor? Or simply a vehicle that combines several of Elon Musk's most capital-intensive ambitions on one balance sheet?

The Largest IPO Ever — and Still a Tiny Float

SpaceX initially sold 555.56 million shares at $135, raising $75 billion. After the underwriters fully exercised the overallotment option, the number of newly issued shares rose to roughly 638.9 million and net proceeds reached $85.7 billion. The company entered the market at a valuation of roughly $1.77 trillion.³⁴

That combination was unusual from the beginning: an enormous valuation, but only a relatively small portion of the total share count available for trading. SpaceX also reserved an unusually large portion of the offering for retail investors. The first trading days therefore became a mixture of price discovery and scarcity. When a company is worth $1.8 trillion but only a fraction of its shares can genuinely change hands, the market price says something not only about enterprise value. It also says something about how scarce the paper happens to be.

That is why I never considered the $225.64 high a particularly useful valuation anchor. At that price, the market capitalization briefly climbed above $2.6 trillion. For a moment, SpaceX was more valuable than companies that have spent decades producing hundreds of billions of dollars in annual revenue. That may eventually prove justified. But a thin float and a famous story over four trading days are not the same thing as fundamental value.

Then the Numbers Arrived — and They Were Better and Worse at the Same Time

On August 4, SpaceX reported its first quarter as a public company. Revenue jumped 92% in the second quarter to $7.814 billion. The net loss roughly halved, from $1.008 billion to $541 million. Adjusted EBITDA rose from $1.214 billion to $3.538 billion.⁵

Normally those numbers would be celebrated. Revenue nearly doubled, the net loss shrank substantially, and adjusted EBITDA almost tripled. The stock fell anyway. That is where the real SpaceX analysis begins.

The headline “SpaceX generates $7.8 billion in quarterly revenue” hides more than it reveals. The SEC filings now show three operating segments: Space, Connectivity and AI. Economically, they could hardly be more different.

The Space segment generated $962 million in revenue and lost $542 million at the operating level. Connectivity generated $4.291 billion in revenue and $1.656 billion in operating profit. AI generated $2.561 billion in revenue — and an operating loss of $1.257 billion. Put differently: the part of the company most people still associate with SpaceX is not currently the part generating the profits.⁴

The cash engine is Starlink.

Starlink Is the Foundation Everything Else Sits On

By the end of June, Starlink had 12 million subscriptions, twice as many as a year earlier. At the same time, monthly ARPU fell from $85 to $66. Normally, that is not a combination I love: more customers, but less revenue per customer. At Starlink, however, the result is still impressive because operating profitability is expanding at the same time.⁶

Connectivity revenue grew 65.8% in the second quarter to $4.291 billion. Operating income rose even faster, by 79.4%, to $1.656 billion. That implies an operating margin of almost 39%. On an adjusted EBITDA basis, the segment generated $2.597 billion — more than 60% of revenue.⁷

But I think the most interesting number sits one level deeper. The incremental revenue did not come primarily from ordinary Starlink households. Compared with the prior-year quarter, Consumer revenue increased by $764 million. Government, Aviation, Maritime and other Enterprise customers contributed $939 million of additional revenue. In absolute dollars, the less visible B2B and government side grew more than the famous satellite internet product for households.

That fits with perhaps the most consequential message from the earnings call. SpaceX no longer wants Starlink to be merely a supplement for dead zones. After acquiring 65 MHz of wireless spectrum from EchoStar for roughly $19.6 billion, the company explicitly plans to build terrestrial mobile infrastructure. President Gwynne Shotwell said the objective is to turn Starlink into a “true mobile service.”⁸

That is a remarkable strategic shift. Until now, Starlink could be viewed as a satellite network with a very large addressable market. SpaceX is now touching business models that have traditionally belonged to Verizon, AT&T and T-Mobile. I would not declare the death of the conventional mobile carrier yet. A terrestrial network cannot simply be replaced by a few satellites. But SpaceX owns something traditional carriers do not: already-global infrastructure above their heads, an in-house launch system and increasingly its own spectrum.

One Footnote in the 10-Q Changes How I Look at the Rocket Business

The SEC filing contains one sentence I consider more important than many analyst price targets: for launches of its own Starlink satellites, SpaceX records no intersegment revenue. Instead, launch costs are capitalized into the satellites. The Space segment therefore shows only customer launches and external development contracts.⁶

That sounds like accounting. Economically, it is much more than that.

SpaceX is vertically integrated. Falcon 9 does not only fly for NASA, the military or commercial customers. A meaningful share of its capacity is used to build the company's own Starlink network. When a SpaceX rocket carries SpaceX satellites, no artificial revenue is created between two internal divisions. That is sensible. But it also means the Space segment looks worse than an independent launch provider with the same utilization might look.

I do not want to beautify the numbers. Space lost $542 million at the operating level in the second quarter, and research and development expense rose 55% to $1.076 billion, largely because of Starship. But part of the economic return on those rockets is harvested elsewhere. In a sense, the losses in the rocket business finance part of Starlink's moat.

That is precisely why conventional sum-of-the-parts models are difficult here. If I value Space independently, I would theoretically need to assign a price to what Starlink would have paid a third-party launch provider for equivalent capacity. If I do not, I understate the economic value of the launch business. If I do, I invent internal revenue that does not exist in the reported financials.

Starship Is Therefore Not a Mars Bet. Not Yet.

For public-market investors, Starship is often discussed through Mars, lunar missions and enormous space stations. I would begin somewhere far more mundane: Starship is first a cost and capacity bet for Starlink.

On July 24, Starship completed its 13th test flight. For the first time, it deployed 20 production-like Starlink V3 satellites, reignited a Raptor engine in space and demonstrated several reentry and landing objectives. The next flight is planned for late August. SpaceX wants to increase reusability further and eventually reach a launch cadence that sounds absurd by today's standards.⁹

For me, V3 is the more important part of that story. If Starlink is the cash-flow engine and V3 delivers materially more capacity per satellite, then Starship is not merely a new spacecraft. It is the truck with which SpaceX plans to expand its most profitable infrastructure more cheaply and more quickly. The Mars option comes for free. I would not base the next few years of valuation on it.

Then There Is xAI — and Suddenly SpaceX Burns Money on Earth

In February, SpaceX acquired xAI in a stock transaction that valued SpaceX at roughly $1 trillion and xAI at about $250 billion at the time. That brought Grok, X and AI data centers under the SpaceX umbrella.¹⁰ Since the IPO, investors have therefore not been buying a pure SpaceX. They have been buying a combined company in which a highly profitable satellite platform helps finance a highly capital-intensive AI expansion.

The Q2 figures make that painfully clear. The AI segment invested $15.828 billion in the second quarter alone. That was roughly 86% of SpaceX's total capital expenditure of $18.369 billion. Space accounted for $1.174 billion of capex, Connectivity for $1.367 billion. In other words: of every dollar SpaceX invested in property, equipment and infrastructure during the quarter, roughly 86 cents went into AI.¹¹

At the same time, the accounting of the AI business is fascinating. It generated $2.561 billion in revenue and lost $1.257 billion at the operating level. Adjusted for depreciation, stock-based compensation and other items, however, it reported $1.146 billion in segment Adjusted EBITDA. The largest bridge between the two measures was $1.885 billion of depreciation in the quarter.¹²

This is exactly the kind of number that makes me cautious. EBITDA can be useful for an ordinary software company. In a business spending billions on GPUs and data centers, depreciation is not an academic footnote. Servers really do age. GPUs really do become obsolete. Power infrastructure costs real money. Positive EBITDA therefore does not automatically mean AI has already become an outstanding capital-return business.

On the other hand, management said after the results that new compute investments currently have payback periods of less than one year, and that SpaceX had signed another $6.7 billion of cloud contracts since quarter-end. CFO Bret Johnsen also pointed to a $100 billion annualized revenue run-rate by year-end.¹³

If that proves true, it would be extraordinary. But it is precisely the kind of claim I want to verify in the next several quarters rather than simply believe. Hyperscalers earn billions because data centers require enormous utilization. If new AI infrastructure truly pays back in less than twelve months, almost every competitor on earth will have an incentive to pour more capital into the market. The moat then has to come from somewhere else: energy, chips, speed, contracts, vertical integration — or a combination of all of them.

Terafab Shows Musk Has No Intention of Solving the Problem in Small Steps

Two days after earnings came the next announcement. SpaceX and Tesla plan to build a vertically integrated semiconductor plant in Grimes County, Texas. The first phase of “Terafab” is expected to cost more than $16.8 billion and create roughly 3,000 jobs. The long-term objective is enormous in-house chip production for AI, autonomous systems and, eventually, perhaps even space-based data centers.¹⁴

I find the sequence almost more important than the factory itself. Tuesday: SpaceX shows Wall Street that 86% of quarterly capex is flowing into AI. Wednesday: the stock is punished for it. Thursday: the first lock-up expires. Shortly afterward: SpaceX and Tesla unveil one of the most ambitious semiconductor projects in the world.

This is not the capital discipline of a traditional public company. It is the capital allocation of a founder who appears to believe that the bottleneck of the next decade will not be demand, but physical compute capacity. Anyone who owns the stock has to accept that excess capital is unlikely to be turned quickly into dividends or buybacks. It will be turned into rockets, satellites, spectrum, data centers and chip fabs.

$85.7 Billion of Fresh Equity Changes the Risk Equation

Amid all the capex, one important counterweight should not be forgotten. The IPO strengthened the balance sheet enormously. At the end of June, SpaceX held $93.5 billion in cash and another $6.5 billion in short-term securities. Against that stood roughly $38.4 billion of financial debt. On a rough basis, that left more than $60 billion of net cash before the remaining commitments related to the EchoStar spectrum transaction.¹⁵

That is the real purpose of this historic IPO. SpaceX did not go public because it desperately needed a few billion dollars for the next rocket. It created a capital base large enough to attack several industries at once.

Capital expenditure in the first half of the year reached $28.476 billion, up from $6.965 billion in the prior-year period. Operating cash flow, meanwhile, improved from $351 million to $3.466 billion. It is a strange profile: the operating business is improving materially, but investment is accelerating even faster.¹⁶

In a conventional value stock, that would worry me. At SpaceX, that is exactly the bet. The capital market is paying today for assets that are expected to generate revenue tomorrow. The critical question is therefore not whether 2026 free cash flow looks attractive. It is: what return will SpaceX earn on the $28.5 billion it is installing today once that infrastructure is mature three or five years from now?

The Lock-Up Was Fundamentally Meaningless — and Still Hugely Important for the Stock

On August 6, additional blocks of shares became eligible for sale for the first time. Up to roughly 912 million shares became tradable, with further tranches to follow. By mid-2027, a total of another 12.9 billion shares are scheduled to come out of restrictions. Musk himself is subject to a longer restriction and, according to Reuters, cannot sell until one year after the IPO.¹⁷

Fundamentally, a lock-up expiry changes nothing. No Starlink customer cancels because an engineer is allowed to sell shares. No Raptor engine becomes worse. No data center loses compute power. But market mechanics can change dramatically. A company whose early valuation was supported by a tiny float suddenly faces real supply.

Interestingly, the expected wave of selling did not immediately arrive. The stock rebounded sharply. On Friday, August 7, retail investors became net sellers for the first time since the IPO, according to Vanda data — but only by $4.5 million. Before that, retail buyers had spent weeks buying almost every dip. Their average purchase price was estimated at roughly $147.¹⁸

I consider that healthier than the first few trading sessions. The more shares genuinely circulate, the less the stock price is a product of artificial scarcity. A lower price with real liquidity can be more informative than a higher price with a constrained float.

What Are You Actually Paying for at $133?

At the time of this analysis, roughly 13.18 billion Class A and Class B shares are outstanding. At approximately $132.68 per share, that implies a market capitalization of about $1.75 trillion. Subtract the roughly $61.6 billion of net cash represented by cash, short-term securities and financial debt, and a highly simplified enterprise value lands near $1.69 trillion.¹⁹

This is where the valuation becomes uncomfortable.

If I simply annualize Q2 revenue of $7.814 billion, I get a little more than $31 billion of revenue. The enterprise value would equal roughly 54 times that annualized revenue. On annualized adjusted EBITDA, the multiple is roughly 119. Of course, that is an unfair calculation: SpaceX is growing far too quickly for one quarter to be extrapolated sensibly over a full year. But that is exactly why the exercise is useful. It shows how much future is already embedded in the price.

The market is not valuing SpaceX on what it earns today. It is valuing a version of the company in which Starlink becomes much larger, mobile works, AI compute scales, Starship reduces launch costs and perhaps, at some point, space-based data centers become real. Investors are paying today for several unfinished business models at once.

To me, this calculator illustrates the valuation problem better than any fixed price target. Even if I assume $250 billion of revenue in 2030 — more than eight times a simple annualization of Q2 — I still need a 35% EBITDA margin and a 30x multiple to get close to justifying today's valuation. Push revenue to $500 billion or $1 trillion and the stock can suddenly look cheap. But at that point the analysis is essentially built around the assumption that Musk's most extreme revenue targets come true.

That is why I think the question “Is SpaceX cheap at $133?” is poorly framed. A better question is: what combination of Starlink, AI and Starship has to happen for $133 today to produce an attractive long-term return?

What Surprised Me Positively

Before the earnings report, I expected Starlink to be growing quickly. I did not expect the Connectivity segment to already produce an operating margin close to 40% while ARPU was falling materially. That suggests extraordinary scalability. Even more interesting is the Enterprise and Government growth. Starlink appears to be evolving from a consumer product into a platform for airlines, ships, governments and corporations, where willingness to pay can be structurally higher.

The balance sheet also gives SpaceX enough strength that it does not immediately depend on friendly capital markets to finance its ambitions. Roughly $100 billion of cash plus securities is a remarkable cushion. A company running close to $30 billion of half-year capex should be grateful to have one.

Finally, the vertical integration is real. Many companies talk about it. SpaceX builds rockets, satellites, terminals, communications networks, AI data centers and now potentially key parts of the semiconductor supply chain. That creates complexity and concentration risk. But it also creates something competitors will struggle to replicate: optimization across system boundaries.

What Bothers Me

The valuation is the obvious one. At $1.75 trillion, there is little room for an ordinary good outcome. SpaceX has to become extraordinary. The stock can fall 40% from its high and still remain expensive — the last several weeks have demonstrated exactly that.

Second, I do not automatically love the blending of Musk's companies. The xAI transaction may prove strategically brilliant. It may also mean that the highly profitable Connectivity engine continues funding projects that sit outside the original SpaceX mandate. The fact that SpaceX already purchased hundreds of millions of dollars of Tesla Megapacks in the first half and is now building Terafab jointly with Tesla does not make the structure easier to analyze.²⁰

Third, the AI business is extraordinarily capital intensive. If $15.8 billion of quarterly capex truly pays back in less than a year, this will become one of the most impressive infrastructure ramps in modern economic history. If not, Wall Street will eventually stop treating every new data center as a future moat.

And fourth, Musk remains a governance risk. According to Reuters, he controls roughly 82% of voting rights after the IPO. From a founder's perspective, that is consistent. As a minority shareholder, however, I need to understand what I am buying: I am not investing in a company where every major capital-allocation decision can be constrained by an independent shareholder vote. To a large degree, I am investing in Musk's judgment about which industry should be vertically integrated next.

The Numbers I Will Watch Next

• 1. Starlink ARPUCan margins keep expanding even as Consumer revenue per subscription declines through international expansion?
• 2. Enterprise & GovernmentDoes this less visible part of Starlink remain the larger absolute growth contributor?
• 3. AI ContractsDo announced compute contracts turn into recurring revenue with attractive returns on capital?
• 4. CapexDoes AI continue absorbing 80%+ of investment, or does the mix normalize after the current build-out?
• 5. StarshipWhen do successful test flights turn into reliable operating launch capacity for V3 Starlink?
• 6. MobileHow concrete does Starlink's terrestrial mobile strategy become following the EchoStar spectrum acquisition?
• 7. Lock-UpsWho sells in the next release windows — employees diversifying, or long-term core holders?
• 8. Cash ReturnsHow quickly do today's billions of dollars in investment convert into operating cash flow?

My Conclusion: I Like the Company More Than the Stock

I came away from SpaceX's first public earnings report with more respect for the company than I had before. Not because of Mars. Not because of Elon Musk. But because Starlink already looks economically far stronger than I expected for a business growing this quickly. A Connectivity segment generating $4.3 billion of quarterly revenue, an operating margin close to 39% and a double-digit-million subscriber base is no longer a future story.

At the same time, the results made me more cautious about the stock. With xAI, SpaceX has embedded a second, completely different capital cycle inside the company. The profitable satellite business now helps finance both rocket development and AI infrastructure. That may become a brilliant industrial architecture. It may also mean that every new dollar of cash flow immediately becomes the foundation for the next $20 billion project.

Perhaps that is the most important lesson from the first eight weeks of public trading. SpaceX has spent years challenging gravity technically. In the stock market, gravity works differently: the higher expectations climb, the harder every disappointment pulls them back down.

The stock has fallen more than 40% from its high and still trades around its IPO level. That sounds paradoxical. I think it is healthy. For the first time, the market is no longer discussing only how large SpaceX might one day become. It is beginning to ask how much of that future is already priced in at $133.

That is where the stock becomes interesting to me. Not as a bet on whether the next rocket lands. But as a bet on whether one of the most unusual vertically integrated infrastructures in the world can eventually produce enough cash flow to make one of the most unusual valuations in stock-market history look reasonable in hindsight.