Guides
Understand GPU rental quotes before choosing a cloud provider. These guides combine source-linked explanations with the current database snapshot.
Current means eligible observations in this build’s dated database snapshot, not a new availability check or a market-wide minimum. Whole-GPU USD quotes only; standard and unspecified regions, tax excluded or unknown. Spot stays separate. Stale, fractional, advertised starting prices and GPU-only components do not enter the calculations. Source, tax, marketplace, FX and billing qualifiers remain on the offer cards.
Price observations dated: 2026-11-18
- How to choose a GPU cloud for LLM inference
Start with model memory, then compare the billable allocation, region and price type. A low GPU-hour rate alone does not establish whether a model fits or what the full instance costs.
- Spot vs on-demand GPU prices
Compare observed Spot discounts within matching configurations. Cheaper interruptible capacity comes with service continuity risks; a discount is not a reliability or availability guarantee.
- Per-GPU-hour vs instance price
Multi-GPU instances can have an attractive normalized rate and a much larger mandatory bill. Use the observed allocation and billing minimum to compare like for like.
- GPU cloud price glossary
Read the labels attached to each quote. These definitions describe the comparison policy and provider billing concepts; they do not replace a provider contract.
- Cheapest GPU per VRAM GB right now
Compare the lowest observed cost per GB of GPU memory per hour for each exact model and price type, with popular models first. This measures memory cost, not inference speed or overall value.
Know the full job cost before you rent: GPU Budget Kit ($12)