LIVETracking 971 AI models from 268 providers.
Beta
NVIDIA logoNVIDIA

NVIDIA DGX GB300 NVL72

Full rackAnnouncedBlackwell Ultra2025

Next-gen liquid-cooled rack with Blackwell Ultra GPUs. 72x B300 GPUs with 288 GB HBM3e per GPU (vs 192 GB on B200). Designed for reasoning-heavy workloads and the largest language models. 50% more memory per GPU enables larger context windows and batch sizes.

72
GPUs per system
1440 FP8 PFLOPS
All systemsCompute hub
GPU model
NB
NVIDIA B300
GPU count
72x
CPU model
NVIDIA Grace Ultra
36x
Memory type
H
HBM3e
Total HBM
20.7 TB
Host memory
26 TB
Interconnect
NVLink 5.0+
180 TB/s
Networking
3200 Gbps
Storage
120 TB NVMe SSD
Form factor
Full rack (42U)
Weight
1,500 kg
Rack units
42U

Manufacturer datasheet values · aggregate system compute

FP4 PFLOPS2,880
FP8 PFLOPS1,440
FP16 PFLOPS720
BF16 PFLOPS720
Training effective PFLOPS1,080

Thermal envelope · cooling requirements · efficiency

Rack power
140 kW
Per GPU
1200 W
Cooling
liquid
PUE estimate
1.1
Power draw relative to tracked systems140 kW / 2500 kW max
10.29 FP8 PFLOPS per kW · average across all systems is 4.81

Hardware amortized over 3 years · power at $0.05/kWh

List price
$4,000,000
Per GPU effective
$55,556
Cost per GPU per month
$1,543
TCO per PFLOPS per year
$973
PFLOPS per kW
10.29
89% below the average TCO of $9,046/PFLOPS/year across all tracked systems
Available from
NVIDIA logoNVIDIA

Every data point on this page is reproducible

Compare across the system landscape

Google logoTPU v5p Pod
8960x Google TPU v5p · 8100 PFLOPS
Pod / clusterShipping
NVIDIA logoDGX GB200 NVL72
72x NVIDIA B200 · 720 PFLOPS
Full rackShipping
Google logoTPU v6e Pod
256x Google TPU v6e · 230 PFLOPS
Pod / clusterShipping
Microsoft logoMaia 100 Rack
32x Microsoft Maia 100 · 96 PFLOPS
Full rackRamping
NVIDIA logoDGX B200
8x NVIDIA B200 · 80 PFLOPS
Server nodeShipping
AMD logoMI325X Platform
8x AMD MI325X · 48 PFLOPS
Server nodeShipping
Back to all systemsCompute hub