← GPU price tracker

MI325X GPU Rental Price

The AMD MI325X (CDNA 3, 256 GB HBM3e) rents from $3.06/hr, averaged across 5 cloud providers, updated 2026-09-28. On-demand pricing has fallen 6% over the past 90 days.

$3.06
on-demand /hr
$0.00
spot /hr
$1.95
reserved /hr
$2,234
/mo on-demand

Memory-maxed CDNA 3 part (256 GB) aimed at large-model inference where capacity, not raw FLOPs, is the bottleneck. CDNA 3 · 256 GB HBM3e · 6 TB/s memory bandwidth · launched 2025.

▮ on-demand /hr · ▮ spot /hr

Recent daily prices

DateOn-demandSpot
2026-09-28$3.06$—
2026-09-27$3.06$—
2026-09-25$3.06$—
2026-09-24$3.06$—
2026-09-23$3.07$—
2026-09-22$3.06$—
2026-09-21$3.06$—
2026-09-20$3.06$—
2026-09-19$3.06$—
2026-09-18$3.06$—
2026-09-17$3.06$—
2026-09-16$3.06$—
2026-09-15$3.06$—
2026-09-14$3.07$—

MI325X vs other GPUs

GPUOn-demand /hrSpot /hrMemory
MI325X$3.06$0.00256 GB HBM3e
B200$6.49$4.78192 GB HBM3eMI325X vs B200 →
H200$4.48$2.62141 GB HBM3eMI325X vs H200 →
H100$3.39$2.0080 GB HBM3MI325X vs H100 →
MI355X$5.99$2.97288 GB HBM3eMI325X vs MI355X →
MI300X$4.41$1.60192 GB HBM3MI325X vs MI300X →

FAQ

How much does it cost to rent an MI325X per hour?

The MI325X averages $3.06/hr on-demand, and $1.95/hr reserved across 5 cloud providers (updated 2026-09-28).

What is the cheapest MI325X cloud price?

On-demand runs about $3.06/hr; reserved (1–3 year commitment) runs ~$1.95/hr. Compare live per-provider rates on the Bargo GPU price tracker.

How much is an MI325X per month?

At on-demand $3.06/hr, one MI325X runs about $2,234/month (730 hours); reserved pricing lowers that to roughly $1,424/month.

Is the MI325X cheaper than the H100?

On-demand, the MI325X is $3.06/hr vs the H100 at $3.39/hr — cheaper per hour. See the full MI325X vs H100 comparison for specs and cost-per-job context.

What's the difference between MI325X spot and on-demand pricing?

Spot ($0.00/hr) is interruptible spare capacity at a discount to on-demand ($3.06/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.