Skip to main content
Prices on GPU Outlet are set by sellers, not by us — so what a GPU costs shifts with what’s listed. You always see the live rate for a listing before you launch. This page explains how billing works; the numbers below are a snapshot of current headline rates.
The always-current picture — every listing, its price, and availability — is on the marketplace (login required).

​
Current headline rates

The from price is the cheapest live rate for each GPU family right now. It tracks the marketplace automatically. A machine may carry more than one GPU; renting more than one gives you the whole machine with its GPUs pooled. See the marketplace for what each listing offers.

​
What’s included

  • The GPU itself (all listed VRAM)
  • A reasonable CPU allocation (8+ vCPU per GPU typical)
  • 64+ GB host RAM per GPU
  • 100 GB scratch SSD per pod
  • Outbound bandwidth (no per-byte egress charges)

​
What’s NOT included

  • Persistent storage — pod disk is wiped on stop. Push to S3/HF/etc.
  • Public IPv4 — every pod gets one. Want a static IP? Not in the catalog.
  • Reserved instances / committed contracts — talk to business@gpuoutlet.ai.

​
Per-second metering

The hourly rate is divided by 3,600 and debited per second. A 17-minute session on H100 charges:
Not 1.95(thefullhour),not1.95 (the full hour), not 2.00 (rounded up). Exactly $0.55 (the displayed cents).

​
Volume tiers

Spending more than $1,000/month? Email business@gpuoutlet.ai — we offer:
  • Discount tiers for committed monthly volume
  • Net-30 invoicing (vs prepaid)
  • Higher rate limits
  • Dedicated capacity reservations