NVIDIA AI Updates: May 8, 2026

1. NVIDIA and DOE detail Genesis Mission for AI-driven energy R&D

NVIDIA. US Energy Secretary Chris Wright and NVIDIA VP Ian Buck discussed how AI accelerates energy solutions through the DOE’s Genesis Mission. NVIDIA is building two AI supercomputers at Argonne National Laboratory targeting fusion, materials science, and grid optimization. Source

2. GB200 NVL72 efficiency comes from rack-scale Slurm block scheduling

NVIDIA. A developer-blog deep dive explains how GB200 NVL72 extends NVLink coherence across full racks and why “rack-scale locality” matters — workload performance drops sharply when crossing NVLink domains. The post details Slurm block-scheduling techniques that keep jobs aligned to NVL72 topology for exascale efficiency. Source

3. Walkthrough on post-training quantization with Model Optimizer

NVIDIA. A new how-to demonstrates NVIDIA Model Optimizer for post-training quantization, cutting VRAM and accelerating inference on consumer GPUs. The piece covers calibration, accuracy/recovery tradeoffs, and deployment paths for resource-constrained environments. Source