Blackwell Architecture GPUs Cloud Pricing
Blackwell is NVIDIA's latest datacenter GPU architecture, succeeding Hopper. It introduces second-generation Transformer Engines with FP4 support, doubling inference throughput over H100. The B200, GB200, and GB300 models target large-scale AI training and inference with up to 192 GB HBM3e per GPU and NVLink 5.0 interconnects.
Blackwell Architecture GPUs Available in the Cloud
B100
ServerB200
GB200
ServerGB300
ServerHGX B300
ServerRTX 5060
RTX 5060 Ti
RTX 5070
RTX 5070 Ti
RTX 5080
RTX 5090
RTX 6000 Pro
RTX PRO 6000
ServerSample Blackwell Architecture GPUs Pricing
Showing 9 of 126 price points. Visit individual GPU pages above for full pricing.
Frequently Asked Questions
What is new in NVIDIA Blackwell vs Hopper?
Blackwell adds FP4 precision for faster inference, doubles the Transformer Engine throughput, increases memory bandwidth with HBM3e, and introduces NVLink 5.0 for higher inter-GPU bandwidth. The GB200 pairs two B200 GPUs with a Grace CPU in a single module.
Which cloud providers have Blackwell GPUs?
Blackwell availability is expanding across cloud providers. Check the pricing table above for current provider availability and pricing.