NVIDIA has announced the Vera Rubin AI platform, its next-generation AI server. First shipments to major AI customers will begin in July 2025. Pilot production is scheduled to start in June 2025. Mass production is expected in the second half of 2026.
Pilot production starts June 2025
The Vera Rubin platform is built on TSMC's 3nm process node, which began mass production early 2025. Each rack is estimated to cost about $180 million. NVIDIA aims to achieve 40 million times the compute output within 10 years with this platform.
Pricing for each Vera Rubin AI server rack is estimated at 180,000,000 USD (about $180,000,000). Initial availability in North America starts in July 2025. Global mass production shipments are expected to begin in the third quarter of 2026.
Key customers for the Vera Rubin platform include Microsoft, Google, Amazon, Meta, and Oracle. The platform is designed to deliver massive compute performance for AI workloads. NVIDIA has not officially confirmed the production version details, but recent rumors about design or spec changes appear to be based on outdated information.
Discussion
0 comments
Log in to join the thread with a thoughtful take, question, or correction.