Skip to main content
PromptQuorumPromptQuorum

How Much Does a Cloud GPU Cost Per Hour in 2026?

Read in:
🇺🇸en🇩🇪de🇫🇷fr🇯🇵ja🇨🇳zh🇪🇸es🇧🇷pt🇸🇦ar🇰🇷ko

This page contains links to third-party products for reference. PromptQuorum is not enrolled in any affiliate program — these are plain links that earn no commission. Clicking links and your next steps are entirely your own responsibility. These links do not represent any endorsement or verification by PromptQuorum.

Quick Answer

RTX 4090: $0.15-0.44/hr. A100 80GB: $1.10-2.00/hr. H100: $2.89-4.00/hr. Cheapest for inference: Vast.ai spot.

  • ▸Vast.ai spot RTX 4090 is cheapest at ~$0.15/hr but can be preempted mid-job.
  • ▸RunPod secure cloud charges ~$0.44/hr for RTX 4090 with guaranteed availability.
  • ▸H100 80 GB on Lambda Labs: ~$3.29/hr — use only for 70B+ models or large batch jobs.

Updated: 2026-07

Hardware-SpecificIntermediate

Key Takeaways

  • ✓RTX 4090 24 GB is the cheapest viable option — $0.30-0.80/hr on marketplaces, ideal for 13B-30B inference
  • ✓A100 80 GB at $0.90-2.00/hr is the workhorse for 70B inference and most training jobs
  • ✓H100 80 GB at $2.50-4.00/hr is the fastest option but only worth it for large-scale training or production serving
  • ✓All ranges are July 2026 approximate — check live provider dashboards before booking

Best Pick: Match the Card to the Workload

The cheapest viable cloud GPU is the one that fits your model with the smallest VRAM headroom. Renting a $4/hr H100 to run a 13B model wastes 60+ GB of VRAM you are paying for.

For 7B-13B inference: an RTX 4090 24 GB on a marketplace (Vast.ai, RunPod community pool) at $0.30-0.80/hr. The 24 GB of VRAM is plenty, and consumer-card marketplaces undercut managed clouds.

For 70B inference or mid-scale fine-tuning: an A100 80 GB at $0.90-2.00/hr. The 80 GB of VRAM fits a 70B model at Q4 with context room. For frontier-model training or production serving with strict latency targets: an H100 80 GB at $2.50-4.00/hr — only worth it when sustained throughput is the constraint.

Check current RunPod GPU ratesproduct link · disclosedCheck current Vast.ai GPU ratesproduct link · disclosedCheck current Lambda Labs GPU ratesproduct link · disclosed

Cloud GPU Hourly Rates by Card (July 2026)

Ranges below are approximate July 2026 figures across major providers (RunPod, Vast.ai, Lambda Labs, and others). The low end is typically interruptible or marketplace pricing; the high end is on-demand managed cloud.

GPUVRAMHourly rate (approx)Best for
RTX 409024 GB$0.30-0.80/hr7B-30B inference, light fine-tuning
A100 80 GB80 GB$0.90-2.00/hr70B inference, most fine-tuning
H100 80 GB80 GB$2.50-4.00/hrLarge-scale training, latency-critical serving

Related Reading

Quick Answers About Cloud GPU Pricing

When is renting a cloud GPU cheaper than buying one?▾
Renting wins for short, bursty workloads — a few hours per week. Buying wins for sustained daily use. A used RTX 4090 at ~$2,500 pays for itself in roughly 3,000-8,000 cloud-rental hours at $0.30-0.80/hr.
Why does the same GPU cost so differently across providers?▾
Managed clouds (Lambda, AWS, GCP) include support, SLAs, and dedicated hardware — they cost more. Marketplaces (Vast.ai) source from individual hosts, which can be interruptible. Region and demand also shift prices.
Are quoted rates inclusive of storage and bandwidth?▾
Usually not. Persistent storage typically costs $0.05-0.20/GB-month. Outbound bandwidth can add cents per GB. For large model weights or datasets, factor these into the total.
How do I find the cheapest GPU for my workload right now?▾
Check at least two providers before booking — RunPod (managed) and Vast.ai (marketplace) cover both ends of the spectrum. Filter by required VRAM, then sort by price.

Want the full breakdown?

Read the complete guide →

Related Prompt Bites

← Back to Prompt Bites