Nvidia-SpaceX Deal Intensifies Chip Scramble, Threatening Neoclouds

SpaceX, the aerospace company led by Elon Musk, has announced its exclusive reliance on Nvidia's chips for its artificial intelligence (AI) computing needs. This strategic decision, emphasizing confidence in Nvidia's forthcoming Vera Rubin architecture, has sent ripples through the AI infrastructure market. Analysts suggest that this exclusive arrangement could grant SpaceX preferential access to Nvidia's already constrained chip supply.

Musk stated that the company plans to build its AI computing infrastructure “exclusively on Nvidia,” praising Nvidia's Vera Rubin architecture as the “best AI computer.” SpaceX aims to significantly expand its AI compute footprint from approximately 2 gigawatts (GW) to somewhere between 5 GW and 10 GW by the end of 2027. This ambitious expansion is critical for projects like Starlink network optimization and autonomous spacecraft navigation, while also serving as part of the company's efforts to commercialize its AI business.

SpaceX is actively moving to monetize its AI computing capacity. The company has reportedly secured agreements with major tech firms like Google and Anthropic to provide access to billions of dollars worth of AI processing power. This positions the AI business to potentially become a significant revenue stream for SpaceX, rather than a financial drain. However, these developments raise concerns for neocloud companies such as CoreWeave and Nebius, whose business models heavily depend on access to Nvidia chips. These neoclouds had previously received equity investments from Nvidia, which implicitly came with preferential access to hardware.

The demand for AI compute power remains exceptionally high, yet the supply of advanced Graphics Processing Units (GPUs) continues to be globally constrained. Bottlenecks in advanced packaging, particularly at foundries like TSMC, are negatively impacting chip availability. Nvidia, in an effort to meet this surging demand and expand its customer base, has been offering financing models, such as minimum revenue guarantees, to neocloud operators to help them acquire GPUs. However, the entry of a major player like SpaceX further intensifies competition within an already challenging supply chain.

Bernstein analysts suggest that Nvidia could bump SpaceX “to the front of the line” when it comes to its limited supply of chips. Stefan Slowinski, an analyst at BNP Paribas, also notes that SpaceX’s aggressive expansion of its AI infrastructure increases the risk of neoclouds like CoreWeave and Nebius being “crowded out.” Slowinski emphasizes that the AI compute shortage is unlikely to ease until at least 2027, and GPU availability could once again become a key constraint in the market. This situation will further intensify competition among AI infrastructure providers and challenge the market position of smaller players.