Skip to content
Launch imminent

Official date to be announced shortlyGPU offers available at launch; figures shown are indicative.

Priority access: sign up
EverXCloud

GPU Rental Price Report, November 2026

Public GPU-hour prices for H100, H200, B200, A100, L40S, RTX 4090 and RTX 5090 across 16 cloud providers, with medians and spreads.

Executive summary

EverXCloud read the public pricing pages of 16 GPU cloud providers on 2026-10-26 and recorded 75 prices per GPU-hour for eight GPUs. On-demand prices for the same GPU still vary widely between providers, and spot, preemptible or community capacity costs far less than on-demand capacity.

This report covers third-party list prices only. Once the EverXCloud exchange launches, its own price index will be published from real transactions.

Key findings

  1. The median on-demand H100 SXM price across 11 providers is $3.99 per GPU-hour, from $3.14 to $6.16: the most expensive listing costs 2.0× the cheapest.
  2. Spot, preemptible and community H100 SXM capacity has a median of $2.30 per GPU-hour across 3 providers, 42% below the on-demand median.
  3. The H200 carries a 24% premium over the H100 SXM at the median ($4.94 vs $3.99 on demand, 10 providers).
  4. On-demand B200 prices range from $6.00 to $8.60 per GPU-hour across 7 providers, with a median of $8.19.
  5. The A100 80GB shows the widest relative spread among data center GPUs: 2.6× between the cheapest ($1.09) and the most expensive ($2.79) on-demand listing, across 9 providers.
  6. AWS Capacity Blocks, a short-term reservation, list the H100 at $5.97 and the B200 at $14.21 per GPU-hour, 1.5× and 1.7× the on-demand medians of the other providers.

Prices by GPU

On-demand, USD per GPU-hour
GPUMinMedianMaxMax ÷ minProviders
H100 SXM$3.14$3.99$6.162.0×11
H100 PCIe$2.89$3.29$3.501.2×3
H200$3.62$4.94$6.301.7×10
B200$6.00$8.19$8.601.4×7
A100 80GB$1.09$1.49$2.792.6×9
L40S$0.97$1.56$2.252.3×6
RTX 4090$0.39$0.64$0.892.3×3
RTX 5090$0.65$0.92$1.191.8×2
Spot, preemptible or community, USD per GPU-hour
GPUMinMedianMaxMax ÷ minProviders
H100 SXM$1.99$2.30$2.691.4×3
H100 PCIe$1.99$1.99$1.991.0×1
H200$2.99$3.00$3.591.2×3
B200$4.09$4.26$5.981.5×3
A100 80GB$1.08$1.08$1.191.1×3
L40S$0.79$0.85$0.911.2×2
RTX 4090$0.16$0.34$0.654.1×3
RTX 5090$0.25$0.47$0.692.8×2

On-demand prices by provider

H100 SXM: on-demand price per GPU-hour, 11 providers
H100 SXM on-demand price per GPU-hour by providerMassed Compute: $3.14; Hyperstack: $3.20; Thunder Compute: $3.20; JarvisLabs: $3.49; Lambda: $3.99; Runpod: $3.99; Together AI: $3.99; Oblivus: $4.31; DigitalOcean: $4.41; Verda (DataCrunch): $4.61; CoreWeave: $6.16. Median $3.99 per GPU-hour.median $3.99Massed Compute$3.14Hyperstack$3.20Thunder Compute$3.20JarvisLabs$3.49Lambda$3.99Runpod$3.99Together AI$3.99Oblivus$4.31DigitalOcean$4.41Verda (DataCrunch)$4.61CoreWeave$6.16
H100 PCIe: on-demand price per GPU-hour, 3 providers
H100 PCIe on-demand price per GPU-hour by providerRunpod: $2.89; Lambda: $3.29; Oblivus: $3.50. Median $3.29 per GPU-hour.median $3.29Runpod$2.89Lambda$3.29Oblivus$3.50
H200: on-demand price per GPU-hour, 10 providers
H200 on-demand price per GPU-hour by providerMassed Compute: $3.62; Hyperstack: $3.99; Oblivus: $3.99; DigitalOcean: $4.47; JarvisLabs: $4.59; Runpod: $5.29; Nebius: $5.40; Together AI: $5.99; Verda (DataCrunch): $6.01; CoreWeave: $6.30. Median $4.94 per GPU-hour.median $4.94Massed Compute$3.62Hyperstack$3.99Oblivus$3.99DigitalOcean$4.47JarvisLabs$4.59Runpod$5.29Nebius$5.40Together AI$5.99Verda (DataCrunch)$6.01CoreWeave$6.30
B200: on-demand price per GPU-hour, 7 providers
B200 on-demand price per GPU-hour by providerHyperstack: $6.00; Lambda: $6.69; Runpod: $7.99; Together AI: $8.19; Nebius: $8.50; Verda (DataCrunch): $8.52; CoreWeave: $8.60. Median $8.19 per GPU-hour.median $8.19Hyperstack$6.00Lambda$6.69Runpod$7.99Together AI$8.19Nebius$8.50Verda (DataCrunch)$8.52CoreWeave$8.60
A100 80GB: on-demand price per GPU-hour, 9 providers
A100 80GB on-demand price per GPU-hour by providerThunder Compute: $1.09; Hyperstack: $1.35; Massed Compute: $1.35; Oblivus: $1.47; JarvisLabs: $1.49; Runpod: $1.79; Verda (DataCrunch): $2.16; CoreWeave: $2.70; Lambda: $2.79. Median $1.49 per GPU-hour.median $1.49Thunder Compute$1.09Hyperstack$1.35Massed Compute$1.35Oblivus$1.47JarvisLabs$1.49Runpod$1.79Verda (DataCrunch)$2.16CoreWeave$2.70Lambda$2.79
L40S: on-demand price per GPU-hour, 6 providers
L40S on-demand price per GPU-hour by providerMassed Compute: $0.97; Runpod: $1.09; Nebius: $1.55; DigitalOcean: $1.57; Verda (DataCrunch): $1.83; CoreWeave: $2.25. Median $1.56 per GPU-hour.median $1.56Massed Compute$0.97Runpod$1.09Nebius$1.55DigitalOcean$1.57Verda (DataCrunch)$1.83CoreWeave$2.25
RTX 4090: on-demand price per GPU-hour, 3 providers
RTX 4090 on-demand price per GPU-hour by providerCloudRift: $0.39; Oblivus: $0.64; Runpod: $0.89. Median $0.64 per GPU-hour.median $0.64CloudRift$0.39Oblivus$0.64Runpod$0.89
RTX 5090: on-demand price per GPU-hour, 2 providers
RTX 5090 on-demand price per GPU-hour by providerCloudRift: $0.65; Runpod: $1.19. Median $0.92 per GPU-hour.median $0.92CloudRift$0.65Runpod$1.19

Methodology

  • Prices were read on 2026-10-26 on each provider's official public pricing page, in USD per GPU-hour, excluding tax. No price was estimated, negotiated or taken from a third-party aggregator.
  • One price per provider, GPU and offer type. When a provider lists several configurations for the same GPU, the lowest listed price was kept, and the listing is recorded in the data.
  • Multi-GPU instances are converted to a per-GPU price by dividing the instance price by the number of GPUs (for example CoreWeave and AWS 8-GPU instances).
  • On-demand and spot, preemptible or community prices are reported separately. AWS prices are Capacity Block rates, a short-term reservation, and are not included in the on-demand statistics.
  • Providers whose pricing page could not be read, showed prices only after sign-in or on request, or did not map prices to GPUs unambiguously were left out: Vast.ai, Crusoe, Vultr, TensorDock, Fluidstack, Genesis Cloud, Google Cloud, Microsoft Azure, Voltage Park, Cudo Compute, Scaleway, Paperspace and SF Compute.
  • The median of an even number of prices is the mean of the two middle values.

Data and license

Download the data (CSV, 75 prices). The data and charts are published under the Creative Commons Attribution 4.0 license. Suggested citation: Source: EverXCloud GPU Rental Price Report, November 2026.

Related: H100 price per hour, GPU cost calculator, press room.

Sources

Be ready on launch day

Create your account now: you can top up your balance and launch a GPU as soon as the market opens.

Get priority accessExplore the live market