Needham: WFE & Memory
WFE Requirements for Vera Rubin PODs> 1 Million GPUs Impact: Building out greenfield fab capacity to support 1 million Rubin GPUs (roughly 5 GW of data center capacity) drives $14 billion in WFE, split roughly evenly between DRAM/HBM and logic.> WFE Intensity Floor: The total WFE intensity for a Vera Rubin POD sits around 9%, which is expected to act as the intensity floor moving forward.> Growth in Spend: Compared to an individual Rubin GPU package, a full Vera Rubin POD sees a 225% total WFE increase, driven heavily by surges in advanced logic (up 256%) and DRAM/HBM (up 164%).Silicon Content & Ratio Shifts> Total Die Count: Nearly 20,000 Nvidia dies are packed into a single Vera Rubin POD.> Wafer Area Shift: The rise of low-latency inference and agentic AI has shifted the DRAM-to-logic ratio (wafer area) from 7:1 down to 4:1.> Supply Dynamics: While traditional DRAM remains in short supply, the market is seeing advanced logic supply tighten correspondingly.> POD Subtotals: Across its 40 total racks (spanning NVL72, Groq 3 LPX, Vera CPU, Bluefield-4 STX, and Spectrum-6 SPX), a single Rubin POD requires roughly 1,384 total 300mm wafers, dominated heavily by DRAM (958.6 wafers) and advanced logic (248.9 wafers).$NVIDIA(NVDA.US) $Micron Tech(MU.US) $DRAM $EWY
The copyright of this article belongs to the original author/organization.
The views expressed herein are solely those of the author and do not reflect the stance of the platform. The content is intended for investment reference purposes only and shall not be considered as investment advice. Please contact us if you have any questions or suggestions regarding the content services provided by the platform.
Post your comment
No Comments