SpaceX Targets 10 GW AI Compute, $500B Revenue by 2028

Elon Musk used a 30-minute all-hands video released on X on August 11, 2026, to tell SpaceX employees that the company's artificial intelligence revenue will "probably in September, like next month" exceed all other SpaceX revenue combined. The statement, covered by The Motley Fool on August 12, assigns a hard near-term crossover date and sets an aggressive path: Musk wants 10 gigawatts of AI computing capacity by the end of 2027, which he values at roughly $30 to $50 per watt, implying $300 billion to $500 billion in annual AI revenue by 2028. The timing is important because SpaceX's Q2 2026 results show the AI segment is growing quickly but has not yet overtaken the rest of the company. SpaceX reported total Q2 revenue of $7.81 billion, up 92% year over year. AI contributed about $2.56 billion, while Connectivity (Starlink) and Space (rockets) generated a combined $5.25 billion. On a monthly basis, the AI segment averaged roughly $0.85 billion in Q2, while the rest averaged about $1.75 billion. For Musk's September crossover to occur, AI revenue would need to more than double from that Q2 monthly average, assuming the other segments do not decline. The recent compute deals with Anthropic and Alphabet's Google are the likely drivers, and Goldman Sachs has modeled about $15.6 billion in SpaceX AI revenue for all of 2026, which would imply a monthly run-rate near $1.3 billion for the year but with a steeper exit rate in the second half. The source notes SPCX moved 9.65% higher, underscoring investor appetite for an aerospace leader with an AI compute growth story. For investors, the shift matters because it re-rates SpaceX from a capital-intensive launch and broadband operator to an AI infrastructure company with compute monetization. However, the transition is not without risks. The 10 GW target depends on power availability, grid interconnection, chip supply, and construction timelines; the article's final section explicitly frames the road to 10 gigawatts as paved with obstacles. Customer concentration is another concern: much of the near-term AI revenue appears tied to just two hyperscalers, Anthropic and Google. If those contracts are renegotiated or absorbed by in-house capacity, the trajectory could slip. Musk also said Grok, SpaceX's AI model, will be trained on the company's entire store of internal information, including the work of employees. He framed this as "in a way, it'll be trained on you." That announcement expands the AI story from infrastructure to internal model development, but it raises data governance, privacy, and security questions for a company with sensitive launch and defense-adjacent work. It also suggests SpaceX may be positioning Grok as a proprietary productivity and operational model, potentially reducing reliance on external AI vendors. Looking ahead, the key metric to watch is whether AI quarterly revenue can reach around $5.25 billion or more in Q3 2026 to validate the crossover claim. If Goldman's $15.6 billion full-year estimate is realized, SpaceX could exit 2026 with an AI run-rate above $2 billion per month. The real test, however, is the 10 GW buildout; even half of Musk's $300B-$500B annual revenue target would transform SpaceX's financial profile, but execution timelines in physical infrastructure often slip. Investors and AI competitors should monitor contract expansion, compute capacity disclosures, and any Grok enterprise deployment as leading indicators.