Elon Musk announced that SpaceX and xAI will exclusively use Nvidia graphics processing units (GPUs) going forward, citing the company's Vera Rubin architecture as superior to all competing designs. During SpaceX's earnings call on August 4, 2026, Musk revealed ambitious plans to deploy Nvidia's latest hardware not only in terrestrial data centers but also in space, marking a significant bet on the chipmaker's technology for powering next-generation artificial intelligence (AI) systems.

Why Is Nvidia's Vera Rubin Architecture So Important?

Musk's endorsement centers on Nvidia's Vera Rubin NVL72 VR200 rack-scale system, which he described as "the best AI computer" available today. The architecture's most distinctive feature is its cable-less design for compute trays, a seemingly technical detail that has major practical implications. By eliminating manual cable and hose connections, the system improves assembly speed, serviceability, and reliability while reducing installation errors and failures.

"We think the design of the NVL72 VR[200] computer is a much better design than, say, having a standard rack style design," Musk stated during the earnings call. "We greatly value our close cooperation and partnership on many levels with Nvidia. We are exclusive to Nvidia".

While xAI has historically relied on Nvidia's Hopper and Blackwell accelerators to train multiple generations of its Grok AI model, the formal exclusivity announcement represents a public commitment that signals confidence in Nvidia's roadmap. The company's Colossus supercomputers, which power Grok training, have used Nvidia hardware for years, making the exclusivity declaration more of a formality than a dramatic pivot.

What Are SpaceX's Plans for Orbital AI Compute?

Perhaps the most ambitious aspect of Musk's announcement involves deploying Vera Rubin hardware in space. SpaceX plans to launch what it calls the "Starmind AI satellite," essentially an optimized version of the Vera Rubin NVL72 computer, beginning next year. This represents a far more aggressive deployment than Nvidia's own Space-1 Vera Rubin Module, which is designed to place several Rubin AI accelerators in orbit.

The full NVL72 system includes 36 Vera CPUs and 72 Rubin AI GPUs, offering substantial computing power. However, significant engineering challenges remain regarding cooling and reliability in the space environment, where traditional data center infrastructure cannot function.

How Is SpaceX Building Its AI Infrastructure on the Ground?

SpaceX's terrestrial AI ambitions are equally impressive. The company disclosed that it will bring approximately 2 gigawatts of compute capacity online by the end of 2026, with plans to reach between 10 and 20 gigawatts by the end of 2027. To put this in perspective, a gigawatt represents one billion watts of power, enough to supply a city of several hundred thousand people.

The company's AI division, SpaceXAI, is already generating significant revenue. During the second quarter of 2026, SpaceXAI contributed $2.6 billion to SpaceX's total revenue of $7.8 billion, which exceeded analyst expectations of $6.93 billion. More impressively, SpaceXAI has already contracted $6.7 billion in cloud computing deals in the first weeks of the third quarter, with sub-one-year payback periods.

• Compute Deployment Timeline: SpaceX expects to have 2 gigawatts operational by end of 2026, scaling to 10 to 20 gigawatts by end of 2027
• Revenue Model: Only 10 percent of SpaceXAI's compute capacity is reserved for training Grok, with the remainder available for inference or commercial rental
• Cloud Deal Momentum: The company has already secured $6.7 billion in cloud contracts in early Q3 2026, demonstrating strong enterprise demand

What's Driving the Memory Shortage Crisis?

During the earnings call, Musk highlighted a critical bottleneck affecting the entire AI industry: memory supply. He noted that memory output is increasing by approximately 20 percent annually, which would normally be considered exceptional growth. However, demand for memory is rising at roughly 200 percent per year, creating a severe supply-demand imbalance.

"If you've got demand increasing much faster than supply then Economics 101 would suggest that the price increases. It does not decrease," Musk explained, underscoring why companies like SpaceX are aggressively securing Nvidia hardware and investing in data center infrastructure.

How to Understand SpaceX's Broader AI Strategy

• Vertical Integration: SpaceX is building its own data centers, satellites, and AI infrastructure rather than relying on cloud providers, giving it control over costs and supply
• Satellite-Based Compute: By deploying AI compute in orbit, SpaceX can offer low-latency services globally and potentially bypass terrestrial infrastructure limitations
• Revenue Diversification: SpaceX is monetizing excess compute capacity through cloud deals while reserving capacity for its own Grok AI model development

Musk also projected that SpaceX's total revenue, not annual recurring revenue, will reach $1 trillion by 2030, one year earlier than previously forecast. The company expects to achieve more than $100 billion in annual recurring revenue by December 2026.

The exclusive partnership with Nvidia represents more than a simple hardware procurement decision. It signals that SpaceX and xAI are betting heavily on Nvidia's technological leadership in AI accelerators, even as competition from AMD and other chipmakers intensifies. By publicly committing to Nvidia's architecture and deploying it in space, Musk is making a statement about where he believes the future of AI compute lies.