What changed
NVIDIA announced the Blackwell Ultra platform, including GB300 NVL72 and HGX systems, positioning added memory and compute for inference workloads that spend more tokens on reasoning.
Why it matters
The platform made memory capacity and long-running inference economics central to infrastructure planning for reasoning and agent workloads.
What to do
Model total tokens, latency, power and rack constraints on representative workloads instead of comparing only peak accelerator specifications.
Caveat
Performance, revenue and availability statements were forward-looking vendor claims at announcement, not independent production measurements.