GB200 NVL72 (per GPU) specs

NVIDIA · Shipping now · first available Feb 2025. Dense compute, memory, bandwidth and power from the vendor's documents; each precision stays separate.

All accelerators · Compare with B200 (HGX)

Specifications

GB200 NVL72 (per GPU) specifications
Dense BF16 · FP32 accumulate2,500 TFLOPS
Dense FP8 · FP32 accumulate5,000 TFLOPS
Dense FP4 · FP32 accumulate10,000 TFLOPS
Memory186 GB · HBM
Memory bandwidth8 TB/s
Maximum chip power1,200 W
InterconnectNVLink 5 · 1.8 TB/s
Process nodeTSMC 4NP
First availabilityFeb 2025
AvailabilityShipping

One of the 72 Blackwell GPUs in a GB200 NVL72 rack (two per Grace Blackwell superchip); rentals are priced per GPU. Launch month is an estimate: the first generally available cloud instances were announced in February 2025.

Dense tensor rates with FP32 accumulation, never sparse or FP16-accumulate headline figures. Memory is in decimal GB. The est marker flags a value its vendor-document review did not confirm.

Rent it

$16.00/GPU-hrCheapest on-demand · Oracle OCI · observed

1 provider: Oracle OCI. Quotes checked through .

GB200 NVL72 rental prices from every provider →

Value ratios

GB200 NVL72 (per GPU) derived ratios
$ per dense BF16 PFLOP-hour$6.40
Dense BF16 TFLOPS per watt2.08
Bandwidth bytes per 1,000 BF16 FLOP3.20
$ per dense FP8 PFLOP-hour$3.20
Dense FP8 TFLOPS per watt4.17
Bandwidth bytes per 1,000 FP8 FLOP1.60
$ per GB of memory per hour$0.086

Prices divide the cheapest on-demand list price; theoretical ratios do not measure application speed.

Sources

Specifications reviewed . Datasheet's individual-GPU column for GB200 NVL72: its sparse FP4/FP8/BF16 figures (20/10/5 PF) halved, as its footnote directs, to 10/5/2.5 PF dense; the rack page's totals divide to the same. 186 GB HBM3E at 8 TB/s and a TDP configurable up to 1,200 W per GPU.

First availability: NVIDIA blog: CoreWeave GB200 NVL72 instances ↗ (2025-02-04): “first cloud service provider to make the NVIDIA Blackwell platform generally available”

Rental price: Oracle OCI published pricing ↗ (BM.GPU.GB200.4; 4-GPU GB200 NVL72 instance, 189GB per GPU).

Every accelerator as CSV · JSON