SpaceX Earnings: What SPCX’s First Report Must Prove

Last updated: August 3, 2026, 4:00 p.m. ET

Click here to listen to this episode on Spotify!

SpaceX reports second-quarter results after the market closes on Tuesday, August 4 — the first earnings report in the company’s 24-year history, and the first hard financial disclosure since a June listing that raised more money than any initial public offering ever completed. Wall Street’s consensus calls for roughly $6.8 billion to $6.9 billion in revenue and a loss of about 26 cents a share.

The number that matters most probably isn’t revenue.

Space Exploration Technologies Corp. went public at $135 a share on June 12, valuing the company at approximately $1.77 trillion. It closed its first session at $160.95. Seven weeks later the stock has spent time below $107, roughly half its mid-June peak, and short sellers have built a position equal to about a third of the tradable float. The company’s first-quarter figures, disclosed in the S-1 registration statement, showed a $4.28 billion net loss on $4.70 billion of revenue — and $10.10 billion of capital expenditure in a single three-month period.

That is the tension the August 4 report has to resolve. SpaceX is not a company that needs to prove it can generate revenue. Starlink already does that, at scale and at high margins. It is a company that needs to prove the cash it is pouring into artificial-intelligence infrastructure and Starship development produces something an investor can eventually own a claim on. Two days after the report, on August 6, insiders become free to sell a tranche of stock that some estimates put at roughly 911.5 million shares.

The earnings call, in other words, arrives about 48 hours before the float starts expanding. That sequencing is not an accident of the calendar — the lockup release was structured to trigger off the earnings date — and it goes a long way toward explaining why the market has treated the past three weeks as a referendum rather than a routine quarter.

Key Takeaways

• Main development: SpaceX reports Q2 2026 results after the close on Tuesday, August 4, 2026 — its first quarterly report as a public company following the June 12 IPO.
• Consensus expectations: Analysts project roughly $6.82 billion to $6.9 billion in revenue and a loss of approximately $0.26 per share. S&P Global Market Intelligence models a 35.9% Connectivity-segment margin offsetting operating losses in the Space and AI segments.
• The stock so far: Priced at $135, peaked near $225 in mid-June, and reached an all-time low of $107.01 on July 28. Market data showed SPCX around $108.37 as of August 2, 2026 — roughly 20% below the IPO price and about half the June high.
• Why it matters: SpaceX’s first-quarter net loss of $4.28 billion came alongside $10.10 billion of capital expenditure, $7.72 billion of it AI-related. Morgan Stanley estimates the company will need roughly $84 billion a year in external funding from 2027 through 2034 and won’t reach positive free cash flow until 2035 — while simultaneously maintaining a Street-high $300 price target.
• What comes next: The first major insider lockup tranche opens August 6, two trading days after earnings. Short interest stood near 32% of float in late July.

How SpaceX got from the largest IPO in history to below its offer price in seven weeks

The chronology is unusually compressed, which is part of why the stock has behaved the way it has. Most companies that go public spend two or three quarters establishing a reporting rhythm before the market forms a firm view. SpaceX has had one.

The S-1 landed on May 20, 2026. It was the first time anyone outside the company and its private shareholders had seen audited consolidated financials, and the disclosure did two things at once: it confirmed that Starlink is a genuinely excellent business, and it revealed that essentially all of Starlink’s profit is being consumed by everything else.

Pricing came on June 12 at $135 a share. Demand was extraordinary. Shares opened around $150, closed at $160.95, and by June 16 had run to roughly $225 — a level that implied a valuation well north of $2.9 trillion on the basic share count. Underwriters collected approximately $100 million in fees. The company entered the Nasdaq 100.

Then the direction reversed, and it did so steadily rather than in a single shock. The stock broke below its $135 offer price for the first time on July 15. Bloomberg reported that on July 28 it traded as much as 20% below the IPO price, an erosion the wire service characterized as roughly $1.2 trillion of market value from the peak. It set an all-time low of $107.01 that day. As of August 2, market data showed SPCX at about $108.37 against a previous close of $112.20, with a session range of $107.57 to $113.64.

Three forces did most of the work. The first was a general withdrawal from expensive technology and AI-linked equities in the second half of July, a rotation violent enough to destroy a $45 billion hedge fund in the process — more on that below. The second was Morgan Stanley’s published estimate of SpaceX’s long-run external funding requirement, which gave bears a specific, sourced number to organize around. The third was arithmetic: a staggered lockup structure that will multiply the free float over the back half of 2026, starting two days after this earnings report.

Needham raised its price target to $250 on July 15, and Ark Invest bought about $16.6 million of stock on the decline, so the selling was not uncontested. But the tape has been one-directional enough that the August 4 report is being treated as the first real opportunity for management to change the story.

Timeline: from xAI merger to first earnings

What the S-1 actually showed: three companies wearing one ticker

The most useful way to read SpaceX’s financials is to stop thinking of it as a rocket company. The registration statement breaks the business into three reporting segments, and they are so different from one another in economics, maturity, and capital intensity that consolidated figures obscure more than they reveal.

Connectivity — the part that works

Starlink sits inside the Connectivity segment, and in 2025 it produced $11.4 billion of revenue, about 61% of the company total. It was the only segment that earned money, generating $4.42 billion of segment income. On an EBITDA basis it threw off roughly $7 billion at margins near 63%.

Those are not satellite-industry margins. They are closer to software margins, and they are the reason the IPO was possible at all. The subscriber trajectory explains how it happened: 2.3 million subscribers at the end of 2023, 4.4 million at the end of 2024, 8.9 million at the end of 2025, and 10.3 million by the end of March 2026. Connectivity’s share of total revenue rose to 69% in the first quarter of 2026, meaning the profitable segment is growing faster than the company as a whole.

Morgan Stanley’s model has consumer and business broadband subscribers reaching 16.8 million by the end of 2026. That is the single most important assumption in most bullish SpaceX valuations, and it is worth watching Tuesday’s disclosure closely for the inputs behind it: gross additions, churn, average revenue per user, and the enterprise-versus-consumer mix.

Which brings us to the number almost nobody talks about.

Starlink’s quiet problem: subscribers are up, revenue per subscriber is falling fast

According to disclosures in the S-1 reported by multiple outlets, Starlink’s monthly average revenue per user fell from $86 for the three months ended March 31, 2025, to $66 for the three months ended March 31, 2026. On an annual basis the decline runs from $91 per month in 2024 to $81 in 2025. That is roughly a 23% drop in ARPU year over year in the most recent quarter.

The company does not hide this, and it does not expect it to stop. The filing is reported to state that Starlink anticipates subscriber ARPU will continue to decline over the next several years as the proportion of subscribers outside North America grows, as lower-priced service plans are introduced, and as monthly service fees for broadband offerings are adjusted. Investors should confirm the precise figures against the company’s own quarterly disclosure on Tuesday, which will be the first time these metrics are reported outside a registration statement.

This is the most important dynamic in the entire business, and it is genuinely ambiguous rather than simply bad.

The bear reading is direct: Starlink is buying subscriber growth with price. Headline subscriber counts that look like compounding demand are partly a mix shift toward cheaper markets and cheaper plans. If ARPU keeps sliding at anything like 23% a year, subscriber growth has to accelerate just to hold revenue flat, and the whole edifice of Starlink-funds-everything begins to wobble.

The bull reading is equally coherent, and the 2025 numbers support it. Despite ARPU falling from $91 to $81 across 2025, the Connectivity segment’s operating income rose approximately 120% to $4.4 billion and adjusted EBITDA rose about 86% to $7.1 billion. Falling revenue per user coincided with sharply rising segment profitability. That happens when incremental subscribers cost very little to serve on capacity that is already in orbit and already paid for. A satellite passing over a given patch of ground carries a fixed capacity cost whether it serves one customer or a thousand; the marginal economics of filling unused capacity are extraordinary even at low prices.

So which is it? The honest answer is that both are true at different points on the curve, and the distinguishing variable is capacity utilisation. In markets where the constellation has spare capacity, adding low-ARPU subscribers is close to pure margin. In markets approaching saturation — dense areas of North America, for instance — adding subscribers requires launching more capacity over those specific geographies, and the relevant test becomes whether ARPU in that cell covers the cost of the incremental satellites serving it.

That is the question Tuesday’s disclosure should be read against. A subscriber number that hits Morgan Stanley’s target on the back of low-ARPU international accounts filling spare capacity is a good result. The same number achieved by discounting in saturated markets is not. Investors should look for regional revenue detail, any commentary on capacity constraints, and the enterprise, aviation, maritime, and government mix — those categories carry far higher ARPU than residential consumer plans and are the natural offset to consumer price erosion.

The depreciation question nobody asked before the IPO

There is a second, related issue that private-company status kept comfortably out of view, and it will now appear in every quarterly report SpaceX files.

SpaceX estimates a five-year useful life for Starlink broadband satellites and depreciates them accordingly. Low-Earth-orbit satellites face atmospheric drag and atomic-oxygen degradation, and industry estimates put practical operational lifespans in the five-to-seven-year range. In practice, the company frequently deorbits satellites ahead of their true end of life — reporting indicates roughly 260 were scrapped in a recent six-month window — largely to reduce the risk that degraded hardware compromises the autonomous collision-avoidance system that keeps the constellation safe.

Deorbiting hardware early is sound engineering and responsible orbital stewardship. It is also expensive, and it lands on the income statement.

The arithmetic matters. A constellation of roughly 10,876 active satellites on a five-year replacement cycle implies replacing on the order of 2,000 satellites a year simply to stand still. At reported manufacturing costs near $500,000 per satellite — down from around $1 million originally — plus launch costs, that is a permanent, structural capital expenditure floor that has nothing to do with growth. It is maintenance.

This is the sense in which Starlink’s 63% EBITDA margin can mislead. EBITDA excludes depreciation, and for a business whose core productive asset burns up in the atmosphere every five years, depreciation is not an accounting abstraction — it is the cash cost of remaining in business, merely recognised on a delay. The gap between the 63% full-year 2025 segment EBITDA margin and the roughly 35.9% Q2 Connectivity operating margin that S&P Global models is, to a substantial degree, exactly this.

Two further details compound it. SpaceX sells user terminals below cost to lower the barrier to adoption, meaning each new subscriber carries an upfront loss that must be recovered across the subscription life — a payback period that lengthens as ARPU falls. And the early-generation satellites launched during the initial buildout are now rolling off their depreciation schedules, which flatters near-term reported margins in a way that will reverse as newer, more expensive V3 hardware enters service.

None of this is a scandal, and none of it was concealed; it is in the filing. But it is the sort of detail that private companies are never forced to explain quarterly and public ones are. How management discusses satellite depreciation, replacement cadence, and terminal subsidies on its first earnings call will tell experienced investors a good deal about how candid the next several years of reporting are likely to be.

The launch business: a monopoly that no longer drives the valuation

Launch services and NASA crew work generated $4.0 billion in 2025 — roughly 21% of revenue, and shrinking as a proportion of the whole. This is the business that made SpaceX famous. It is no longer what the valuation rests on, which is a strange thing to write about the most dominant launch provider in history.

The dominance is not in dispute. SpaceX flew approximately 130 missions in 2025, more than every other operator on Earth combined, and reached its 50th launch of 2026 by mid-year. Falcon 9 has sustained a multiple-launches-per-week cadence since 2023 to the point where individual missions no longer register as news even within the industry.

The competitive field has moved, but not by as much as the headlines suggest. Blue Origin’s New Glenn actually exceeds Falcon 9 on payload capacity and flew its first commercial missions in early 2026, picking up a NASA lunar cargo contract worth up to $468 million. It has not flown often enough to demonstrate that capacity translates into repeatable service, and the company has been grounded since a launchpad explosion on May 28, 2026. Rocket Lab has the strongest claim to being a genuine second source — chief executive Peter Beck’s assessment that nobody outside SpaceX and Rocket Lab is demonstrating cadence and reliability is self-interested but not unreasonable — and its backlog has more than doubled year over year as the reusable Neutron vehicle approaches debut. United Launch Alliance completed six launches in 2025 and has said it aims for 20 to 25 in 2026; Vulcan Centaur costs an estimated $100 million to $120 million per flight with no recovery, which on a per-kilogram basis is not competitive with a high-cadence reusable Falcon 9.

So why does a business with those competitive dynamics account for barely a fifth of revenue and almost none of the valuation debate?

Because the commercial launch market is small relative to what SpaceX has become. Global demand for launch simply does not scale to the size required to justify a trillion-dollar valuation, and SpaceX’s most important launch customer is SpaceX — Falcon 9 exists at its current cadence largely to deploy Starlink. The launch segment’s real contribution is not its $4.0 billion of external revenue. It is that it makes constellation deployment cheaper for SpaceX than for anyone attempting to compete, which is the structural advantage underpinning Starlink’s margins and, by extension, everything else.

That vertical relationship is also why Starship’s timeline matters more than the launch segment’s revenue line. V3 Starlink satellites are too large for Falcon 9 to deploy economically at the volumes required.

AI — the reason for the losses

The AI segment arrived through the xAI acquisition, which closed in early 2026 and valued the combined entity at roughly $1.25 trillion. It contributed $3.2 billion of revenue in 2025. It is also where the money is going. In the first quarter of 2026, $7.72 billion of the company’s $10.10 billion in capital expenditure was AI-related. Reporting on the S-1’s quarterly detail indicates xAI generated $818 million of revenue against a $2.47 billion operating loss in that quarter.

Put plainly: Starlink’s profits are funding an AI buildout that currently loses more than it earns, and the market spent seven weeks deciding how it feels about that.

The compute contracts that changed the revenue model

Before the IPO, SpaceX signed two agreements that reframed what the AI segment is. Neither involves launching anything.

The first, disclosed in the S-1 and reported by Axios on May 20, commits Anthropic to pay SpaceX $1.25 billion per month through May 2029 for dedicated access to the Colossus I and Colossus II GPU clusters — a footprint reported at more than 200,000 Nvidia GPUs. Over the life of the agreement that is approximately $45 billion. Payments were structured to ramp in the early months.

The second came on June 5, when CNBC and TechCrunch reported that Google will pay $920 million per month for compute capacity at xAI data centers, running from October 2026 through June 2029 and covering roughly 110,000 Nvidia GPUs plus associated CPU, memory, and infrastructure. Full term, that is about $30 billion. Google characterized it publicly as bridge capacity to meet demand for its Gemini Enterprise agent platform that ran ahead of internal forecasts.

Together those two contracts represent something like $75 billion of contracted revenue, which against a company that booked $18.7 billion in all of 2025 is not a rounding error. They are the strongest single argument the bulls have, and they will be a major focus of Tuesday’s call.

They also carry a detail that deserves more attention than it has received. Both agreements contain exit provisions. The Google contract allows either party to terminate after December 31 with 90 days’ notice; the Anthropic agreement has been reported with a similar 90-day termination right. Contracted revenue with a three-month escape hatch is not the same thing as contracted revenue, and it is reasonable for investors to ask what portion of the AI backlog is genuinely locked.

There is a second-order question, too. Anthropic and Google are both, in the AI model business, competitors of xAI — which SpaceX now owns. The company is simultaneously selling compute to rivals and using that compute to train Grok. That is a workable arrangement, and the cloud industry has lived with versions of it for years, but it introduces a customer-concentration risk with an unusual flavor: the two largest AI-segment customers have every commercial incentive to build their own capacity and exit as soon as it is practical to do so.

The AI segment sells compute, not intelligence — and that distinction matters

There is a widespread assumption that SpaceX’s AI segment is a bet on Grok competing with ChatGPT and Claude. The revenue composition suggests something rather different, and investors should be clear about which business they are actually buying.

By most external estimates, xAI’s standalone AI revenue — model subscriptions, API access, enterprise deployments — runs near $500 million annualized. Anthropic is reported at roughly $47 billion annualized and OpenAI near $24 billion. In U.S. consumer chatbot market share, one January 2026 measure placed ChatGPT at 52.9%, Gemini at 29.4%, and Grok at 17.8%. Anthropic led global large-language-model revenue share in the first quarter of 2026 at 31.4%, narrowly ahead of OpenAI at 29%.

Note that these external estimates are not directly comparable to the segment figures in the S-1, which reported $3.2 billion of AI revenue in 2025 and, per reporting on the filing’s quarterly detail, $818 million of xAI revenue against a $2.47 billion operating loss in the first quarter of 2026. The segment number is broader than model revenue; the gap between them is largely compute leasing.

Which is the point. The Anthropic and Google contracts — $1.25 billion and $920 million per month respectively — dwarf anything Grok generates on its own. SpaceX’s AI business, as it exists today and as it will appear in Tuesday’s report, is overwhelmingly a data-centre landlord renting GPU capacity, and its two largest tenants are the companies whose models are beating Grok in the market.

That is not a criticism. Renting infrastructure to competitors is a fine business — it is roughly how Amazon built AWS, and it carries far more predictable economics than winning a model race against firms with more enterprise traction. Anthropic reportedly wins around 70% of new enterprise deals it contests against OpenAI; xAI has a materially weaker enterprise foothold, and reporting has noted co-founder departures and difficulties converting technical capability into commercial adoption.

But it reframes the risk. If SpaceX’s AI revenue is principally compute leasing, then the segment’s value depends on the durability of tenancy, not on Grok’s competitive standing. And tenancy, as noted, is terminable on 90 days’ notice by counterparties with strong strategic reasons to build their own capacity and the balance sheets to do it — Google alone has guided to $175 billion to $185 billion of 2026 capital expenditure, a fraction of which would replicate what it currently rents.

The orbital data centre programme is best understood as the attempt to make that tenancy durable. If compute in orbit is genuinely cheaper than terrestrial compute at scale, SpaceX owns a cost advantage no hyperscaler can replicate without a launch business. If it is not, SpaceX is a landlord whose tenants are also its landlords’ competitors, on short leases.

What Wall Street is actually forecasting for the second quarter

Consensus revenue sits at roughly $6.82 billion, with S&P Global Market Intelligence’s preview using $6.9 billion. Either figure represents sequential growth of about 45% from the first quarter’s $4.70 billion, driven principally by Connectivity. Consensus earnings per share is a loss of approximately $0.26.

S&P Global models a Q2 Connectivity margin of 35.9%, with that segment’s profitability offsetting operating losses in Space and AI. Note that this is a different measure from the 63% EBITDA margin cited for full-year 2025 — segment operating margin and segment EBITDA margin are not interchangeable, and the gap between them is largely satellite depreciation, which is substantial and rising as the constellation grows.

A caution on the EPS line. SpaceX has been public for seven weeks and has never reported a quarter. Consensus formed on the basis of a single registration statement, which means dispersion around that $0.26 figure is likely far wider than the point estimate suggests, and the reference share count is itself in flux as lockups release. Treating a beat or miss against that number as a meaningful signal would be a mistake. The disclosure quality — segment detail, subscriber metrics, capex phasing, and whatever management chooses to say about the second half — will carry more information than the headline.

The second-half ramp is the real test

Earnings previews have converged on a point that deserves more prominence than a beat-or-miss framing allows: the question is not whether SpaceX clears a $6.82 billion bar in the second quarter. It is whether management can justify the far steeper revenue and earnings trajectory embedded in full-year expectations.

Work through the arithmetic. First-quarter revenue was $4.70 billion. Second-quarter consensus is roughly $6.85 billion. Sustaining sequential growth of that magnitude through the back half of the year requires the AI segment to ramp substantially — and the largest new contract, the Google agreement, does not begin until October. That places an unusual amount of weight on the fourth quarter.

Two consequences follow. The first is that a good second quarter tells you relatively little about the full year, because the hardest growth is scheduled for the period after the one being reported. The second is that the Google contract’s October start makes the fourth quarter the first period in which SpaceX’s AI revenue reflects both major compute agreements running simultaneously — and also the first period in which the December 8 lockup expiry and any year-end contract renegotiations come into view at once.

If management declines to guide, as first-time reporters frequently do, analysts will be left modelling a second-half acceleration on the strength of two contracts with 90-day exit clauses. That is a thin foundation, and it is the single strongest argument for why specific, quantified guidance would do more for the share price on Tuesday than any revenue beat.

The capital expenditure problem

This is the heart of the bear case, and it deserves to be stated at full strength rather than waved at.

SpaceX spent $10.10 billion on capital expenditure in the first quarter of 2026. Annualized, that is north of $40 billion, against full-year 2025 revenue of $18.7 billion. Estimates cited in earnings previews put 2026 capex near $48.7 billion, rising toward $118.4 billion by fiscal 2028.

Morgan Stanley put a number on the cumulative requirement that has since been quoted in nearly every bearish note written about the stock. The bank estimates SpaceX will need approximately $84 billion a year in external funding between 2027 and 2034 — about $672 billion in total — with capital expenditure peaking around $300 billion in 2031 and free cash flow not turning positive until 2035. The bank explicitly warned that if debt markets cannot absorb financing on that scale, SpaceX may be forced to issue equity, reduce growth investment, or slow deployment.

Jim Chanos, the short seller best known for identifying Enron, drew attention to the figure publicly in July.

What makes the Morgan Stanley note genuinely interesting is that the same firm carries an Overweight rating and a $300 price target — the highest on the Street. That is not incoherent, but it does require holding two ideas simultaneously: that SpaceX will consume more external capital over the next eight years than almost any private enterprise in history, and that the terminal value of what it builds justifies paying for the journey. Whether an investor finds that persuasive depends almost entirely on their view of orbital compute, which currently generates no revenue at all.

The $85.7 billion raised at IPO covers roughly one year of the projected funding requirement. That is the arithmetic the market has been absorbing since mid-July.

SpaceX is not the only company spending like this — and that is part of the problem

Context matters here, and it cuts against the idea that SpaceX’s capital programme is uniquely reckless. It also explains why the stock fell when it did.

Amazon, Alphabet, Meta, and Microsoft have collectively guided to roughly $725 billion of capital expenditure in 2026, up about 77% from approximately $410 billion in 2025. Amazon alone projects around $200 billion, Alphabet $175 billion to $185 billion, Meta $115 billion to $135 billion, and Microsoft $110 billion to $120 billion. Goldman Sachs has estimated combined capital spending for those four companies at roughly $5.3 trillion across fiscal 2025 through fiscal 2030.

Against that backdrop, SpaceX’s projected $48.7 billion of 2026 capital expenditure is not an outlier in absolute terms. It is an outlier in one specific respect: the hyperscalers are funding their buildouts largely from operating cash flow generated by enormously profitable existing businesses. SpaceX is funding its buildout from a Starlink segment producing roughly $7 billion of annual EBITDA and an $85.7 billion IPO. Amazon’s free cash flow is projected to turn negative under the weight of its programme; SpaceX’s free cash flow has never been positive and is not forecast to be until 2035.

The timing of SpaceX’s July decline is directly connected to this. On July 28, hyperscalers came under renewed scrutiny after Alphabet’s results were poorly received; Alphabet shares fell about 7% and Amazon, Meta, and Microsoft declined alongside it. Investors were reassessing capital-intensive AI infrastructure spending with uncertain returns across the entire category, and SpaceX — the most capital-intensive, least cash-generative, most recently listed name in that category — took the worst of it. Its all-time low of $107.01 came on the same day.

This suggests something worth holding onto through Tuesday’s report: a meaningful portion of SPCX’s decline reflects a sector-wide repricing of AI capital expenditure rather than a company-specific verdict. That does not make the funding question less real. But it does mean a strong SpaceX quarter may not be sufficient to reverse the move if the broader reassessment of AI infrastructure spending continues, and a weak one will be read as confirmation of a thesis investors have already applied to companies with far stronger balance sheets.

The lockup calendar is the reason this week matters more than a normal quarter

SpaceX did not use the conventional 180-day lockup. It used a staggered structure with tranches releasing at approximately days 70, 90, 105, 120, and 135 after the IPO, plus a price-triggered release.

The first major tranche opens on August 6 — two trading days after Tuesday’s earnings. Reporting indicates insiders become free to sell up to 20% of restricted holdings, roughly 911.5 million shares, worth close to $123 billion at recent prices. Additional releases of about 7% are scheduled around August 21 and September 10. A separate 10% tranche unlocks only if the stock trades 30% above the IPO price, above roughly $175 — a threshold that has moved from plausible to distant since mid-July. The final employee lockup expires December 8, 2026.

Elon Musk’s own shares carry a 366-day lockup and cannot be sold until approximately June 2027, a structural detail that matters for the “founder selling” narrative and one that management will likely emphasize.

Note that reported share counts for the August tranche vary across outlets — some cite roughly 911.5 million shares, others approximately 91.5 million. That is a tenfold discrepancy and it has not been reconciled in public reporting. The larger figure is consistent with the widely quoted $123 billion notional value at prevailing prices; readers should treat the precise count as unsettled pending the company’s own disclosure.

Either way, the direction is unambiguous. The float expands substantially, beginning 48 hours after management speaks. That sequencing gives the call outsized importance: whatever tone management sets on Tuesday evening will be the frame through which the market absorbs the supply that arrives Thursday.

Index inclusion means many holders had no choice

SpaceX entered the Nasdaq 100 shortly after listing, and that detail has consequences worth separating from the fundamental debate.

Index membership compels passive funds tracking the benchmark to hold the stock in proportion to its weight, regardless of any view on valuation, governance, or funding risk. Given SpaceX’s market capitalisation, that weight is substantial. A large cohort of investors is therefore exposed to a pre-revenue orbital-compute programme and a dual-class structure with no shareholder recourse — not because they assessed those risks and accepted them, but because their fund tracks an index that added the company.

This produces two effects that pull against each other, and both are visible in the July tape.

Passive demand provides a persistent bid that does not care about price, which cushions declines and helps explain why a stock down 52% from its peak has not fallen further given the short interest against it. But index inclusion combined with a small free float also amplifies volatility in both directions. When the tradable share count is limited and a third of it is borrowed and sold short, relatively modest changes in active sentiment move the price a long way. That mechanic worked in reverse during the June run to $225, and it has been working against the stock since mid-July.

The lockup expiries change this arithmetic materially. As the float expands through August, September, and December, the same dollar flow moves the price less. Volatility should compress. That is good for long-term holders and bad for anyone counting on a short squeeze — which is another reason the sequencing of Tuesday’s earnings and Thursday’s unlock matters more than either event would in isolation.

It is also worth noting who bought during the decline. Ark Invest added roughly $16.6 million on the dip in mid-July, and Needham raised its target to $250 on July 15, the same day the stock first closed below its IPO price. Conviction buying at lower levels existed; it simply was not large enough to absorb the selling.

A third of the float is short

CNBC reported on July 21 that bearish positions in SPCX had grown to about 32% of the tradable float, with short interest rising from roughly 40 million shares a month earlier to about 206 million. Benzinga put the notional value above $26 billion. Short sellers have reportedly accumulated paper profits in the region of $15.5 billion as the stock declined.

Musk has publicly warned short sellers, a pattern familiar to anyone who followed Tesla through the 2010s. The warning carries a specific mechanical logic here: with a small float and a very large short position, a positive surprise could produce a violent squeeze. That is a real possibility and it cuts both ways — it is also why the lockup expansion matters so much to the bears, since a larger float makes the squeeze arithmetic far less dangerous.

On valuation, SPCX has traded at a price-to-sales ratio near 58, below its post-IPO median but extremely elevated against any conventional industrial or telecommunications benchmark. Price-to-earnings is not calculable; the company is loss-making.

The short thesis is not that Starlink is a bad business. Almost nobody argues that. It is that the market capitalized a decade of speculative orbital-compute revenue into a single quarter of trading in June, and that the funding required to reach that future is large enough to dilute existing holders substantially along the way.

What public shareholders actually bought

If the funding requirement is the central financial risk, governance is the central structural one, and it received less attention during the IPO frenzy than it deserved.

SpaceX listed with a dual-class share structure. Class B shares carry ten votes each; the Class A shares sold to the public carry one. Musk held approximately 85% of voting power before the offering and retains a majority afterward. He can be removed as chief executive, chief technology officer, or chairman only by a vote of Class B shareholders — which is to say, by himself.

The structure goes further than voting control. Reporting on the filing indicates SpaceX adopted mandatory binding arbitration for shareholder disputes, requires shareholders to irrevocably waive the right to a jury trial, and prohibits class actions against the company, its directors, officers, or controlling shareholders. The Harvard Law School Forum on Corporate Governance published an assessment in May under the heading “Top IPO, Weak Governance.”

Reasonable people differ on dual-class structures. The case for them is that they insulate long-horizon strategy from quarterly pressure, and SpaceX is close to the strongest available example: no public market would have funded eighteen years of reusable-rocket development, and a Mars programme is not a quarterly earnings story. Musk’s ability to ignore investors who want him to stop spending is, from one angle, precisely what the bull case requires.

The case against is more concrete than usual here. An investor buying SPCX is funding a capital programme that, on the underwriter’s own estimate, requires $672 billion of external financing over eight years and produces no free cash flow until 2035 — while holding shares that cannot vote out management, cannot bring a class action, cannot demand a jury trial, and are subject to dilution decisions made by a controlling shareholder whose own stock is locked until June 2027.

That combination — extreme capital intensity, an extended path to cash generation, and minimal shareholder recourse — is the reason governance specialists flagged this offering specifically. It does not predict a bad outcome. It does mean the ordinary mechanisms by which public shareholders express dissatisfaction are largely unavailable, leaving the share price itself as the only real feedback channel. Some of what happened in July was that channel being used.

Starship: a good flight, and an unresolved engineering question

Starship’s thirteenth integrated test flight in late July was, by the standard of the program’s history, close to a complete success. The vehicle completed its mission profile end to end and executed a controlled splashdown in the Indian Ocean rather than breaking up. The flight tested new tile attachment methods, modified heat shield designs, and load-sensing tiles able to measure stresses during ascent and reentry. Six satellites carrying cameras photographed the vehicle in flight, giving engineers detailed imagery of thermal protection performance.

Then the analysis started, and it complicated the picture.

Bloomberg reported on July 29 that outside experts examining post-flight imagery identified white streaking that appeared to originate at boundaries between heat shield tiles — a pattern consistent with hot gas penetrating those joints. Dan Rasky, a former NASA engineer who spent nearly four decades studying heat shield materials at the Ames Research Center, was quoted assessing that Starship’s current thermal protection system is “a dead-end for all missions that require full and rapid reusability.”

That is a strong claim from a credentialed source, and it should be weighed against an equally important fact: the flight worked. The vehicle survived reentry. The distinction Rasky is drawing is between surviving a reentry and surviving many reentries in quick succession with minimal refurbishment, which is the entire economic premise of Starship.

Two further constraints sit behind that debate. Starship must be refuelled in orbit to reach the Moon or Mars, and cryogenic propellant transfer has never been demonstrated at anything close to the required scale. And the thermal protection problem is one SpaceX is largely solving alone; reporting on the Bloomberg piece noted that NASA has not invested meaningfully in thermal protection research for decades.

For Tuesday’s purposes, the relevant question is narrower than “will Starship work.” It is whether management provides a credible timeline for Starship reaching operational satellite-launch cadence, because the deployment schedule for the V3 Starlink satellites — and for anything resembling an orbital data center constellation — depends on it.

Why V3 satellites are the hinge between Starship and Starlink

The dependency between Starship and Starlink is the most underappreciated relationship in SpaceX’s financial model, and it is where the engineering risk becomes a revenue risk.

V3 Starlink satellites represent a step change in capacity. Published specifications indicate roughly 1 terabit per second of downlink bandwidth and 160 to 200 gigabits per second of uplink per satellite — figures described as approximately ten times the downlink and twenty-four times the uplink capacity of the current V2 Mini generation. V3 Direct-to-Cell variants are intended to deliver full 5G cellular connectivity from orbit, comparable to a terrestrial experience, rather than the constrained messaging and basic data service available today.

Different sources cite different throughput figures for V3 hardware, and readers should treat the specifics as provisional until the constellation is operating at scale. The direction, though, is not in dispute: V3 is substantially more capable, and substantially larger, than anything currently flying.

That size is the problem. V3 satellites are too large for Falcon 9 to deploy economically at the volumes required. They need Starship. Which means the answer to Starlink’s capacity constraints — the constraint that determines whether the company can keep adding subscribers in saturated markets without margin destruction — runs directly through the vehicle whose thermal protection system a former NASA specialist has publicly described as a dead end for rapid reuse.

Follow that chain through and the stakes on Tuesday become clearer. Declining ARPU is tolerable while spare capacity exists. Spare capacity in high-demand markets requires V3. V3 requires Starship at operational cadence. Starship at operational cadence requires a heat shield that survives repeated reentry with minimal refurbishment. If that last link slips by two years, the ARPU decline stops being a mix-shift story and starts being a margin story — at precisely the moment Amazon Leo enters at $50 to $100 per month.

This is why “when does V3 materially expand network capacity” is a more revealing question than anything about the quarter just ended, and why management’s answer to it deserves more attention than the revenue line.

Regulatory exposure: one million satellites is a policy question, not just an engineering one

SpaceX has asked the Federal Communications Commission for authority to operate a constellation of up to one million satellites for orbital compute. For scale, roughly 10,876 Starlink satellites are currently active, and the total number of objects humanity has placed in orbit across all of spaceflight history is a small fraction of the requested figure.

A request of that magnitude will not be processed as a routine licensing matter. It raises orbital debris and collision-avoidance questions that regulators, astronomers, and other operators have already been raising at current constellation sizes. It touches spectrum coordination through the International Telecommunication Union, where competing operators have standing to object. And it invites scrutiny from national regulators in every market where SpaceX wants to sell service.

The debris question is not abstract for a company that deorbits satellites early specifically to protect the integrity of its automated collision-avoidance system. That practice is responsible, and it is also an implicit acknowledgment that constellation safety degrades as hardware ages — a concern that scales unfavourably with a hundredfold increase in satellite count.

The reminder arrived, conveniently, this week. A spent Falcon 9 upper stage left in space after an earlier mission is on a trajectory intersecting the Moon, with impact expected in the coming days and a debris plume that may be visible from Earth. It is harmless — other spent stages have hit the Moon before, and several missions have done so deliberately — but it is a visible demonstration that hardware SpaceX no longer controls remains in the environment long after its mission ends.

Regulatory approval is not a near-term earnings input, and it will not be resolved on Tuesday’s call. It belongs in the risk assessment of anyone valuing SpaceX on orbital compute, because the timeline for that business is set as much by the FCC as by Starship.

Starmind and the orbital data centre thesis

The strategic logic behind the xAI acquisition is not diversification. It is a specific bet that terrestrial AI compute runs into a power and cooling wall, and that orbit is the way around it.

In the memo accompanying the merger, Musk argued that global electricity demand for AI cannot be met with terrestrial solutions in the near term without imposing hardship on communities and the environment. The project has a name — Starmind — and a first article: a satellite designated AI1, described as functioning as an orbiting server rack delivering roughly 150 kilowatts of peak computing power, spanning about 70 metres tip to tip, and carrying chips running xAI’s Grok models. SpaceX has filed with the FCC for authority to operate a constellation of up to one million satellites.

The vertical integration argument is genuinely strong. A company that builds the rockets, operates the launch cadence, manufactures the satellites, runs the ground network, and owns the AI models is positioned to attempt something no competitor can currently assemble.

The skeptical reading is equally straightforward, and it is mostly about physics and arithmetic rather than ambition. A 150-kilowatt orbital node is a rounding error against terrestrial data centres that consume hundreds of megawatts — the Colossus clusters SpaceX already operates are in that range. Reaching parity would require launching and maintaining an enormous number of satellites, which is precisely why the FCC filing contemplates a million of them, and which in turn depends on Starship achieving rapid reuse. Thermal management in vacuum is a hard problem: there is no air to carry heat away, only radiators. Servicing failed hardware in orbit is not currently practical. And the economics have to beat terrestrial power and cooling by a wide enough margin to justify the launch cost.

None of that makes the thesis wrong. It does mean that orbital compute is, today, a research programme with a valuation attached rather than a business with revenue. Investors paying $1.4 trillion or more for SpaceX are paying for the option, and Tuesday is the first opportunity to hear management quantify the milestones against which that option should be judged.

Competition: Starlink’s lead is real but no longer unchallenged

Starlink operates roughly 10,876 active satellites and serves subscribers across more than 100 countries. That constellation, and the Falcon 9 cadence that built it, constitute a lead measured in years rather than months.

Amazon rebranded Project Kuiper as Amazon Leo in November 2025 and has been deploying since. As of July 2026, approximately 309 production satellites were in orbit across a dozen missions, with commercial service targeted for the United States, Canada, the United Kingdom, France, and Germany. Amazon has committed more than $10 billion, and has signalled consumer pricing in the range of $50 to $100 per month against Starlink’s roughly $120 for standard U.S. residential service.

Amazon’s structural advantages are not in space. They are AWS integration, a Prime base exceeding 200 million customers, and a balance sheet that does not require the satellite business to be self-funding. If Amazon prices aggressively to acquire subscribers, the pressure lands on Starlink’s ARPU — the metric that supports the 63% segment EBITDA margin that makes the rest of SpaceX’s spending possible.

That is the transmission mechanism worth watching. SpaceX’s AI ambitions are financed by Starlink’s margins. Anything that compresses those margins tightens the funding constraint that Morgan Stanley already describes as the central risk.

The AI unwind that framed the quarter

SpaceX’s July decline did not happen in isolation, and the most vivid illustration of the environment came from a fund that no longer exists in its previous form.

Situational Awareness, the AI-focused hedge fund founded by former OpenAI researcher Leopold Aschenbrenner, had returned more than 1,000% since launching in July 2024. In late July 2026 it collapsed from roughly $45 billion in assets to about $10 billion. CNBC reported that margin calls from prime brokers forced a distressed sale of its entire public equity portfolio to Ken Griffin’s Citadel, at a discount. Reported leverage ran as high as 400%; holdings included semiconductor and AI-infrastructure names such as SK Hynix and CoreWeave. TechCrunch reported the firm retained its private positions, including a stake in Anthropic.

The episode matters here for two reasons. It demonstrates how quickly leveraged AI exposure unwound in the back half of July — the same window in which SPCX fell through its IPO price. And it is a reminder that concentrated bets on AI infrastructure carry funding risk that can materialize faster than the underlying thesis resolves. That is, in a different form and on a vastly larger scale, the same question the market is asking about SpaceX.

What history says about mega-IPOs that break their issue price

SpaceX is the largest IPO ever completed, so precise analogies do not exist. But the pattern of a heavily oversubscribed offering that runs hard, peaks within weeks, and then breaks issue is well documented, and the outcomes diverge in instructive ways.

Facebook’s 2012 listing raised approximately $16 billion at a valuation near $100 billion and was widely described as a failure within months as the stock fell well below its $38 offer price. The business, however, was already profitable, cash-generative, and required little incremental capital to grow. The stock recovered because earnings arrived. The lesson usually drawn — that breaking issue price says nothing about long-term value — is true, but it is true specifically because Facebook’s problem was valuation rather than funding.

Rivian is the sharper comparison, and the less comfortable one. It went public in November 2021, raised $11.9 billion at a $66.5 billion valuation, rose nearly 30% on debut, peaked within days, and has traded roughly 77% below its $78 offer price. Rivian’s problem was not that the market misjudged demand for its product. It was that scaling manufacturing consumed vastly more capital than the IPO provided, and every subsequent raise diluted the holders who bought at listing. The trajectory of the business and the trajectory of the shareholder’s return separated, and stayed separated.

Uber sits between them: an $8.1 billion raise at $82.4 billion in May 2019, priced near the bottom of its range, down almost 8% on day one, followed by years of pressure until the company demonstrated it could stop losing money.

The structural question SpaceX poses is which template applies. Its Connectivity segment resembles Facebook — genuinely profitable, high margin, growing fast. Its consolidated financial profile resembles Rivian — enormous capital requirements, negative free cash flow for a decade on the underwriter’s own forecast, and a funding path that runs through repeated external raises.

The distinction that will determine the answer is not whether SpaceX succeeds. It is whether SpaceX succeeds without diluting the people who bought in June. Those are different questions, and only the second one is about the stock. Rivian’s vehicles exist and work; its shareholders lost three-quarters of their money anyway.

The bull case, stated properly

The strongest argument for SpaceX does not depend on orbital data centres working.

Start with what exists. Starlink generated $11.4 billion of revenue in 2025 at roughly 63% EBITDA margins, grew subscribers from 8.9 million to 10.3 million in a single quarter, and is the dominant provider in a category with enormous unserved demand. That business alone, growing at its current rate, supports a substantial valuation. Add a launch business with effective monopoly economics in commercial launch and deep government relationships.

Then add the compute contracts: roughly $75 billion committed across Anthropic and Google, revenue that did not exist eighteen months ago and that arrives at high incremental margin against infrastructure already being built.

The bull case is that the market is applying a conglomerate discount and a funding-risk discount to a collection of assets that are individually excellent, at a moment when a technical selloff and a lockup overhang have created forced selling unrelated to fundamentals. Thirty analysts covering the stock produce 23 Buy ratings, six Holds, and one Sell, with an average target of $239.04 — implying roughly 120% upside from early-August levels. Morgan Stanley’s $300 is the high; Needham moved to $250 on July 15.

The most compelling version of the argument is that SpaceX is the only company that can vertically integrate launch, satellite manufacturing, global connectivity, and frontier AI, and that optionality of that kind is systematically underpriced by discounted cash flow models that cannot handle a decade of negative free cash flow.

The bear case, stated properly

The strongest argument against does not depend on Starlink failing.

It is that SpaceX is a company with $18.7 billion of annual revenue, a $4.28 billion quarterly loss, $40 billion-plus of annualized capital expenditure, and a projected eight-year external funding requirement of $672 billion — carrying a valuation above $1.4 trillion, or roughly 58 times sales.

Every element of the bull case requires capital that SpaceX does not currently generate. The $85.7 billion IPO covers about a year. Morgan Stanley’s own analysis warns that if debt markets cannot absorb the requirement, the company issues equity — diluting holders — or slows deployment, undermining the growth that justifies the multiple. Free cash flow arrives in 2035 on the bank’s estimate, which is eleven reporting years of trusting management through quarterly losses.

Meanwhile the two largest AI customers can exit on 90 days’ notice and are strategically motivated to build their own capacity; Amazon is entering satellite broadband with a cost structure that does not need to be profitable; the thermal protection system that underpins Starship’s economics is described by a former NASA specialist as a dead end for rapid reuse; orbital compute generates no revenue; and roughly $123 billion of insider stock becomes sellable two days after Tuesday’s report.

A third of the float is already short. Those investors are not betting that SpaceX fails. They are betting that the price paid in June was for a 2035 outcome, and that the intervening decade contains dilution.

How do you actually value this company?

Conventional valuation methods struggle with SpaceX, which is part of why analyst targets range so widely — from Morgan Stanley’s $300 to a consensus average of $239.04 to at least one Sell rating, against a share price near $108.

Discounted cash flow, the standard tool, requires cash flow. On the underwriter’s own model, SpaceX produces none until 2035. A DCF that discounts nine years of negative free cash flow before terminal value arrives is essentially a terminal-value calculation with extra steps, and terminal-value calculations are exquisitely sensitive to assumptions nobody can currently test — chiefly whether orbital compute becomes a real market.

Price-to-sales, at roughly 58 times, is the metric most commonly cited. It is close to meaningless in isolation. It would be a preposterous multiple for a launch company and an unremarkable one for a high-growth software business in a strong market, and SpaceX contains elements of both plus a data-centre operator.

The framework that survives contact with the facts is sum-of-the-parts, and it is instructive to attempt it even roughly.

The Connectivity segment generated $11.4 billion of 2025 revenue with approximately $7 billion of segment EBITDA, growing subscribers at better than 15% sequentially. As a standalone business with those characteristics — recurring revenue, high margin, structural competitive advantage, declining ARPU offset by volume — it would command a substantial multiple. Reasonable people could argue for a valuation in the high hundreds of billions.

The Space segment at $4.0 billion of revenue, with monopoly-adjacent economics but a small addressable market, is worth a fraction of that. Call it tens of billions.

The AI segment is where the argument happens. It has roughly $75 billion of contracted compute revenue, terminable on 90 days’ notice, generating operating losses today. Grok’s standalone commercial position is weak relative to Anthropic, OpenAI, and Google. Orbital compute generates nothing.

Add those honestly and it becomes difficult to reach $1.4 trillion — the approximate valuation at early-August prices — without assigning several hundred billion dollars to orbital data centres, a business that currently consists of one satellite design, an FCC application, and a physics argument.

That is not an argument that the stock is overvalued. It is an argument about what an investor is actually paying for. At $108, roughly a third to a half of SpaceX’s market capitalisation appears to rest on a programme with no revenue, dependent on a launch vehicle whose reuse economics are contested, requiring regulatory approval on an unprecedented scale.

People who believe orbital compute works should probably think the stock is cheap. People who do not should struggle to justify it at any price near current levels on the operating businesses alone. The June-to-August decline is best read not as the market changing its mind about Starlink but as the market reducing the probability it assigns to Starmind.

Three ways Tuesday could go

Scenario analysis is not prediction, and none of what follows should be read as a forecast of where the stock will trade. But the range of plausible outcomes is unusually wide, and it is worth mapping.

A clean beat with disclosure. Revenue near or above $6.9 billion, Starlink subscriber additions tracking toward 16.8 million, AI revenue matching the contracted ramp, and — most importantly — specific management commentary on capital expenditure phasing, remaining cash, and V3 timing. In this case the setup matters: with roughly 32% of the float sold short and a relatively small tradable share count, a genuine positive surprise could force a violent covering rally. That is the scenario Musk has been publicly hinting at.

An in-line quarter with vague guidance. Revenue close to consensus, no capital expenditure guidance, expansive language about Mars and orbital compute, little detail on ARPU or capacity. This is the most likely outcome for a first-time reporter and probably the worst one, because it resolves nothing two days before $123 billion of stock becomes sellable. Ambiguity and expanding supply are a poor combination.

A miss on AI revenue. The specific risk flagged by several previews: even if total revenue looks acceptable, AI compute revenue falling short of the contracted ramp would undercut the segment that justifies the valuation premium. Headline revenue could beat and the stock could still fall sharply.

Worth noting that the lockup arrives regardless of which scenario occurs, and that insiders’ willingness to sell into it will itself be information. Heavy selling from early employees at $108 — down 52% from the June peak — would say something about internal conviction that no earnings release can offset.

Material risks

• Funding and dilution. The central risk. An external requirement near $84 billion annually through 2034 implies repeated capital raises. Equity issuance at depressed prices is materially more dilutive than at June levels.
• Customer concentration in the AI segment. Anthropic and Google together represent the overwhelming majority of contracted AI revenue, both with reported 90-day termination rights, and both with strategic incentives to insource.
• Starship technical risk. Rapid reuse is unproven; orbital cryogenic refuelling has never been demonstrated at scale. Both gate the V3 Starlink deployment and any orbital compute constellation.
• Starlink margin compression. Amazon Leo’s entry at lower price points threatens the ARPU underpinning the segment margins that fund everything else.
• Lockup supply. Substantial float expansion between August and December 2026, into a market that has already repriced the stock downward by roughly half from its peak.
• Key-person concentration. Strategy, capital allocation, and public communication are unusually dependent on one individual who also runs several other large enterprises.
• Regulatory exposure. The FCC filing for up to one million satellites, orbital debris considerations, spectrum coordination, and international market access are all live regulatory questions.
• Government revenue dependence. Launch and crew services revenue is meaningfully tied to NASA and defence contracting, subject to appropriations and political risk.

What else has happened since the IPO

Two items outside the earnings frame are worth recording, because both have moved the stock.

On July 31 the Wall Street Journal reported that Tesla executives had been told to prepare for a separation of the company’s China business, and framed it as a step that could clear the way for an eventual Tesla–SpaceX merger — the reasoning being that SpaceX’s role as a U.S. government and defence contractor is difficult to reconcile with substantial Chinese operations. Musk responded on X calling the report “fake news” and denying any discussions over selling or spinning off Tesla’s China business; Tesla’s China unit also dismissed the claims. Bloomberg has not independently confirmed the Journal’s reporting. Speculation about a combination has persisted for months regardless, and Gwynne Shotwell’s non-committal handling of the question when asked directly has been read by some observers as leaving the door open. Nothing here is confirmed, and readers should treat it as a reported claim under dispute rather than a pending transaction.

Separately, a spent Falcon 9 upper stage left in space after a previous mission is on a trajectory that intersects the Moon, with impact anticipated in the coming days. It is not a safety concern — other spent stages have struck the Moon before, and several missions have done so deliberately — but it has renewed attention on orbital debris, an issue with direct regulatory relevance to a company seeking authority for a constellation of up to one million satellites.

What to watch on Tuesday

The headline revenue figure will be reported and analysed within minutes. The material questions take longer to surface, and most of them will be answered on the call rather than in the release.

1. Starlink subscriber additions and ARPU. Is the path to Morgan Stanley’s 16.8 million intact, and what is the revenue quality behind it? Watch churn and the enterprise-versus-consumer split.
2. AI segment revenue against the contracted ramp. The Anthropic agreement was structured to ramp; Google’s begins in October. Whether recognized AI revenue tracks the schedule is the single most consequential line in the release.
3. Capital expenditure phasing and any 2026 capex guidance. The Q1 figure was $10.10 billion. A materially higher Q2 number without corresponding revenue would confirm the bear thesis.
4. Cash position and financing intentions. How much of the $85.7 billion remains, and does management signal an appetite for debt over equity?
5. V3 satellite timeline and Starship cadence. Specifically, when V3 materially expands network capacity, and what the flight-13 heat shield findings imply for the schedule.
6. Second-half guidance, if any. First-time reporters often decline to guide. If management does guide, the gap between Q2 actuals and the full-year ramp implied by consensus becomes the story.
7. Whether Musk appears, and what he says about the lockup. His own shares are locked until roughly June 2027, and saying so plainly would be the cheapest available reassurance.

The communication variable

One factor resists financial modelling entirely, and it would be dishonest to write a preview of this particular earnings call without addressing it.

Elon Musk’s public communication is itself a source of share-price volatility, and the past week supplied a clean example. The Wall Street Journal reported on July 31 that Tesla executives had been told to prepare for a separation of the China business, framing it as a precondition for an eventual Tesla–SpaceX merger. Musk responded on X calling the report “fake news” and “absurdly fake news.” The Journal has not retracted. Investors are left holding a disputed claim about a potential transaction involving both companies, adjudicated through social media posts, four days before an earnings report.

Musk has also warned short sellers publicly — a pattern with a long history at Tesla, where it periodically preceded sharp moves in both directions. With roughly a third of SPCX’s float sold short, that rhetoric has mechanical relevance rather than being merely colourful.

Two things follow for Tuesday. First, whether Musk participates in the call at all is a live question; Gwynne Shotwell, as president and chief operating officer, has historically handled operational communication, and a first earnings call is precisely the venue where a company might prefer its operator to its founder. Second, if Musk does participate, the ratio of specific operational disclosure to long-horizon vision will be read closely. Analysts covering a company that needs $84 billion a year in external funding want capital expenditure phasing and cash balances. A call weighted toward Mars and multiplanetary life would not be new — it would be entirely characteristic — but it would land differently 48 hours before the float expands.

This is not a criticism of ambition, which is the reason SpaceX exists and the reason its constellation was built while incumbents deliberated. It is an observation about audience. The people who funded SpaceX privately signed up for a decades-long vision. The people who bought at $135 in June are now down 20% and hold shares that cannot vote, cannot sue as a class, and will shortly compete with $123 billion of newly tradable stock. Those two constituencies want different things from a conference call.

Frequently asked questions

When does SpaceX report Q2 2026 earnings?

After the U.S. market close on Tuesday, August 4, 2026. It is the company’s first quarterly report as a publicly traded company.

What are analysts expecting?

Consensus revenue of approximately $6.82 billion to $6.9 billion and a loss of about $0.26 per share. Because SpaceX has never reported a quarter, dispersion around those estimates is likely wider than usual.

Why has SPCX stock fallen so far below its IPO price?

Three factors: a broad July selloff in AI-linked technology equities; Morgan Stanley’s published estimate that SpaceX needs roughly $672 billion of external funding through 2034 with no positive free cash flow until 2035; and an approaching staggered lockup expiry that will substantially expand the free float starting August 6.

Is SpaceX profitable?

No. The company reported a $4.28 billion net loss on $4.70 billion of revenue in Q1 2026. The Connectivity segment containing Starlink is profitable — $4.42 billion of segment income in 2025 — but those profits are being consumed by AI and Starship investment.

How much revenue does Starlink generate?

The Connectivity segment produced $11.4 billion in 2025, approximately 61% of company revenue, rising to 69% of revenue in Q1 2026. Subscribers reached 10.3 million by the end of March 2026.

What are the SpaceX AI compute contracts worth?

Anthropic pays $1.25 billion monthly through May 2029, roughly $45 billion in total. Google pays $920 million monthly from October 2026 through June 2029, roughly $30 billion. Both agreements have been reported to include 90-day termination provisions.

When does the SpaceX lockup expire?

The structure is staggered rather than a single 180-day cliff. The first major tranche opens August 6, 2026, two trading days after earnings, with further releases around August 21 and September 10 and a final employee expiry on December 8, 2026. Musk’s shares are locked for 366 days, until approximately June 2027.

How much of SpaceX stock is sold short?

CNBC reported short interest at roughly 32% of the tradable float in late July 2026, approximately 206 million shares, with notional value above $26 billion.

What is the analyst price target for SPCX?

The average across 30 analysts is $239.04, with 23 Buy, six Hold, and one Sell rating. Morgan Stanley carries the Street-high $300 target; Needham moved to $250 on July 15, 2026. These are analyst opinions, not forecasts of where the stock will trade.

Did Starship Flight 13 succeed?

Yes — the vehicle completed its mission profile and splashed down intact in late July 2026. However, Bloomberg reported on July 29 that experts analysing post-flight imagery identified evidence of hot gas penetrating heat shield tile boundaries, raising questions about whether the current thermal protection system can support the rapid reuse the programme’s economics depend on.

Are Tesla and SpaceX merging?

No transaction has been announced. The Wall Street Journal reported on July 31, 2026, that Tesla was preparing to separate its China business in a step that could facilitate an eventual merger. Musk publicly called the report “fake news” and denied the China separation. The claim is disputed and unconfirmed.

Why is Starlink’s ARPU falling?

Monthly average revenue per user fell from $86 in Q1 2025 to $66 in Q1 2026, and from $91 to $81 on a full-year basis between 2024 and 2025. The S-1 attributes this to a growing share of subscribers outside North America, the introduction of lower-priced plans, and adjustments to broadband service fees, and states that the company expects ARPU to keep declining for several years. Notably, Connectivity segment operating income still rose approximately 120% in 2025 despite the decline.

How long do Starlink satellites last?

SpaceX estimates a five-year useful life for depreciation purposes. Satellites are often deorbited before true end of life to protect the constellation’s automated collision-avoidance system. This creates a structural replacement capital expenditure requirement independent of subscriber growth.

Does Elon Musk control SpaceX after the IPO?

Yes. SpaceX listed with a dual-class structure in which Class B shares carry ten votes to Class A’s one. Musk held roughly 85% of voting power before the offering and retains a majority afterward, and can be removed from his executive roles only by a Class B vote.

Can SpaceX shareholders sue the company?

Reporting on the filing indicates SpaceX adopted mandatory binding arbitration for shareholder disputes, a jury-trial waiver, and a prohibition on class actions against the company, its directors, officers, or controlling shareholders. The Harvard Law School Forum on Corporate Governance published a critical assessment of the structure in May 2026.

How does SpaceX’s spending compare with the big cloud companies?

Amazon, Alphabet, Meta, and Microsoft have collectively guided to roughly $725 billion of 2026 capital expenditure, up about 77% year over year. SpaceX’s projected $48.7 billion is smaller in absolute terms but far larger relative to its revenue and cash generation, and unlike the hyperscalers it is not funded from profits of an existing dominant business.

What should investors watch beyond the headline numbers?

Starlink subscriber additions and ARPU, whether AI-segment revenue tracks the contracted ramp, capital expenditure phasing, the remaining cash balance, and any commentary on Starship cadence and V3 satellite deployment.

Final assessment

SpaceX arrives at its first earnings report carrying an unusual burden. Most newly public companies spend a first quarter establishing credibility on small questions. SpaceX has to address the largest question in the market about it — whether a business generating $18.7 billion of annual revenue can responsibly commit to a capital programme that Morgan Stanley sizes at $672 billion — while a third of its float is sold short and $123 billion of insider stock becomes tradable 48 hours later.

The verified evidence points in two directions at once, and honest analysis should say so. Starlink is a genuinely exceptional business: 63% segment EBITDA margins, subscribers growing from 8.9 million to 10.3 million in one quarter, and a constellation lead measured in years. The compute contracts with Anthropic and Google represent roughly $75 billion of demand that did not exist eighteen months ago. These are not promises; they are signed agreements and audited results.

Against that: the first quarter of 2026 produced a $4.28 billion loss and $10.10 billion of capital expenditure, three-quarters of it AI-related. The orbital data centre thesis that underpins the most bullish valuations currently generates no revenue and depends on a Starship reuse capability that a former NASA thermal protection specialist has publicly called a dead end in its current form. The two largest AI customers can leave on 90 days’ notice.

What changed between June and August was not the facts — nearly all of this was disclosed in the S-1 before pricing. What changed was the market’s willingness to fund a decade of losses at a $1.77 trillion valuation. That repricing has already happened; the stock is roughly half its June peak.

The most useful frame for Tuesday is therefore not whether SpaceX beats a consensus formed from a single filing. It is whether management’s first sustained public communication as a listed company demonstrates that it understands the funding question and has a specific answer to it. Investors will accept losses from a company that explains precisely what the losses purchase and when they stop. They react badly to ambition without a timetable.

The unresolved questions are concrete: whether AI revenue is tracking its contracted ramp, whether Starlink’s margins survive Amazon’s entry, whether Starship achieves reuse economics, and how the next $84 billion gets raised. Tuesday will answer the first of those and, at best, gesture at the others. The lockup arrives Thursday regardless.

Sources

• Space Exploration Technologies Corp., Form S-1 Registration Statement, U.S. Securities and Exchange Commission, May 20, 2026
• CNBC: SpaceX IPO takeaways — SPCX closes at $161, jumping 19% after record debut
• CNBC: SpaceX is heavily reliant on Starlink for growth and profit as it marches toward Nasdaq listing
• CNBC: SpaceX snaps 7-day losing streak, sets earnings date that triggers first big share unlock
• CNBC: Bets against SpaceX grow to 32% of float as Elon Musk warns short sellers won’t survive
• CNBC: AI investor Leopold Aschenbrenner forced to unwind all public stock positions after steep losses
• Bloomberg: Experts warn SpaceX Starship heat shield may delay rapid flight plans
• Bloomberg: SpaceX shares rebound after briefly falling 20% below IPO price
• Axios: Anthropic is paying SpaceX $15 billion per year
• TechCrunch: Google will pay SpaceX $920M per month for compute
• TechCrunch: AI hedge fund Situational Awareness may have sold its public portfolio, but it still has its Anthropic shares
• S&P Global Market Intelligence: SpaceX earnings preview, Q2 2026
• Morningstar: SpaceX’s IPO filing — big spending, big losses
• Morningstar: 5 charts on what to watch in SpaceX earnings
• Via Satellite: SpaceX acquires xAI to pursue orbital data center constellation
• Data Center Dynamics: SpaceX IPO filing reveals Anthropic set to pay $1.25bn a month for xAI data center space
• GuruFocus: SpaceX faces $672 billion funding risk, warns Morgan Stanley
• Benzinga: SpaceX stock now has $26 billion in short bets
• Investing.com: SpaceX IPO lockup expiry — $123B in shares set to unlock in early August 2026
• TipRanks: SpaceX (SPCX) stock forecast, price targets and analyst predictions
• Quartz: Leopold Aschenbrenner’s AI hedge fund collapses after margin calls
• Harvard Law School Forum on Corporate Governance: Top IPO, Weak Governance
• Fortune: SpaceX’s supervoting shares put a decades-old governance debate back in play
• The National Law Review: SpaceX IPO raises major governance and investor-protection risks
• CNBC: Amazon, Meta and Microsoft face skeptical investors after Google report sparked sell-off
• CNBC: Tech AI spending approaches $700 billion in 2026, cash taking big hit
• CNBC: SpaceX is poised to be the biggest IPO ever — here are the top U.S. deals to date
• The Motley Fool: The biggest IPOs in history — and how they performed
• The Motley Fool: SpaceX reports its first-ever earnings Tuesday — the five numbers that will define the valuation
• TradingKey: Starlink scraps 260 satellites in six months as depreciation ledger surfaces
• SatNews: SpaceX advances direct-to-device constellation with Starlink V3 hardware
• Trefis via Yahoo Finance: SpaceX vs. Blue Origin, Rocket Lab — what the numbers show
• TheStreet: Analyst sends strong signal on SpaceX stock before earnings
• The Washington Post: Trump threatens more strikes on Iran, and other developments in the Middle East

This article is provided for general informational purposes and does not constitute financial, investment, tax, or legal advice.

Affiliate disclosure: Businessfinance.news may earn compensation from qualifying actions completed through selected links on this website, at no additional cost to the reader. Affiliate relationships do not influence our editorial reporting, analysis, or conclusions.