SpaceX’s Super Intelligence business generated $7.5 billion in third-quarter revenue and $431 million in profit, according to TechCrunch’s reporting on the company’s regulatory disclosure. The segment combines Grok, the recently acquired Cursor coding tool, and a neocloud infrastructure business reported to have 1.7 gigawatts of compute capacity.
That puts AI at just over half of SpaceX’s reported $14.6 billion in quarterly revenue. It is a substantial business by the disclosed numbers, although the combined segment does not reveal how much each product contributes or whether its infrastructure investment is generating cash.
The distinction matters because Elon Musk paired those results with a much larger prediction: annual recurring revenue of $500 billion to $1 trillion by the end of 2027. The quarter establishes meaningful revenue and a reported segment profit. It does not establish that trajectory.
The $7.5 Billion Combines Three Different Businesses
SpaceX’s October 20, 2026, Form 8-K identifies an earnings release covering the quarter ended September 30. The segment figures discussed here come from TechCrunch’s account of those results; the filing’s cover document confirms the reporting period but does not itself reproduce the financial breakdown.
Super Intelligence bundles businesses with different economics. Grok is an AI product, Cursor sells AI-assisted coding tools, and the neocloud operation supplies computing infrastructure. Their combined revenue measures the breadth of the business, not the commercial performance of any single model or application.
TechCrunch describes neocloud as the most important contributor to the segment’s revenue, but its report does not provide separate dollar amounts for Grok, Cursor, and infrastructure. Readers looking for a standalone xAI revenue figure should therefore be cautious about using the $7.5 billion total as a substitute.
That missing split changes the competitive interpretation. Strong infrastructure sales can demonstrate demand for computing capacity without establishing that Grok subscriptions or API usage are driving equivalent growth. Conversely, grouping everything together can obscure whether software sales are producing attractive margins alongside a more capital-intensive infrastructure operation.
Using the rounded reported figures, Super Intelligence supplied approximately 51% of SpaceX’s quarterly revenue. “More than half” is the defensible takeaway, rather than an overly precise percentage derived from rounded inputs.
The $431 Million Profit Needs an Accounting Definition
The reported $431 million profit represents a margin of approximately 5.7% on $7.5 billion in revenue. That is a useful starting point, but TechCrunch’s account does not specify enough about the segment measure to equate it with net income, operating income, or free cash flow.
The largest identified complication is $1.2 billion in one-time stock-based compensation related to the Cursor acquisition. According to the report, that expense reduced the AI segment’s profit.
Mechanically adding $1.2 billion back to $431 million produces approximately $1.63 billion. That arithmetic illustrates the charge’s size; it is not a separately verified adjusted-profit measure and should not be presented as SpaceX’s underlying earnings without a proper reconciliation.
Stock compensation also deserves more scrutiny than the phrase “one-time” sometimes receives. It generally does not require an equivalent cash payment when recognized, but it remains compensation with potential consequences for shareholders through dilution. An acquisition-related charge may be unusual without being economically irrelevant.
The quarter therefore supports a narrower conclusion than either an enthusiastic or dismissive reading would suggest. The segment reported a profit despite a large compensation expense. The available breakdown does not establish how much cash it generated or what margin it can sustain after further expansion.
The 1.7-Gigawatt Footprint Signals Scale, Not Utilization
The reported 1.7 GW of compute capacity is one of the clearest indications of the infrastructure operation’s scale. It is also easy to misinterpret.
Gigawatts describe power capacity, not a standardized quantity of AI computation. The useful computing output depends on the hardware, networking, workload, and how much of the facility is operating. Capacity alone does not establish customer utilization or revenue per installed accelerator.
TechCrunch’s report does not provide a breakdown showing how much of the 1.7 GW is actively serving customers, used internally, or otherwise available. It should not automatically be read as fully occupied, revenue-producing capacity.
For a vertically integrated AI business, the allocation matters. Infrastructure can support internal model development and inference, external customers, or both. Those uses produce different financial signals, and a combined segment disclosure does not separate them.
The competitive advantage would come from deploying and selling useful capacity faster or more economically than rivals. The reported footprint suggests SpaceX has built at considerable scale. It does not, by itself, prove superior infrastructure returns.
The Capital Plan Is Much Larger Than the Quarterly Profit
TechCrunch reports $19.6 billion in total capital expenditures alongside the results, with CFO Bret Johnsen forecasting another $35 billion to $40 billion in capital investment during the fourth quarter.
Those are company-wide figures, not a disclosed AI-only spending budget. The report connects the investment program to the Colossus II data center project and to Starship, which SpaceX sees as important for future Starlink deployments and orbital data centers. Assigning the entire amount to Grok or the neocloud business would overstate what the disclosure establishes.
Still, the scale of planned spending is central to understanding the earnings. Accounting profit and capital expenditure measure different things. Spending on long-lived assets generally reaches the income statement over time through depreciation rather than being deducted entirely when the cash is spent. A company can therefore report a profitable quarter while consuming substantial cash on construction and equipment.
The reported figures are not enough to calculate AI free cash flow. That would require, among other things, a clearer allocation of capital spending and operating cash flows to the segment.
Musk’s explanation for the infrastructure push also deserves attribution. According to TechCrunch, he argued that power, cooling, and infrastructure were becoming greater constraints on deploying advanced AI than chips. He described using SpaceX’s engineering and metal-casting capabilities to accelerate access to turbine components.
That is a plausible industrial strategy, but the account does not quantify its cost advantage or demonstrate how much faster it makes deployment. The evidence supports substantial spending and an effort to remove physical bottlenecks, not a verified lead in infrastructure efficiency.
Musk’s 2027 ARR Claim Is a Different Category of Number
Sources
- TechCrunch’s reportingtechcrunch.com
- October 20, 2026, Form 8-Ksec.gov
