← GPU price tracker

B200 GPU Rental Price

The NVIDIA B200 (Blackwell, 192 GB HBM3e) rents from $3.74/hr, averaged across 8 cloud providers, updated 2026-08-14. On-demand pricing has risen 15% over the past 70 days.

$6.84
on-demand /hr
$3.74
spot /hr
$5.89
reserved /hr
$4,993
/mo on-demand

Blackwell-generation flagship — roughly 2–3× H100 training throughput at FP8/FP4 and a far lower cost-per-token on large models. Blackwell · 192 GB HBM3e · 8 TB/s memory bandwidth · launched 2025.

on-demand /hr · spot /hr

Recent daily prices

DateOn-demandSpot
2026-08-14$6.84$3.74
2026-08-13$6.83$3.69
2026-08-12$6.90$3.76
2026-08-11$6.91$3.54
2026-08-10$6.86$3.59
2026-08-09$6.84$3.60
2026-08-08$7.06$3.60
2026-08-07$7.10$3.60
2026-08-05$7.11$3.39
2026-08-04$7.01$3.73
2026-08-03$7.00$3.67
2026-08-02$6.86$3.77
2026-08-01$6.87$3.71
2026-07-31$6.96$3.54

B200 vs other GPUs

GPUOn-demand /hrSpot /hrMemory
B200$6.84$3.74192 GB HBM3e
H200$4.37$2.98141 GB HBM3eB200 vs H200 →
H100$4.00$1.7180 GB HBM3B200 vs H100 →
MI355X$8.60$5.29288 GB HBM3eB200 vs MI355X →
MI325X$3.18$2.00256 GB HBM3eB200 vs MI325X →
MI300X$3.56$1.65192 GB HBM3B200 vs MI300X →

FAQ

How much does it cost to rent an B200 per hour?

The B200 averages $6.84/hr on-demand, $3.74/hr spot, and $5.89/hr reserved across 8 cloud providers (updated 2026-08-14).

What is the cheapest B200 cloud price?

The lowest tier is spot at about $3.74/hr on average; reserved (1–3 year commitment) runs ~$5.89/hr. Compare live per-provider rates on the Bargo GPU price tracker.

How much is an B200 per month?

At on-demand $6.84/hr, one B200 runs about $4,993/month (730 hours); reserved pricing lowers that to roughly $4,300/month.

Is the B200 cheaper than the H100?

On-demand, the B200 is $6.84/hr vs the H100 at $4.00/hr — pricier per hour. See the full B200 vs H100 comparison for specs and cost-per-job context.

What's the difference between B200 spot and on-demand pricing?

Spot ($3.74/hr) is interruptible spare capacity at a discount to on-demand ($6.84/hr); use spot for checkpointed or fault-tolerant jobs and on-demand or reserved for long, uninterrupted runs.

Cross-provider averages from public cloud GPU listings. Informational only.