Rack & Pod Scale
AI GPU Platforms
Enterprise AI runs on systems that combine many GPUs into a single product. Compare the rack-scale and pod-scale configurations that power today's frontier AI models. See their average FLOPS, power draw, and bandwidth side by side, and understand the machines behind world-class AI training and inference.
Filters
Available Systems
| Compare | |||
|---|---|---|---|
HW Huawei
Atlas 650E Atlas· rack | 12.5 PFLOPS FP4 | View → | |
HW Huawei
Atlas 950 SuperPoD Atlas· pod | 2.00 EFLOPS FP4 | View → | |
NVIDIA
DGX Rubin NVL8 DGX· rack | 400 PFLOPS NVFP4 | View → | |
NVIDIA
Groq 3 LPX LPX· rack | 307 PFLOPS FP8 | View → | |
NVIDIA
HGX Rubin NVL8 HGX· baseboard | 400 PFLOPS NVFP4 | View → | |
AMD
Helios Helios· rack | 2.90 EFLOPS FP4 | View → | |
Google
TPU 8i Pod TPU· pod | 10.34 EFLOPS FP4 | View → | |
Google
TPU 8t Superpod TPU· pod | 121 EFLOPS FP4 | View → | |
NVIDIA
Vera Rubin NVL72 NVL· rack | 3.60 EFLOPS NVFP4 | View → | |
HW Huawei
Atlas 800I A3 Atlas· rack | 4.5 PFLOPS FP16 | View → | |
HW Huawei
Atlas 800T A3 Atlas· rack | 6.0 PFLOPS FP16 | View → | |
HW Huawei
Atlas 900 A3 SuperPoD Atlas· pod | 614 PFLOPS INT8 | View → | |
HP HPE
Compute XD690 Compute XD· rack | 144 PFLOPS FP4 | View → | |
NVIDIA
DGX B300 DGX· rack | 144 PFLOPS FP4 | View → | |
NVIDIA
DGX GB300 NVL72 NVL· rack | 1.44 EFLOPS FP4 | View → | |
NVIDIA
DGX SuperPOD GB200 SuperPOD· pod | 46.08 EFLOPS FP4 | View → | |
NVIDIA
GB200 NVL32 NVL· rack | 640 PFLOPS FP4 | View → | |
NVIDIA
GB200 NVL36 NVL | 720 PFLOPS FP4 | View → | |
NVIDIA
GB200 NVL576 NVL | 11.52 EFLOPS FP4 | View → | |
NVIDIA
GB200 NVL72 NVL· rack | 1.44 EFLOPS FP4 | View → | |
NVIDIA
HGX B200 8-GPU HGX· baseboard | 144 PFLOPS FP4 | View → | |
NVIDIA
HGX B300 8-GPU HGX· baseboard | 144 PFLOPS FP4 | View → | |
AMD
Instinct MI355X Platform Instinct Platform· baseboard | 80.5 PFLOPS FP8 | View → | |
HP HPE
ProLiant Compute XD685 (HGX B200) ProLiant Compute XD· rack | 144 PFLOPS FP4 | View → | |
HP HPE
ProLiant Compute XD685 (HGX B300) ProLiant Compute XD· rack | 144 PFLOPS FP4 | View → | |
HP HPE
ProLiant Compute XD685 (HGX H200) ProLiant Compute XD· rack | 15.8 PFLOPS FP8 | View → | |
HP HPE
ProLiant Compute XD685 (Instinct MI355X) ProLiant Compute XD· rack | 80.5 PFLOPS FP8 | View → | |
Google
TPU7x (Ironwood) Pod TPU· pod | 42.52 EFLOPS FP8 | View → | |
AWS AWS
Trn3 UltraServer UltraServer· rack | 362 PFLOPS FP8 | View → | |
CS Cerebras
CS-3 CS-3· rack | — | View → | |
CS Cerebras
CS-3 Inference Cluster (16-node) CS-3· cluster | — | View → | |
CS Cerebras
CS-3 Inference Cluster (32-node) CS-3· cluster | — | View → | |
CS Cerebras
CS-3 Inference Cluster (64-node) CS-3· cluster | — | View → | |
CS Cerebras
CS-3 Inference Cluster (8-node) CS-3· cluster | — | View → | |
CS Cerebras
CS-3 Inference Cluster (88-node) CS-3· cluster | — | View → | |
CS Cerebras
CS-3 Rack CS-3· rack | — | View → | |
HP HPE
Cray XD670 Cray· rack | 15.8 PFLOPS FP8 | View → | |
NVIDIA
DGX B200 DGX· rack | 144 PFLOPS FP4 | View → | |
NVIDIA
DGX H200 DGX· rack | 15.8 PFLOPS FP8 | View → | |
AWS AWS
EC2 trn2.48xlarge EC2 Instance· rack | 41.0 PFLOPS FP8 | View → | |
GI Gigabyte
G593-SD1-LAX3 G593· rack | 15.8 PFLOPS FP8 | View → | |
SU Supermicro
GPU SuperServer AS -8125GS-TNHR (HGX H100) GPU SuperServer· rack | 15.8 PFLOPS FP8 | View → | |
SU Supermicro
GPU SuperServer AS -8125GS-TNHR (HGX H200) GPU SuperServer· rack | 15.8 PFLOPS FP8 | View → | |
SU Supermicro
GPU SuperServer SYS-821GE-TNHR GPU SuperServer· rack | 15.8 PFLOPS FP8 | View → | |
Intel
Gaudi 3 HLB-325 Baseboard Gaudi· baseboard | — | View → | |
NVIDIA
HGX B100 8-GPU HGX· baseboard | 56.0 PFLOPS FP4 | View → | |
NVIDIA
HGX H200 HGX· baseboard | 15.8 PFLOPS FP8 | View → | |
DE Dell
PowerEdge XE9680 PowerEdge· rack | 15.8 PFLOPS FP8 | View → | |
Google
TPU v6e Pod TPU· pod | 470 PFLOPS INT8 | View → | |
AWS AWS
Trn2 UltraServer UltraServer· rack | 164 PFLOPS FP8 | View → |
per page
1/3
Rent These Systems
Cloud providers offering managed instances with these GPU systems.
| Provider | Instance | Price/GPU-hr | Checked | |
|---|---|---|---|---|
DataCrunch | 1H100.80S.30V (spot) | $1.63 | 3h ago | View → |
Cudo Compute | sapphire-rapids-h100 | $1.79 | 3h ago | View → |
RunPod | H100 SXM (Community Cloud) | $2.69 | 3h ago | View → |
Massed Compute | gpu_1x_h100 | $2.73 | 3h ago | View → |
DataCrunch | 1H100.80S.30V (on-demand) | $3.25 | 3h ago | View → |
RunPod | H100 SXM (Secure Cloud) | $3.29 | 3h ago | View → |
Scaleway | H100-SXM-2-80G | $3.83 | 3h ago | View → |
Lambda Labs | gpu_2x_h100_sxm5 | $4.19 | 3h ago | View → |
DigitalOcean | gpu-h100x1-80gb | $4.41 | 3h ago | View → |
Paperspace | H100 | $5.99 | 3h ago | View → |
Oracle Cloud (OCI) | H100T | $10.75 | 3h ago | View → |
DataCrunch | 4B200.120V.CC (spot) | $3.12 | 3h ago | View → |
Beyond.pl | NVIDIA DGX B200 | $5.10 | 3h ago | View → |
Vast.ai | — | $5.31 | 3h ago | View → |
RunPod | B200 (Community Cloud) | $5.98 | 3h ago | View → |
DataCrunch | 2B200.60V.CC (on-demand) | $6.23 | 3h ago | View → |
RunPod | B200 (Secure Cloud) | $6.79 | 3h ago | View → |
Oracle Cloud (OCI) | B200 | $14.00 | 3h ago | View → |
DataCrunch | 1H200.141S.44V (spot) | $2.00 | 3h ago | View → |
RunPod | H200 SXM (Community Cloud) | $3.59 | 3h ago | View → |
DataCrunch | 1H200.141S.44V (on-demand) | $4.00 | 3h ago | View → |
DigitalOcean | gpu-h200x1-141gb | $4.47 | 3h ago | View → |
RunPod | H200 SXM (Secure Cloud) | $4.59 | 3h ago | View → |
Oracle Cloud (OCI) | H200 | $10.00 | 3h ago | View → |
DataCrunch | 1B300.30V.CC (spot) | $3.83 | 3h ago | View → |
RunPod | B300 (Community Cloud) | $6.94 | 3h ago | View → |
DataCrunch | 1B300.30V.CC (on-demand) | $7.65 | 3h ago | View → |
RunPod | B300 (Secure Cloud) | $7.89 | 3h ago | View → |
Scaleway | B300-SXM-8-288G | $8.68 | 3h ago | View → |