Elon Musk Says AI Revenue Will Eclipse the Rest of SpaceX by September. Here’s What Investors Need to Know

Space Exploration Technologies (NASDAQ:SPCX) CEO Elon Musk just told employees that artificial intelligence (AI) will soon out-earn everything else the company does, like launching rockets or beaming broadband internet down from space. His prediction was made during a 30-minute all-hands video released yesterday on X.

Here’s the full quote: "Probably our AI revenue -- not probably, definitely -- our AI revenue will exceed all other SpaceX revenue probably in September, like next month."

Musk bets big on AI as SpaceX's future growth engine

In the video, Musk called AI "an extremely important part of SpaceX's future" and laid out a target of 10 gigawatts (GW) of AI computing capacity by the end of next year. He put the value of that capacity at roughly $30 to $50 per watt. By his math, that works out to annual revenue of $300 billion to $500 billion from AI alone by 2028 -- "big numbers," as he puts it.

He also described plans to train Grok, SpaceX's AI model, on the company's entire store of internal information, including the work of its employees. "So in a way, it'll be trained on you," he told staff.

Q2 results show AI gaining ground, but it's not there yet

In its second-quarter report, SpaceX brought in total revenue of $7.81 billion, up 92% from a year earlier. The AI segment contributed about $2.56 billion of that, while Connectivity -- the Starlink satellite internet business -- and Space -- the rocket business -- brought in a combined $5.25 billion.

So, while AI growth has been rapid recently thanks to new compute contracts with Anthropic and Alphabet’s Google, the segment needs to more than double, assuming the rest of the business isn’t done growing.

This isn’t out of the question in the short term, given its new contracts, however. Goldman Sachs modeled about $15.6 billion in AI revenue for all of 2026 after SpaceX signed its two massive compute deals.

The road to 10 gigawatts is paved with obstacles

As for Musk’s 10 GW goal by the end of next year, the company has its work cut out for it. Access to power is a huge constraint, and SpaceX is already facing enormous pushback for the methane-producing gas turbines it is using to power its current data centers: Colossus 1 and 2.

And that many chips alone will be difficult to come by as well. By my math, it represents a very sizeable share of all the chips Nvidia will make in that time.

The bottom line

I think the direction Musk is pointing is possible, but the timing and scale deserve skepticism. There is certainly demand, but the ability to build out this much capacity at this point is hard to wrap my head around. It will also require an absolutely enormous capital spend.

SpaceX said it spent over $25 billion last quarter on AI. If it wants to add more than 8 GW of capacity in the next 18 months, it will need to spend a whole lot more than that. The plans leave a pretty big shortfall between what it needs to spend and what it will bring in before 2027, and the company may need to find some outside funding.

And finally, while these contracts are enormous, they’re about as loose as lease deals come. Either party can exit with just a 90-day notice. That would leave SpaceX having spent all that capital without the customer it spent it on.

Johnny Rice has no position in any of the stocks mentioned. The Motley Fool has positions in and recommends Goldman Sachs Group. The Motley Fool has a disclosure policy.