SpaceX will build all its AI on Nvidia, Elon Musk tells investors

SpaceX Goes All-In on Nvidia

SpaceX will build its next generation of artificial intelligence infrastructure exclusively on Nvidia platforms, Elon Musk confirmed during the company's first earnings call as a public entity. The decision ends the aerospace company’s previous multi-supplier approach to AI computing, standardizing its future on what Musk described as the industry’s best hardware.

"Going forward, we’ve decided to build exclusively on NVIDIA because we think the Vera Rubin architecture is the best AI computer," Musk told investors, as first reported by Yahoo Finance. "We greatly value our close cooperation and partnership with NVIDIA on many levels. So, we’re exclusive to NVIDIA."

A Terrestrial and Orbital Compute Blueprint

The exclusive commitment covers both ground-based data centers and a new class of space-based AI hardware. SpaceX plans to deploy Nvidia’s Vera Rubin NVL72 rack-scale systems on satellites as part of its “Starmind” AI satellite program, according to a Facebook summary of the earnings call.

Musk outlined an aggressive timeline for scaling this infrastructure. The company expects to end the year with more than 2 gigawatts of AI compute capacity and aims for close to 10 gigawatts by the end of the following year, as detailed by Wccftech. This computational power will be concentrated in the “Colossus” infrastructure, a cluster already being commercialized for external customers.

A specific beneficiary is the AI startup Reflection AI, which has signed a compute contract with SpaceX reportedly valued at $150 million per month starting July 1, 2026, and running through the end of 2029, a deal that could total $6.3 billion, according to CNBC. The contract uses Nvidia GB300 systems, the precursor to the Vera Rubin architecture that will anchor future expansion.

Building the Ultimate AI Foundation

The Nvidia partnership is one pillar of a broader “space compute” blueprint. This includes a new 11-million-square-foot “Gigasat” satellite factory in Bastrop, Texas, dedicated to producing orbital AI data centers. A report by Moomoo notes the factory targets deploying 1 GW of space-based AI compute by the end of 2027, with full commercialization in 2028 and an ultimate ambition of reaching 100 GW by 2030.

This pivot to an Nvidia-exclusive strategy streamlines a sprawling hardware ecosystem. Musk’s other ventures, xAI and Tesla, are also heavy Nvidia consumers for frontier model training and autonomous driving programs, a reliance Musk says will continue at scale, per a related video analysis. By standardizing on one architecture, SpaceX simplifies procurement and integration across terrestrial and orbital assets, as noted by Roic.ai.

Market Ripples and Long-Term Ambitions

The exclusive deal reinforces Nvidia’s dominance in the AI hardware supply chain, which has direct consequences for global tech markets. For African startups and cloud providers, Nvidia’s entrenched position, now fortified by space-bound data centers, signals that cutting-edge GPU access will remain expensive and channeled through a few large US-based infrastructure providers for the foreseeable future.

The integration of AI compute with Starlink’s satellite network presents a future where connectivity and processing power are bundled. As Starlink expands in markets like Nigeria, the prospect of low-latency AI services delivered from orbit will raise new questions for regulators like the Nigerian Communications Commission regarding spectrum, data sovereignty, and cross-border data flows.

However, the near-term exclusive commitment does not represent a permanent state. Musk has privately told investor Ron Baron that he is working on an in-house AI chip he claims will be “two to three times better than Nvidia” at about a tenth of the cost, an ambition reported by multiple outlets that suggests a long-term challenge to the very supplier he is now anchoring his strategy on. For now, his public posture is firm: SpaceX’s AI future is printed on Nvidia silicon, from Earth to orbit.