The latest Data Phoenix news tagged H200.
Dharma AI says a model trained on 272 internal H200 fine-tuning runs estimated job duration more accurately than compute-only baselines, though the…
Nvidia has stopped sending licensed H200 data-center chips to China, narrowing a recently permitted channel for Hopper-generation AI compute.
ByteDance and Tencent have each reportedly received about 10,000 Nvidia H200 AI chips, the first meaningful flow of export-approved H200s into…