Skip to content
Cloud Provider Benchmarks

Pricing

GPU cloud cost comparison: H100, H200 and L40S prices per GPU-hour

Published H100, H200 and L40S cloud prices per GPU-hour with sources, normalised per GB of memory and per TB/s of bandwidth, and what they leave out.

Published: Updated: Türkçe oku

Direct answer

Among the published USD list prices at the access date, the L40S is the lowest-cost GPU per hour and the H200 the highest. Per TB/s of memory bandwidth, however, the published H100 and H200 prices are lower than the L40S. Compare across currencies only after applying current exchange rates and tax treatment.

Executive summary

This analysis lists the published on-demand list prices per GPU-hour for H100, H200 and L40S at the tracked providers, each with its source and access date, and normalises them per GB of GPU memory and per TB/s of memory bandwidth. Prices stay in each provider's own currency and billing basis. Several providers offer a GPU without publishing an hourly price; they appear as available but unpriced.

Key findings

  1. 01 Prices are published in different currencies (USD and TRY), so no cross-provider ranking is computed on this page.
  2. 02 Billing granularity differs: per second with a minimum charge, per minute, or hourly. Short jobs can cost more than the hourly price implies on coarse billing.
  3. 03 Some GPUs are offered without public prices (for example PortvMind H200/L40S and several hyperscaler configurations); they are marked as unpriced rather than estimated.
  4. 04 Per-GPU-hour list prices exclude storage, data transfer and host resources that some providers bundle and others bill separately.

Published prices

Every row below comes from the provider’s own pricing page, with the access date linked to the source. The basis column explains what the price covers.

Published per-GPU-hour list prices
Provider GPU Per GPU-hour Per GB-hour Basis
PortvMind NVIDIA H100 TRY 195.835 TRY 2.4479 gpu.h100s.xlarge (1× H100) hourly list price, tr-ist-01, excluding tax. Multi-GPU profiles price from ₺141.12 per GPU-hour (3× and 4× profiles). · 6 November 2026
DigitalOcean NVIDIA H100 $4.410 $0.0551 On-demand, 1× H100 GPU Droplet. · 6 November 2026
DigitalOcean NVIDIA H200 $4.470 $0.0317 On-demand, 1× H200 GPU Droplet. · 6 November 2026
DigitalOcean NVIDIA L40S $1.570 $0.0327 On-demand, 1× L40S GPU Droplet. · 6 November 2026
Vultr NVIDIA H100 $2.300 $0.0287 On-demand "starting at" per-GPU rate, 8-GPU HGX configuration. · 6 November 2026
Vultr NVIDIA L40S $1.671 $0.0348 On-demand per-GPU rate, 1× L40S. · 6 November 2026

Why there is no single ranking

  • Currency. PortvMind lists prices in Turkish lira excluding tax; the other providers list in US dollars. Converting at a fixed rate would make the page wrong within weeks.
  • Configuration. Some providers sell one GPU per instance; others price an 8-GPU node and the per-GPU figure assumes the full node.
  • Billing granularity. DigitalOcean bills per second with a 60-second minimum, PortvMind bills per minute, and Vultr bills hourly. Hetzner bills per started hour with a monthly cap but does not offer these GPU models.
  • Bundled resources. vCPU, host RAM and local NVMe are included at some providers and priced separately at others.

AWS, Google Cloud and Azure price these GPUs as multi-GPU instance sizes that also bundle vCPU, memory and networking, and prices vary by region. They are not reduced to a per-GPU figure here, because splitting a bundled instance price across GPUs would be an estimate.

Normalising by memory and bandwidth

Price per GB-hour of GPU memory is a useful view for capacity-bound workloads (large models, long contexts). Price per TB/s of bandwidth is a first-order proxy for decode cost. Using the published USD prices above, an L40S costs about $1.82–1.93 per TB/s-hour, while the H100 and H200 list prices work out to roughly $0.69–1.32 and $0.93 respectively. The GPU price-performance page charts both, grouped by currency.

Availability without a public price

GPU availability and published per-GPU-hour list prices by provider
Provider NVIDIA H100NVIDIA H200NVIDIA L40S Türkiye infrastructure
PortvMind Cloud Generally available gpu.h100s / gpu.h100m profiles TRY 195.835 / GPU-h 1, 2, 3 or 4 GPUs per server; 80 GB (81,920 MB) VRAM per GPU; 26 vCPU and 496 GB RAM per GPU; local NVMe disk included. Generally available GPU / AI Compute (H200 option) Listed as an available GPU option; configurations and prices are not included in the public calculator table and are provided via the console or sales. Generally available GPU / AI Compute (L40S option) Listed as an available GPU option; configurations and prices are not included in the public calculator table and are provided via the console or sales. Yes
Amazon Web Services Generally available EC2 P5 Single-GPU p5.4xlarge and 8-GPU p5.48xlarge sizes; availability varies by region. Generally available EC2 P5e / P5en 8-GPU instances; availability varies by region. Generally available EC2 G6e 1 to 8 L40S GPUs per instance. Yes
Microsoft Azure Generally available NC H100 v5 / ND H100 v5 NC H100 v5 with one or two H100 NVL GPUs; ND H100 v5 with eight H100 GPUs. Generally available ND H200 v5 Eight H200 GPUs per VM. Not offered L40S SKUs are documented for Azure Local / Azure Stack Hub (customer-site hardware). No
Google Cloud Generally available A3 High / A3 Mega Up to 8 H100 GPUs per VM; zone availability varies. Generally available A3 Ultra 8 H200 GPUs per VM; zone availability varies. Not offered Announced
DigitalOcean Generally available GPU Droplet — NVIDIA HGX H100 $4.410 / GPU-h 1× or 8× H100 80 GB; 1× plan with 20 vCPU and 240 GiB RAM. Generally available GPU Droplet — NVIDIA HGX H200 $4.470 / GPU-h 1× or 8× H200 141 GB; 1× plan with 24 vCPU and 240 GiB RAM. Generally available GPU Droplet — NVIDIA L40S $1.570 / GPU-h 1× L40S 48 GB with 8 vCPU and 64 GiB RAM. No
Vultr Generally available Cloud GPU — NVIDIA HGX H100 $2.300 / GPU-h 8× H100 80 GB SXM (640 GB GPU RAM, 216 vCPU, 1,914 GB RAM). Unverified Not listed on the public pricing page H200 was not listed on the Vultr pricing page at the last review date. Generally available Cloud GPU — NVIDIA L40S $1.671 / GPU-h 1, 2, 4 or 8× L40S 48 GB; 1× plan with 16 vCPU and 180 GB RAM. No
Hetzner Not offered Not offered Not offered No

A GPU marked as available but absent from the price table is offered through a console, quote or reservation rather than a public list price.

Methodology

See the pricing methodology. Prices are recorded with source URL and access date, normalised to one GPU-hour where the provider’s pricing allows it, and never estimated.

Limitations

  • List prices only; committed-use, spot and negotiated discounts are excluded.
  • Regional price differences are not tracked; the region is noted where the provider specifies one.
  • Taxes are excluded where the provider lists prices excluding tax, which is not uniform.

Conclusion

Use this table to shortlist providers that offer the GPU you need at a published price, then compare in one currency at today’s rate and, ideally, on measured cost per million tokens for your model.

Frequently asked questions

Which cloud has the cheapest H100?

It depends on currency, configuration and commitment. Among the USD list prices tracked here at the access date, Vultr's 8-GPU H100 rate is the lowest per GPU-hour. PortvMind publishes TRY prices, which must be converted at a current rate before comparing.

Why are some providers missing from the price table?

They either do not offer the GPU or do not publish an hourly price for it in a form that can be normalised to one GPU-hour. Estimated or third-party prices are not used.

Is price per GPU-hour the right metric?

It is the starting point. Cost per million tokens at a target latency is the more meaningful metric, and it can only come from measured benchmarks.

Sources

  1. [1] PortvMind Cloud pricing — PortvMind Supplied by provider team accessed:
  2. [2] DigitalOcean GPU Droplet pricing — DigitalOcean Official documentation accessed:
  3. [3] How are GPU products billed differently? (Vultr) — Vultr Official documentation accessed:
  4. [4] Vultr pricing — Vultr Official documentation accessed:
  5. [5] Amazon EC2 On-Demand pricing — AWS Official documentation accessed:
  6. [6] Google Compute Engine VM instance pricing — Google Cloud Official documentation accessed:
  7. [7] Azure Linux Virtual Machines pricing — Microsoft Official documentation accessed: