SpaceX to Fully Adopt NVIDIA Architecture Amid Memory Supply Bottlenecks
Elon Musk announced that SpaceX will fully adopt NVIDIA's architecture platform, specifically praising the next-generation Rubin architecture. He projected that SpaceX's compute capacity will exceed 2 gigawatts (GW) by the end of this year and could reach up to 10 GW by the end of next year. This transition highlights the critical role of NVIDIA's hardware in scaling massive aerospace and AI compute infrastructures. Furthermore, Musk's warning about memory supply growing at only 20% compared to a 200% demand surge underscores a severe industry-wide hardware bottleneck. Musk identified memory, rather than raw compute power, as the primary limiting factor for scaling AI, noting that economic principles dictate rising prices due to this supply-demand mismatch. The decision to commit to NVIDIA is heavily driven by the capabilities of the upcoming Rubin architecture.
## BACKGROUND
NVIDIA's Rubin architecture is a next-generation GPU platform designed to address end-to-end bottlenecks in AI workloads by integrating GPUs, CPUs, and high-speed networking. High Bandwidth Memory (HBM) has become a major bottleneck in AI hardware because modern AI models require massive data transfer rates that traditional memory technologies cannot support.