Subject Line
Starlink’s Profits vs. SpaceX’s AI Capex Shock
Preheader
Q2’s AI spending surprised again, and the lock-up calendar makes August’s supply test immediate.
Meta Description
SpaceX beat Q2 revenue and narrowed losses, but AI capex surged to $18.37B. Starlink remains the profit engine as a major lock-up unlock hits August 6.
Starlink Is Funding a Black Hole
Subtitle: Starlink drove the profit. AI capex consumed most of Q2’s $18.37B spend. That math has a limit.
Markets do not panic because a company is spending money. They panic when the spending outpaces any credible story about how it gets paid back. SpaceX gave investors both problems simultaneously on Tuesday night, and the stock reacted immediately after the report.
The headline read like a victory: revenues of $7.81 billion, up 92% from $4.1 billion, net loss of $541 million, an improvement from the prior year, and adjusted EBITDA of $3.5 billion. The per-share loss narrowed to nine cents versus a higher consensus forecast. Every segment beat. SpaceX also issued full-year revenue guidance for the first time in its 24-year history.
None of it mattered. Shares fell after the report as a surge in AI spending rattled investors and clouded an otherwise expectation-beating quarter.
The Only Number That Moved the Stock
Capital expenditures jumped more than sixfold from a year earlier to $18.37 billion in the second quarter. Most of that, $15.83 billion, was in AI. The total capex figure exceeded the $13.22 billion average analyst estimate, according to FactSet. SpaceX missed that figure by about 39%, and more than double its total quarterly revenue left the building in a single capex line.
The trajectory is the more unsettling data point. In the first quarter of 2026, SpaceX spent $7.7 billion building out AI infrastructure, according to Fortune’s reporting and the company’s disclosures. That number more than doubled in a single quarter.
“The stock is down because the capex for the AI segment was more than double what was expected,” Melissa Otto, head of Visible Alpha Research, told Fortune. The company did not provide forward capex guidance for the rest of the year. Investors are being asked to fund a capital sprint with no disclosed finish line.
Inside the Three-Segment Structure
SpaceX now reports three business lines, and understanding the internal transfer of resources is the only way to understand the investment question.
Connectivity (Starlink): Revenue from the connectivity segment rose 66% year-over-year to about $4.29 billion. Connectivity was the only segment reported to generate operating profit, at $1.66 billion for the quarter. Starlink subscriber figures are widely described as having grown sharply into mid-2026, but the company’s exact quarter-end subscriber count and year-over-year doubling claim were not consistently confirmed across primary reporting, so treat precise subscriber totals cautiously.
The compression, however, is also real. Reports have pointed to declining average revenue per user as Starlink expanded internationally and introduced lower-priced plans, but specific ARPU figures and the timing of any June price increase effects were not consistently supported by primary reporting in a way that justifies a single precise number here.
AI Segment (xAI, X, cloud services): The AI segment brought in $2.56 billion in revenue, up 247% year-over-year, driven by new cloud service agreements. The AI segment’s operating loss was $1.257 billion, while adjusted EBITDA turned positive to $1.146 billion. The losses are narrowing. But $15.83 billion in quarterly capex against $2.56 billion in revenue means the AI segment is consuming resources at a rate no incremental revenue improvement can currently justify on a stand-alone basis.
Space (launch, Starship R&D): The space segment generated $962 million in revenue, up 29% year-over-year. The segment posted an operating loss of $542 million as research and development costs for the Starship program continued to climb.
The internal structure, stated plainly: Starlink connectivity at roughly $4.29B revenue and $1.66B operating income is the only profit engine, and it funds a money-losing AI segment ($2.56B revenue, a $1.26B operating loss, and roughly $15.8B of the $18.37B quarterly capex) plus the Space business ($962M revenue, a $542M loss). With roughly $100 billion of cash and marketable securities, SpaceX has fuel. The question is how many quarters that fuel lasts at this burn rate.
What Management Said and Why It Did Not Land
SpaceX CFO Bret Johnson said on the earnings call that the company is on pace to reach $100 billion in annualized recurring revenue by the end of the year. He also said the company has an additional $6.7 billion of cloud services revenue under contract over a six-month period that begins ramping starting in October of this year.
Musk went further on the call. “To be clear, the $100 billion ARR in December is not a question mark,” Musk said. “That’s what we would achieve if we basically did nothing.” He also said SpaceX plans to build its AI data centers exclusively using Nvidia chips, and he projected compute capacity growth toward 10 gigawatts by the end of 2027.
The message did not resonate. The reason is mathematical. A $100 billion ARR run-rate exiting December implies a near-tripling of the Q2 revenue pace across three segments in two quarters. Even if achievable, the capex required to get there has not been disclosed. Investors are being asked to underwrite a spending program whose scale has already surprised twice in consecutive quarters. The draft’s reference to a specific sell rating, price target, and percent drop from a specific closing price could not be reliably verified from primary reporting, so it has been removed.
The Spending Context: SpaceX Is the Newest Entry in a Crowded Race
Well over 80% of SpaceX’s capex went toward artificial intelligence, where the company is behind major AI labs on models and behind hyperscalers on cloud distribution, while also trying to sell compute capacity.
SpaceX’s position is different in one specific way. Those companies have established, profitable cloud businesses generating tens of billions in operating income to fund the spend. SpaceX’s AI segment produced an operating loss on $2.56 billion in revenue. The cash engine is Starlink, and the trend in per-user economics has been described as weakening even as the base grows.
Beyond the terrestrial data center buildout, Musk has also announced plans for the Terafab chip project. Filings and reporting have described a first phase estimated around $55 billion, with potential scaling toward roughly $119 billion for a larger buildout.
The Cursor acquisition adds another spend layer to account for. SpaceX agreed to buy AI coding assistant Cursor for $60 billion in stock. Cursor, developed by San Francisco-based startup Anysphere, is expected to become a wholly owned subsidiary upon closing the deal in the third quarter of 2026. The draft’s claim about Anysphere’s annualized revenue at announcement could not be verified from primary reporting, so it has been removed.
The August 6 Variable
The capex shock is not the only pressure vector active right now. SpaceX is preparing for a major lock-up unlock, with approximately 911.5 million shares scheduled to become eligible for trading on August 6, 2026, tied to the company’s first earnings release as a public company.
SpaceX went with an unusual lockup approach. Instead of a fixed lockup period after which all insider shares could be sold, the company designed a staggered schedule that spreads sales over multiple dates. On August 6, insiders may sell up to the first 20% of eligible shares. Additional tranches are scheduled later in the summer and fall, but exact dates, percentages, and conditional triggers vary by source and by the prospectus conditions, so treat any single calendar as indicative rather than definitive.
The stock has already seen a sharp decline from its peak since the IPO, and heavy short interest has been widely discussed, though the draft’s specific short-interest characterization and precise percentage drop were not verified in primary reporting and have been softened. The combination of post-earnings disappointment and new float supply arriving Thursday makes the next 48 hours consequential for price discovery in a stock that has had limited public float since June.
Sector Implications
SpaceX’s AI capex posture affects several adjacent trades. Nvidia benefits directly. Musk said the company has decided to build exclusively on Nvidia because it views Vera Rubin as the best architecture. The exclusive chip relationship confirms incremental Nvidia demand on top of already-record hyperscaler ordering.
Pure-play launch competitors gain a cleaner competitive position if SpaceX’s Space segment is effectively deprioritized relative to AI. A company directing the vast majority of its capex toward AI is a meaningfully different competitive threat to other launch providers than it was two years ago.
The broader hyperscaler group faces a read-through on investor patience with unrecouped AI spend. SpaceX is the newest and most aggressive entrant in the capex race, and its post-earnings price action will be monitored closely as a data point on how markets are currently pricing promised AI revenue against actual infrastructure spend.
Options Market Analysis
SPCX has a brief listed options history following its June 12 IPO. Pre-earnings, IV percentile was running in the low-to-mid 20s range based on the available history, indicating options premiums were below the average of the available historical range. That reading made pre-earnings directional purchases relatively inexpensive compared to the actual realized move. The stock’s two-day window from Tuesday open through Wednesday morning represented unusually large price action in either direction.
Post-earnings, IV will compress as the event resolves, but the August 6 lock-up unlock provides an active volatility catalyst within 48 hours. That secondary event prevents a full IV crush and may keep near-term premiums elevated. The skew dynamic is worth monitoring: a stock down sharply from its post-IPO highs with an imminent supply event and a capex disclosure that surprised to the upside will tend to show elevated put demand at near-dated strikes. The draft’s claim about “billions of dollars” of IPO-adjacent options volume and the specific call-flow characterization could not be verified from primary reporting, so it has been removed.
Defined-Risk Framework: Bull Case
For traders expecting the $100 billion ARR target to prove credible and the lock-up supply to be absorbed without structural breakdown: a bull call spread targeting recovery toward the $135 IPO price level, using the October expiry to capture the Q3 earnings cycle where the Cursor acquisition is expected to close and Q3 Starlink economics become clearer. The $6.7 billion in contracted cloud services ramping in October provides a specific, dated catalyst. If you believe the $100 billion ARR is achievable, Q3 is the first opportunity to demonstrate it in reported numbers.
Defined-Risk Framework: Bear Case
For traders expecting lock-up selling to overwhelm absorption capacity and Starlink unit economics to deteriorate: a bear put spread or defined-risk put debit at near-dated strikes captures the August 6 supply event and the next 30 to 45 days of post-lock-up price discovery. The bear case does not require the business to be broken. It requires the supply-demand dynamic for the shares to overwhelm the fundamental improvement story in the near term. Later tranches extend that window.
Neutral Case
If you believe the AI capex trajectory is too uncertain to directionally trade but the Starlink growth and contracted cloud revenue are real: a short iron condor with defined risk on both wings allows premium collection while bounding exposure to both the upside breakout scenario and a downside capitulation event. The next major inflection points are August 6 (unlock), October (cloud revenue ramp and Q3 earnings), and the Cursor deal close.
Risk Analysis
The bull case carries three primary risks. First, if Starlink per-user economics continue to weaken, the sole profit engine of the company softens while the AI build continues unabated. Second, the $100 billion ARR target requires a revenue ramp in cloud services that has not yet been demonstrated at scale. Third, with roughly $100 billion of cash and marketable securities on hand, SpaceX still carries substantial debt and finance lease obligations, and rising leverage against an uncertain free cash flow timeline is a structural concern that compounds with every additional capex quarter.
The bear case’s primary risk is time horizon. Musk has a demonstrated record of executing on capital-intensive visions that appeared implausible at announcement. The lock-up supply event is a one-time float expansion, not recurring dilution. The $6.7 billion in contracted cloud revenue beginning October is not a forecast, it is described as contracted. Even so, Musk’s timelines have a history of slipping, and any multi-year roadmap should be read as an optimistic case, not a firm date.
Forward Outlook
Three dates structure the next chapter.
August 6 is the immediate test. Whether the unlocked shares create sustained selling pressure or are absorbed by institutional buyers will determine the stock’s short-term floor. The draft’s claim about Musk’s exact share count and an absolute lock date could not be verified from primary reporting, so it has been removed, but reporting does suggest meaningful insider restrictions extend into 2027 for some holders.
October is the Q3 earnings window and the Cursor close. The $6.7 billion in contracted cloud services ramps that month. If Starlink unit economics stabilize, and AI cloud revenue begins tracking toward the contracted backlog, the fundamental picture changes materially. This is the quarter that either validates or discredits the $100 billion ARR path.
December is Musk’s self-imposed benchmark. He put a specific number on the record with no hedging. If SpaceX exits 2026 at or near that annualized run-rate, every analysis written this week about unsustainable capex will look premature. If the target slips, the credibility cost is substantial.
Action Checklist
- Monitor the August 6 open closely: Watch volume and price behavior in the first two hours of trading as the lock-up unlock takes effect. Heavy volume with price stabilization suggests institutional absorption. Heavy volume with accelerating decline suggests insider supply is overwhelming demand.
- Track Connectivity margin as the single most important number: Starlink is the profit engine in the reported segment view. If margins weaken while AI spending accelerates, the valuation argument gets harder.
- Watch contracted cloud revenue recognition in October: The $6.7 billion in cloud services contracts is described as beginning to ramp this fall. Early Q3 disclosures or management commentary on the recognition timeline are the most important forward-looking data points between now and the next earnings report.
- Price the lock-up calendar into position sizing: The staggered tranche schedule means supply events can continue through the fall, and some insider restrictions extend into 2027. Treat it as a ladder, not a single cliff.
- Monitor Terafab milestones: The project represents a multi-year capital commitment with no near-term revenue. Delay announcements are negative catalysts. Acceleration announcements, including regulatory approvals and construction progress in Texas, provide the long-duration bull case with concrete dates rather than promises.
- Distinguish Starlink economics from AI promises: The only verified, recurring, profitable business inside SpaceX right now is Starlink Connectivity. Any thesis that relies on AI revenue that has not yet been consistently demonstrated as durable, high-margin revenue is a forward bet, not a present valuation. The two must be priced separately until future quarters validate the cloud ramp.
