SpaceX AI to Surpass $5.25B Launch + Starlink Revenue by September

For an industry that has long measured SpaceX by launch cadence, Starlink subscriber growth, and defense launch awards, Musk's September crossover date is a strategic inflection point. Q2's $2.56 billion AI segment is already nearly half the $5.25 billion launch-plus-connectivity total, and contracted capacity with Anthropic and Google suggests the rocket and satellite businesses may soon become supporting actors in a much larger AI infrastructure play.

Elon Musk has drawn a new strategic boundary inside SpaceX. In a 30-minute all-hands video released on X on August 11, 2026, the CEO told employees that artificial intelligence revenue will 'definitely' exceed all other SpaceX revenue 'probably in September, like next month.' The statement, reported by The Motley Fool as shares of Space Exploration Technologies (SPCX) rose 9.65%, is more than a growth boast. It reframes SpaceX as an AI-compute infrastructure business that also launches rockets and operates Starlink, rather than a launch and satellite company dabbling in data centers.

“

Musk values that capacity at roughly $30 to $50 per watt, yielding an annualized AI revenue run-rate of $300 billion to $500 billion by 2028.

The numbers give the claim a concrete floor. In Q2 2026, SpaceX reported total revenue of $7.81 billion, up 92% year over year. The AI segment contributed $2.56 billion of that, while connectivity and space combined produced $5.25 billion. That puts AI at roughly 33% of total revenue, up sharply but still below the other segments. To overtake them by September, AI revenue must more than double from the Q2 quarterly level if the launch and Starlink businesses simply hold flat. Since the non-AI businesses generated about $1.75 billion per month in Q2, Musk's September crossover implies AI monthly revenue will need to clear roughly that figure—or a higher bar if launch and connectivity keep growing. Goldman Sachs has modeled $15.6 billion in AI revenue for all of 2026, which averages about $1.3 billion a month; the September prediction therefore requires an acceleration well beyond the full-year consensus.

That acceleration path is underpinned by two large compute contracts with Anthropic and Alphabet's Google. Those deals, cited by Musk and reflected in Goldman's model, explain why AI revenue grew quickly from a small base. The announcement also includes a longer-horizon target: 10 gigawatts (GW) of AI computing capacity by the end of 2027. Musk values that capacity at roughly $30 to $50 per watt, yielding an annualized AI revenue run-rate of $300 billion to $500 billion by 2028. To contextualize, SpaceX's Q2 total revenue annualizes to about $31 billion, so the 2028 AI ambition would be roughly 10 to 16 times the company's current total revenue run-rate. Even if one discounts the upper bound sharply, it would make AI compute the dominant value driver in the company.

For investors, the disclosure creates a tension between a near-term crossover and an aspirational capacity target. The September crossover may be achievable because compute contracts can produce step-function revenue recognition as capacity comes online, but revenue from long-term contracts is not always recurring in the same way as Starlink subscriptions or launch backlog. Concentration risk is significant: Anthropic and Alphabet are both massive customers and, in Alphabet's case, a potential competitor in AI infrastructure through Google Cloud. The buildout to 10 GW also faces physical and regulatory constraints—power generation, grid interconnection, land, cooling, and specialized AI chips—that are not entirely within SpaceX's control. Musk's plan to train Grok on SpaceX's internal information, including employee work, adds a workforce and legal dimension, raising questions about privacy, intellectual property, and retention.