SpaceX's first earnings report since its June initial public offering revealed a company whose center of gravity is quietly shifting from rockets to computing. Capital spending on artificial intelligence infrastructure jumped more than sixfold in the quarter to $18.4 billion, according to figures the company disclosed and reported by CNBC — a sum that outstripped the $7.8 billion in quarterly revenue the company posted, itself up 92 percent from a year earlier.
More than 80 percent of that capital outlay went into AI compute rather than launch vehicles or satellites. The buildout traces to Elon Musk's decision earlier this year to fold xAI into SpaceX, turning Colossus — the Memphis, Tennessee supercomputer complex xAI built to train its Grok chatbot — into a rentable cloud-compute business that now competes directly with Amazon, Microsoft and Google Cloud, as Fortune detailed in an examination of the business.
A Rocket Company's Compute Bet
Executives say the wager is paying off faster than a conventional data center investment. The company's chief financial officer told investors that AI compute deployments were generating a payback period of under a year. Supporting that claim are a string of capacity-leasing deals: Google agreed to pay up to $920 million a month for compute access, while Anthropic committed to as much as $1.25 billion a month for three years to use capacity at the Colossus 1 facility. SpaceX said it signed another $6.7 billion in cloud-computing contracts since the quarter closed, and is targeting a $100 billion annualized revenue run rate by year-end.
Investors were not entirely convinced. Shares fell after the earnings release despite the revenue growth and the payback claims, with analysts flagging the sheer scale of ongoing capital commitments and asking whether compute demand — and pricing — can hold up long enough to justify them. It is the same tension playing out across the AI infrastructure buildout generally, where combined capital expenditure plans across the largest technology companies now run to roughly $600 billion a year, according to estimates cited by eMarketer. Supporters of the spending argue it will accelerate automation and cut the cost of AI services over time; skeptics warn that if demand growth slows even slightly, the industry's newest and biggest bets could sour quickly.
SpaceX's case is unusual mainly for how it arrived at this position. A company built to launch satellites and eventually reach Mars now derives a growing share of its near-term financial story from renting out server capacity, illustrating how thoroughly the AI boom has reordered corporate priorities even among firms with no obvious footprint in the industry a few years ago.
The Custom Silicon Rush
The spending boom is also reshaping who builds the chips underneath it. On August 5, Anthropic confirmed it is assembling an in-house chip-design team, following a report from TechCrunch that the company is recruiting semiconductor engineers with salaries reaching $485,000. Anthropic says the goal is to co-design hardware and models together so future chips run its Claude systems faster and more cheaply, easing the cost of the very compute expansion driving deals like the one with SpaceX.
Anthropic has been careful to frame the move as additive rather than a break from Nvidia, which along with AMD, AWS and Google will remain part of its infrastructure mix. The company has reportedly been scouting Samsung as a potential manufacturing partner for the new chips. The strategy mirrors moves already made by Google, whose Tensor Processing Units now underpin much of its own AI workloads, Amazon's Trainium and Inferentia lines, and Meta's MTIA chips — all designed to chip away at Nvidia's dominant share of AI accelerator spending while giving each company more control over its own compute costs.
Taken together, the two developments point to the same underlying dynamic: as the dollar figures behind AI infrastructure climb, control over both the compute itself and the silicon that powers it has become as central to competitive position in the industry as the models running on top of it.
