When Nvidia’s most powerful chips are already harder to get than a dinner reservation at Dorsia, the last thing CoreWeave and Nebius needed was another heavyweight cutting in line.
SpaceX’s exclusive partnership with Nvidia for its AI infrastructure buildout could bump Elon Musk’s rocket company to “the front of the line” for chip allocations, according to analysts at Bernstein. That’s a problem for the so-called neoclouds, the upstart AI-focused cloud providers whose entire business models depend on getting Nvidia silicon before everyone else.
The GPU queue just got longer
SpaceX announced in early August that it would exclusively use Nvidia chips to power its AI compute needs. CoreWeave and Nebius have each received $2 billion equity investments from Nvidia, investments that historically came with the implicit perk of preferential access to Nvidia’s latest and greatest hardware. Both companies have been among the first to deploy cutting-edge platforms like the GB200 NVL72 and Blackwell-based systems, giving them a competitive edge over hyperscalers that spread their chip procurement across multiple vendors.
Both CoreWeave and Nebius have targeted more than 5 GW of AI compute capacity by 2030. That kind of scale requires a steady, predictable flow of next-generation GPUs. Any delay in receiving new silicon doesn’t just slow expansion, it risks losing customers to competitors who can deliver compute capacity on time.
Why Nvidia loves this deal
Bernstein raised its price target on SpaceX from $239 to $248 following the company’s Q2 2026 earnings, citing its AI revenue outlook as a key driver. Nvidia’s stock has also seen positive reactions tied to the partnership announcement.
But Nvidia is playing a delicate balancing act. The company has invested billions in CoreWeave and Nebius precisely because those neoclouds serve as distribution channels for Nvidia hardware into the broader AI economy. Small and mid-sized AI companies that can’t negotiate directly with Nvidia rely on neoclouds to provide access to GPU clusters. If CoreWeave and Nebius can’t scale fast enough because chips are being rerouted to SpaceX, that downstream demand goes unserved, and Nvidia’s total addressable market shrinks at the edges.
The neocloud squeeze
CoreWeave and Nebius occupy a peculiar position in the AI infrastructure stack. They’re not hyperscalers like AWS, Google Cloud, or Azure, which have the leverage to build custom chips and negotiate from positions of enormous scale. Instead, neoclouds differentiate by offering purpose-built GPU clouds optimized specifically for AI workloads, essentially betting everything on Nvidia’s roadmap.
The broader GPU supply chain remains constrained despite Nvidia’s efforts to ramp production. TSMC’s advanced packaging capacity, particularly the CoWoS technology essential for assembling Nvidia’s most advanced chips, has been a persistent bottleneck. Adding a customer the size of SpaceX to the priority list means the math gets harder for everyone else, not easier.