A Rocket Company Just Reported a $1.26 Billion AI Loss. That Isn’t a Metaphor.
SpaceX filed its first quarterly earnings report as a public company after the closing bell on Tuesday, August 4 — seven weeks after the largest initial public offering in history. Revenue came in at $7,814,000,000, up 92% from a year earlier and roughly a billion dollars above what Wall Street expected. The net loss narrowed to $541,000,000, or nine cents a share.
All three of the company’s business segments beat revenue expectations. Two of the three lost money on operations — but only one lost it at a scale that reshapes how investors should read the balance sheet. The artificial-intelligence segment booked $2,561,000,000 in revenue and a $1,257,000,000 operating loss, on $15,828,000,000 of capital spending in a single three-month stretch.
The stock fell anyway. And on Thursday, August 6, roughly 911.5 million insider shares become eligible to trade for the first time.
- $7,814,000,000 Q2 2026 revenue, up 92% year over year from $4.07 billion — above the roughly $6.81 billion consensus — SpaceX Form 8-K, SEC EDGAR
- $15,828,000,000 capital expenditure on the AI segment alone in one quarter, against $749 million in the same quarter a year earlier — SpaceX Form 8-K, SEC EDGAR
- $1,257,000,000 AI-segment operating loss for the quarter, even as AI revenue grew 247% year over year — SpaceX Form 8-K, SEC EDGAR
The first thing to understand about this earnings report is that the AI loss is not a figure of speech. In February 2026, xAI merged into SpaceX, folding the Grok model family, the Colossus data-center complex, and the X social platform into the same corporate entity that builds Falcon and Starship rockets. When SpaceX went public four months later, all of it landed on one consolidated income statement. The company now reports three segments — Space, Connectivity, and AI — and files a single set of financials covering all three.
That structure is why a company most Americans think of as a launch provider just disclosed that its largest single line of capital spending went to graphics processors and data centers. Of $18,369,000,000 in total capital expenditure for the quarter, $15,828,000,000 — roughly 86 cents of every dollar — went to the AI segment. Rockets got $1,174,000,000. Satellites got $1,367,000,000.
We posted our second quarter 2026 financial and operational results → Q2 highlights: - Demonstrated the power of extreme vertical integration, delivering revenue growth of 92% year-over-year across Space, Connectivity, and AI - Completed two successful…
SpaceX itself frames the arrangement as a feature rather than a complication, describing what its release calls “the power of extreme vertical integration.” The argument is that one company can launch the satellites, sell the bandwidth, and train the models on compute it owns end to end. The counter-argument, which Wall Street spent Tuesday evening making, is that the same structure means a profitable satellite-broadband business is now funding a capital-hungry AI business inside the same reporting line.
Connectivity is the engine. Starlink and the government-facing Starshield network produced $4,291,000,000 in revenue, up 66% year over year, and $1,656,000,000in income from operations — the only segment in the black. Starlink ended the quarter with 12.0 million subscribers, exactly double the 6.0 million it reported a year earlier, and up 1.7 million in three months. Average revenue per user held at $66 a month, down sharply from $85 a year ago as growth shifted toward lower-priced international tiers.
The Space segment — the original business — brought in $962,000,000 and lost $542,000,000 from operations, a deficit the company attributes almost entirely to Starship research and development it says will eventually cut the cost of reaching orbit by 99% or more. SpaceX flew 38 launches in the quarter and put 485 metric tons into orbit, most of it its own Starlink hardware. The AI segment, meanwhile, turned in the strangest line in the filing: a $1,257,000,000 operating loss that was nonetheless a 49% improvement over the prior quarter, alongside positive segment-adjusted EBITDA of $1,146,000,000— the gap between the two being depreciation on all those newly installed chips.
Total capital expenditure of $18,369,000,000 was more than six times the $2,825,000,000 SpaceX spent in the same quarter of 2025. The AI line alone went from $749,000,000 to $15,828,000,000— a twenty-onefold increase in twelve months. Fortune characterized the quarter as an annualized capex run rate near $73.5 billion against a consensus closer to $48.7 billion. Other outlets publishing the print in real time reported the quarterly capex figure as roughly in line with expectations; the disagreement is noted in the methodology below. What is not in dispute is the direction and the magnitude.
Compute: SpaceX expanded nameplate capacity to 1.4 gigawatts from 1.0 gigawatt in the first quarter and 0.4 gigawatt a year earlier, driven by the continued build-out of Colossus II.
Contracts: The company signed Cloud Services Agreements totaling $14,100,000,000 in contracted sales, which delivered $1,600,000,000of incremental AI infrastructure revenue in the quarter alone — the single biggest reason AI revenue tripled sequentially.
Acquisitions: SpaceX announced an agreement to acquire the AI coding company Cursor for $60,000,000,000, expected to close in the third quarter, and released Grok 4.5 in July.
The balance sheet can absorb it for now. SpaceX ended June with roughly $100,000,000,000 in cash, cash equivalents, and marketable securities, and $47,500,000,000 in backlog — a cash pile assembled from the IPO and from a $25,000,000,000 inaugural investment-grade bond issuance that closed June 26 at a weighted average interest rate of 5.855%. The question analysts kept returning to on the call was not whether the company can fund this year. It was what happens in 2027 and 2028 if the AI segment is still consuming three-quarters of the capital budget.
Chief Financial Officer Bret Johnsen anchored the release on operating leverage rather than the loss line, and the company’s own filing carries his statement verbatim.
SpaceX CFO Bret Johnsen: "2026 has been a momentous year so far, and the second quarter demonstrated the true power of SpaceX. Revenue growth accelerated across all our business segments and we delivered strong operating leverage, with significant margin expansion led by our new…
Chief executive Elon Musk spent much of the call arguing that the market has the satellite business badly mispriced, and, per TechCrunch’s account of the hour, repeatedly stepped over his own executives to say so.
“I think people are really underestimating Starlink. It's not out of the question that at some point Starlink will deliver a majority of the world's internet, at least in countries where we're allowed to operate, which is the vast majority of countries.”
Elon Musk · CEO, SpaceX · Q2 2026 earnings call, August 4, 2026
Musk also told investors that SpaceX’s internal projections for reaching $1,000,000,000,000 in annual revenue had moved forward from 2031 to 2030, with what he called a non-zero chance of 2029. President and Chief Operating Officer Gwynne Shotwell handled the operational case, pointing to more than $6,000,000,000in multi-year U.S. government contracts awarded for Starshield during the quarter — primarily two Space Force awards for low-Earth-orbit communications and sensing constellations — and set a hard date for the program investors most want to hear about: “We want to put boots on the ground — boots on the moon — in 2028.”
SpaceX priced its IPO at $135 a share and began trading on Nasdaq under the ticker SPCX on June 12, 2026. By the time the offering closed on June 15, the company had sold an aggregate 638,888,888 Class A shares for net proceeds of approximately $85,700,000,000 — the largest initial public offering ever completed. CNBC reported the market capitalization topped $2,000,000,000,000 intraday on the debut. As of August 5, the stock traded at $112.20, roughly 17% below its offering price and well under its post-IPO high.
A separate event compounds the pressure and has nothing to do with the earnings themselves. On Thursday, August 6, the first tranche of SpaceX’s staggered lockup expires, making roughly 911.5 million insider shares eligible to trade, with holders permitted to sell up to the first 20% of eligible shares. The public float is currently below 280.1 million shares, meaning the newly unlocked pool is more than three times the stock that trades today. Shares held by Musk and a small group of insiders remain restricted until mid-2027.
Analysts split about where that leaves the stock. Ross Gerber, chief executive of Gerber Kawasaki Wealth and Investment Management, called SpaceX “a compelling investment” in a Bloomberg Television appearance the evening of the report, arguing the Starlink franchise alone justifies the position. Dominic Pappalardo of Morningstar Wealth was more measured: “We want to see some move toward profitability. We know it’s not profitable today and it probably won’t be for some period of time.”
SpaceX beat revenue expectations in every segment, doubled Starlink’s subscriber base, and narrowed its loss — and the stock fell anyway. The reason is legible in the company’s own filing: 86 cents of every capital dollar now goes to an AI business that lost $1,257,000,000in three months, inside a company whose only profitable segment is satellite broadband. After February’s xAI merger, investors buying a rocket company are also buying a data-center build-out — and this week they get to see how many insiders want out.
