- Company leaders expect capital spending to stay at the elevated level for the next two quarters. It indicates the infrastructure buildout remains unfinished.
- Company to build its AI computing platform exclusively on Nvidia’s Vera Rubin architecture.
SpaceX held its first earnings call as a public company on Tuesday to deliver a bold message. The company is positioned to dominate the artificial intelligence computing race. Its secret weapon is its rocket scientists.
During the second quarter alone, Elon Musk invested nearly $16 billion in AI infrastructure. He subsequently defended the company’s astronomical spending unapologetically.
His argument was simple and sweeping. The discipline that keeps rockets safe also covers building and operating data centres. Thus, the approach applies to data-centre challenges as well.
Musk did not mince words when describing how he views the competitive landscape. He compared SpaceX’s edge over data centre rivals to the Yankees facing a Little League team. This comparison did not name which competitors were in the farm system.
Data centres, Musk insisted, represent a “trivial problem” when measured against the daily work of SpaceX’s aerospace engineers.
We are applying a small portion of rocket and satellite engineering to scaling terrestrial data centres. He said this on the call. Even a fraction of that engineering firepower, he argued, produces “an amazing outcome” when redirected toward AI infrastructure.
The bravado is striking given SpaceX’s status as a relative newcomer to the data centre arena. The industry has been dominated by cloud computing titans for a long time. Amazon Web Services, Microsoft Azure, and Alphabet-owned Google Cloud posted record growth in their latest cloud earnings.
A swiftly growing line of “neocloud” startups races to expand capacity. Thus, this effort targets powering the next generation of AI workloads.
Numbers behind the ambition
The scale of SpaceX’s investment is breathtaking by any measure. Total company-wide capital expenditures reached $18.4 billion in the second quarter. They represented a sharp rise from $10 billion in the first three months of the year. Company executives told analysts that capital spending is expected to remain at roughly that elevated level over the next two quarters, signalling that the infrastructure buildout is far from over.
The spending spree follows SpaceX’s February acquisition of xAI, the artificial intelligence company also owned by Musk, and represents a decisive expansion beyond the company’s roots in aerospace and satellite broadband.
AI revenue reached $2.6 billion in the second quarter, a 213 per cent jump sequentially, and the company disclosed $14.1 billion in contracted cloud services agreements — early proof that customers are willing to bet on SpaceX’s computing ambitions. Among the deals powering the nascent AI cloud business are contracts with Anthropic and with Google.
Betting big on Nvidia’s Vera Rubin
Another pillar of SpaceX’s confidence comes from its relationship with Nvidia. Additionally, Musk stated that the AI platform will be built exclusively on Nvidia’s Vera Rubin architecture, which he described as industry-leading.
“Our understanding with Nvidia is that we will receive a very significant percentage of their GPUs next year,” he said, underscoring the depth of the partnership with the chipmaker in a market where access to cutting-edge silicon is a defining competitive moat.
From Gigawatts to grand visions
The company’s compute ambitions are equally outsize. SpaceX ended the quarter with 1.4 gigawatts of nameplate compute capacity and told investors it expects to surpass 2 gigawatts by year end. But when pressed by an analyst, Musk went further, floating what he described as a “tentative target” of 20 gigawatts of power and cooling capacity online by the end of 2027.
Even allowing for significant delays, he said, the company would likely land “close to 15 gigawatts” — a scale that would represent a dramatic reshaping of the AI infrastructure landscape.
While SpaceX has previously explored the concept of space-based data centres capable of harvesting solar energy directly in orbit, Musk kept Tuesday’s discussion grounded firmly on terrestrial facilities. The focus for now, he made clear, is on winning the Earth-bound compute race.
“Rocket science is literally our daily business,” Musk said, noting that engineers spend every flight trying to prevent a rocket from doing what it “desperately wants to do,” which is blow itself apart. Applying even a sliver of that mindset to data centres is, in his view, “kind of ridiculous, frankly,” given the relative simplicity of the problem.