NVIDIA Grace Hopper
NVIDIA GH200 Grace Hopper
GH200 combines an Arm Grace CPU with a Hopper GPU via coherent NVLink, targeting memory-bound AI and HPC workloads.
- Launch year
- 2023
- Memory
- 480 GB LPDDR5X + 96 GB HBM3
- Memory bandwidth
- Grace-Hopper NVLink 900 GB/s
- Peak FP16 / FP32
- 400 TFLOPS · 67 TFLOPS
Market snapshot
$1.49 /hr
Range $1.49 – $1.49
Catalog coverage
17 live offerings
Across 1 providers · 17 regions
- Grace CPU + Hopper GPU module
- Huge unified memory
- PCIe and NVLink deployment options
Last refreshed Oct 20, 2025, 1:53 AM
Performance snapshot
Normalized versus NVIDIA A100 (=1.0). Values use public reference benchmarks for training and inference workloads.
- Memory Capacity×2.20
- AI Training×1.80
- CPU Offload×1.50
Provider availability
Price bands per provider pulled from the live catalog.
Lambda Labs17 offers$1.49 – $1.49 /hr
Popular regions
- asia-northeast-11 offers
- asia-northeast-21 offers
- asia-south-11 offers
- australia-east-11 offers
- europe-central-11 offers
- me-west-11 offers
- us-east-11 offers
- us-east-21 offers
Weighted average price $1.49 /hr · median $1.49
- Lambda Labs$1.49/hr
- Lambda Labs$1.49/hr
- Lambda Labs$1.49/hr
- Lambda Labs$1.49/hr
- Lambda Labs$1.49/hr
- Lambda Labs$1.49/hr
- Lambda Labs$1.49/hr
- Lambda Labs$1.49/hr
- Lambda Labs$1.49/hr
- Lambda Labs$1.49/hr
- Lambda Labs$1.49/hr
- Lambda Labs$1.49/hr
- Lambda Labs$1.49/hr
- Lambda Labs$1.49/hr
- Lambda Labs$1.49/hr
- Lambda Labs$1.49/hr
- Lambda Labs$1.49/hr