H200
NVIDIA's Hopper-generation GPU, a memory-upgraded H100 that serves as the common baseline generation in fleet, confidential-computing, and cross-vendor performance comparisons.
Current numbers
~288 GBHBM4 per Rubin GPU at ~22 TB/s; trajectory H100 80 GB → H200 141 GB → HGX B200 180 GB → B300 288 GB, flat into Rubin, then Rubin Ultra 1 TB
~3–12 moH100/H200 GPU lead times through 2025–2026 (historical bimodal context; current mature-supply system quotes run ~4–12 wk — see the lead-times tracker): spot H100s improving to weeks while H200 and full enterprise systems ran 6–12 mo; why GPUs are owner-furnished and on allocation