Question 1

What is the cheapest cloud GPU available right now?

Accepted Answer

The cheapest cloud GPU changes daily as providers adjust pricing. Typically, NVIDIA T4 instances start under $0.20/hr, while more powerful options like A100s can be found under $1.50/hr on spot markets. This page shows live pricing across 54+ providers, updated continuously.

Question 2

Should I use spot or on-demand GPU instances?

Accepted Answer

Spot instances are 50–80% cheaper but can be interrupted with as little as 30 seconds notice. Use spot for batch inference, data processing, and fault-tolerant training with checkpointing. Use on-demand for production inference endpoints, real-time applications, and training runs that cannot tolerate restarts.

Question 3

How do I choose between GPU providers?

Accepted Answer

Consider four factors: (1) Price — compare $/hr for the same GPU model across providers. (2) Reliability — enterprise clouds (AWS, GCP, Azure) offer SLAs; marketplaces (Vast.ai, RunPod) are cheaper but less predictable. (3) Availability — some GPUs have waitlists. (4) Egress costs — data transfer out can add 10–20% to your real cost on hyperscalers.

Question 4

What GPU do I need for LLM inference?

Accepted Answer

For 7B parameter models (Llama 3 8B, Mistral 7B), a single GPU with 16–24 GB VRAM is sufficient (T4, L4, RTX 4090). For 13B–30B models, you need 24–48 GB (A10G, L40S, A100 40GB). For 70B+ models, you need 80 GB+ or multi-GPU setups (A100 80GB, H100, 2×L40S).

Question 5

Is renting a cloud GPU cheaper than buying hardware?

Accepted Answer

It depends on utilization. If you use a GPU more than 60% of the time (16+ hours/day), buying hardware pays off within 6–12 months. Below 30% utilization, cloud rentals are nearly always cheaper. Use our Buy vs Rent calculator for a personalized break-even analysis.

#	GPU	Provider	Price	VRAM	Count	Type	Region
	V100 V100 Older	Vast.ai	0.0133/hr	16GB	1×	Spot	US-West	Deploy
2	V100 V100 Older	Vast.ai	0.0133/hr	16GB	1×	Spot	US-West	Deploy
3	3080 RTX3080Ti Ampere	Vast.ai	0.0267/hr	12GB	1×	Spot	pt-portugal	Deploy
4	V100 V100 Older	Vast.ai	0.0282/hr	16GB	1×	On-Demand	US-West	Deploy
5	V100 V100 Older	Vast.ai	0.0300/hr	16GB	2×	Spot	US-West	Deploy
6	V100 V100 Older	Vast.ai	0.0305/hr	16GB	1×	On-Demand	US-West	Deploy
7	5060 RTX5060Ti Blackwell	Vast.ai	0.0360/hr	16GB	1×	Spot	Asia-Pacific	Deploy
8	V100 V100 Older	Verda	0.0483/hr	16GB	1×	Spot	EU-Central	Deploy
9	5060 RTX5060Ti Blackwell	Vast.ai	0.0497/hr	16GB	1×	On-Demand	Asia-Pacific	Deploy
10	V100 V100 Older	Vast.ai	0.0533/hr	16GB	4×	Spot	US-West	Deploy
11	3090 RTX3090 Ampere	Vast.ai	0.0533/hr	24GB	1×	Spot	US-East	Deploy
12	GPU A16 Older	Vultr	0.0590/hr	2GB	1×	On-Demand	ewr	Deploy
13	GPU A16 Older	Vultr	0.0590/hr	2GB	1×	On-Demand	ord	Deploy
14	GPU A16 Older	Vultr	0.0590/hr	2GB	1×	On-Demand	US-East	Deploy
15	GPU A16 Older	Vultr	0.0590/hr	2GB	1×	On-Demand	lhr	Deploy
16	GPU A16 Older	Vultr	0.0590/hr	2GB	1×	On-Demand	fra	Deploy
17	GPU A16 Older	Vultr	0.0590/hr	2GB	1×	On-Demand	sjc	Deploy
18	GPU A16 Older	Vultr	0.0590/hr	2GB	1×	On-Demand	nrt	Deploy
19	GPU A16 Older	Vultr	0.0590/hr	2GB	1×	On-Demand	sgp	Deploy
20	GPU A16 Older	Vultr	0.0590/hr	2GB	1×	On-Demand	blr	Deploy

Cheapest Cloud GPU

Cheapest by Performance Tier

High Performance

Mid-Range

Budget

How to Find the Cheapest Cloud GPU in 2026

5 strategies to cut GPU costs

Self-hosting vs API

Frequently Asked Questions

Get notified when prices drop