Public GPU prices, with sources.
On-demand and reserved prices for H100, H200, B200, A100 and L40S from 19 providers with public prices (25 checked). Every row has a link and a date. Unknown stays unknown.
Download CSVDownload JSONFake provider checklist (PDF)
Last data check: 2026-11-02. Rows: 211. Prices are US dollars per GPU-hour, divided from instance prices where the page lists a whole node.
Lowest public on-demand price by GPU
Three lowest on-demand prices per GPU among rows we rate medium or high confidence. Spot, preemptible, serverless and reserved prices are in the full table. The cheapest row is not the best buy. Compare the GPU variant, the region, the term and whether capacity exists.
H100
| Provider | Variant | $ per GPU-hr | Instance | Checked | Source |
|---|---|---|---|---|---|
| Latitude.sh | 80GB | $1.68 | 1x | 2026-11-01 | source |
| Hyperstack | - | $2.50 | 1x | 2026-11-02 | source |
| Massed Compute | 80GB | $2.73 | 1x | 2026-11-02 | source |
H200
| Provider | Variant | $ per GPU-hr | Instance | Checked | Source |
|---|---|---|---|---|---|
| Civo | 141GB | $3.49 | 1x | 2026-11-02 | source |
| Massed Compute | NVL 141GB | $3.62 | 1x | 2026-11-02 | source |
| Hyperstack | SXM | $3.99 | 1x | 2026-11-02 | source |
B200
| Provider | Variant | $ per GPU-hr | Instance | Checked | Source |
|---|---|---|---|---|---|
| Massed Compute | SXM6 | $5.43 | 8x | 2026-11-02 | source |
| Hyperstack | - | $6.00 | 1x | 2026-11-02 | source |
| Lambda | SXM6 180GB | $6.69 | 8x | 2026-11-02 | source |
A100
| Provider | Variant | $ per GPU-hr | Instance | Checked | Source |
|---|---|---|---|---|---|
| Civo | 40GB | $1.09 | 1x | 2026-11-02 | source |
| Thunder Compute | 80GB | $1.09 | 1x | 2026-11-02 | source |
| Hyperstack | - | $1.35 | 1x | 2026-11-02 | source |
L40S
| Provider | Variant | $ per GPU-hr | Instance | Checked | Source |
|---|---|---|---|---|---|
| Massed Compute | 48GB | $0.97 | 1x | 2026-11-02 | source |
| RunPod | - | $1.09 | 1x | 2026-11-02 | source |
| Civo | 48GB | $1.29 | 1x | 2026-11-02 | source |
How to read this. A "from" price can be for a different GPU variant (PCIe vs SXM), a smaller instance, or a region that is out of stock. Marketplaces and brokers set prices live or by quote, so they appear as no public price. We never fill a gap with a guess.
Lead times and minimum terms (public statements)
What providers say on their own pages about how fast you can start and what you must commit to. Most reserved capacity has no published lead time. Ask in writing.
| Provider | Product | GPUs | What the page says | Minimum term or payment | Source |
|---|---|---|---|---|---|
| Lambda | On-demand instances | B200, H100, A100, GH200 | Page: deploy instances in minutes with self-serve, first-come access. | None | source 2026-11-01 |
| Lambda | 1-Click Clusters (reserved) | B200, H100 | Not published. Page says talk to our team. | 2 weeks to 1 year, 16 or more GPUs (page) | source 2026-11-01 |
| Thunder Compute | Instances | A100, H100 | Page: persistent GPU machines ready to SSH into in minutes. | None | source 2026-11-01 |
| Voltage Park | On-demand | H100 | Page: self-serve, up within 15 mins, no minimum terms. | None | source 2026-11-01 |
| Voltage Park | Long-term reserve | H100 | Not published. | 6 or more month contracts, 32 to 8000+ GPUs (page) | source 2026-11-01 |
| Hyperbolic | On-demand | H100, H200, B200 | FAQ: launch an on-demand GPU instance in minutes. | None | source 2026-11-01 |
| Hyperbolic | Reserved | H100, H200, B200 | Not published. | Shorter-term commitment, usually under one year (FAQ) | source 2026-11-01 |
| Compute Exchange | Reserved rental (aggregator quotes) | H100 PCIe, A100 80GB, L40S | Page: most reservations of available models provision within 1 to 2 weeks of contract signing. | Terms 1 to 36 months | source 2026-11-01 |
| Compute Exchange | Reserved rental (aggregator quotes) | H200, B200, B300, GB200 and GB300 NVL72 | Page: may take 4 to 8 weeks or provision on phased delivery schedules. | B300: 6-month minimum, allocation-constrained. B200: 3-month minimum commit (page table) | source 2026-11-01 |
| Compute Exchange | Quote | all | Page: returns a live quote within 24 hours. | None | source 2026-11-01 |
| Catalyst Capacity | Cobalt dedicated rapid GPU deployments | NVIDIA and AMD | Page: 4 to 16 weeks. | 30% deposit, refundable if you are not satisfied within a 3-day validation window (page) | source 2026-11-01 |
| FlexiCloud | Dedicated 8-GPU nodes (broker) | H100, H200, B200, B300 | Not published. Page says lead time per region is put in writing within about one working day of asking. B300 is the tightest. | One-year minimum, one full 8-GPU node, billed upfront for the term | source 2026-11-01 |
| 1Legion | GPU server rental (reseller program) | Blackwell B300 and others | Not published. | 36, 48 or 60 month terms, down payment at signing, bank guarantee or letter of credit for larger allocations, no trials (page) | source 2026-11-01 |
| Hyperstack | B300 | B300 | Page banner: B300 on-demand in August, reserved private clusters in Q4. | Contact sales for reserved | source 2026-11-01 |
| DigitalOcean | Reserved GPU Droplets | H100, H200, B300 | Not published. Contact sales. | 12-month contractual commitment (page) | source 2026-11-01 |
| Verda | Reserved | GPUs listed on the site | Not published. Page says custom quote. | Discounted capacity starting from 1 month (page) | source 2026-11-01 |
| Together AI | Reserved GPU clusters | H100, H200, B200 | Not published. | Terms priced in bands: 3 to 30, 31 to 90, 91 to 180, 181+ days (page) | source 2026-11-01 |
Need a number you cannot find here?
Tell us the GPU, count, term, region and budget. We reply with the public prices that match and the checks to run before you pay. It is free. We do not sell compute.
Send a GPU requestRun the fake provider checklist first
Full table
Filter by GPU, provider or type. For sorting, download the CSV.
| Provider | GPU | Variant | Type | Term | $ per GPU-hr | As listed | Conf. | Checked | Source |
|---|---|---|---|---|---|---|---|---|---|
| Latitude.sh | H100 | 80GB | on_demand | none | 1.68 | Starting at $1.68/hr | medium | 2026-11-01 | source |
| Voltage Park | H100 | HGX | on_demand | none | 1.99 | starting at $1.99 per FAQ | low | 2026-11-02 | source |
| Hyperstack | H100 | on_demand | none | 2.50 | $2.50/hr | high | 2026-11-02 | source | |
| Hyperstack | H100 | NVL | on_demand | none | 2.60 | $2.60/hr | high | 2026-11-02 | source |
| Massed Compute | H100 | 80GB | on_demand | none | 2.73 | $2.73/hr for 1x | medium | 2026-11-02 | source |
| RunPod | H100 | PCIe | on_demand | none | 2.89 | $2.89/hr | medium | 2026-11-02 | source |
| Civo | H100 | 80GB | on_demand | none | 2.99 | $2.99 per hour for 1x | high | 2026-11-02 | source |
| Massed Compute | H100 | SXM5 80GB | on_demand | none | 3.14 | $3.14/hr for 1x | medium | 2026-11-02 | source |
| RunPod | H100 | NVL | on_demand | none | 3.19 | $3.19/hr | medium | 2026-11-02 | source |
| Hyperstack | H100 | SXM | on_demand | none | 3.20 | $3.20/hr | high | 2026-11-02 | source |
| Thunder Compute | H100 | SXM | on_demand | none | 3.20 | $3.20/hr | high | 2026-11-02 | source |
| Lambda | H100 | PCIe 80GB | on_demand | none | 3.29 | $3.29 per GPU-hr (1x instance) | high | 2026-11-02 | source |
| Crusoe | H100 | 80GB HGX | on_demand | none | 3.90 | $3.90/GPU-hr | medium | 2026-11-02 | source |
| Lambda | H100 | SXM 80GB | on_demand | none | 3.99 | $3.99 per GPU-hr (8x instance) | high | 2026-11-02 | source |
| RunPod | H100 | SXM | on_demand | none | 3.99 | $3.99/hr | medium | 2026-11-02 | source |
| Together AI | H100 | HGX | on_demand | none | 3.99 | $3.99 | high | 2026-11-02 | source |
| Lambda | H100 | SXM 80GB | on_demand | none | 4.09 | $4.09 per GPU-hr (4x instance) | high | 2026-11-02 | source |
| Lambda | H100 | SXM 80GB | on_demand | none | 4.19 | $4.19 per GPU-hr (2x instance) | high | 2026-11-02 | source |
| Lambda | H100 | SXM 80GB | on_demand | none | 4.29 | $4.29 per GPU-hr (1x instance) | high | 2026-11-02 | source |
| DigitalOcean | H100 | HGX | on_demand | none | 4.41 | $4.41/GPU/hour | high | 2026-11-02 | source |
| Nebius | H100 | HGX | on_demand | none | 4.50 | $4.50 per GPU-hr | high | 2026-11-02 | source |
| Verda | H100 | SXM5 80GB | on_demand | none | 4.841 | $4.841/h | high | 2026-11-02 | source |
| CoreWeave | H100 | HGX | on_demand | none | 6.155 | $49.24/hr per 8-GPU instance | high | 2026-11-02 | source |
| Amazon Web Services | H100 | SXM 80GB, 8x | on_demand | none | 6.88 | $55.04/hr for p5.48xlarge (Linux, shared tenancy) | high | 2026-11-01 | source |
| Amazon Web Services | H100 | 80GB, 1x | on_demand | none | 6.88 | $6.88/hr for p5.4xlarge (Linux, shared tenancy) | high | 2026-11-01 | source |
| Microsoft Azure | H100 | NVL 94GB, 1x VM | on_demand | none | 6.98 | $6.98/hr for Standard_NC40ads_H100_v5 (Linux) | high | 2026-11-02 | source |
| Oracle Cloud (OCI) | H100 | on_demand | none | 10.00 | $10 GPU Per Hour (B98415) | high | 2026-11-02 | source | |
| Microsoft Azure | H100 | SXM 80GB, 8x VM | on_demand | none | 12.29 | $98.32/hr for Standard_ND96isr_H100_v5 (Linux) | high | 2026-11-02 | source |
| Modal | H100 | SXM5 | serverless_per_second | none | 3.9492 | $0.001097/sec | high | 2026-11-02 | source |
| Hyperstack | H100 | reserved | reservation, term not stated on page | 1.75 | $1.75/hr (starting from) | medium | 2026-11-02 | source | |
| Hyperstack | H100 | NVL | reserved | reservation, term not stated on page | 1.82 | $1.82/hr (starting from) | medium | 2026-11-02 | source |
| Civo | H100 | 80GB | reserved | 36 months | 2.49 | $2.49 per hour for 1x, 36 months commitment | high | 2026-11-02 | source |
| Civo | H100 | 80GB | reserved | 24 months | 2.59 | $2.59 per hour for 1x, 24 months commitment | high | 2026-11-02 | source |
| Civo | H100 | 80GB | reserved | 12 months | 2.69 | $2.69 per hour for 1x, 12 months commitment | high | 2026-11-02 | source |
| Hyperstack | H100 | SXM | reserved | reservation, term not stated on page | 2.72 | $2.72/hr (starting from) | medium | 2026-11-02 | source |
| Civo | H100 | 80GB | reserved | 6 months | 2.79 | $2.79 per hour for 1x, 6 months commitment | high | 2026-11-02 | source |
| FlexiCloud (broker) | H100 | HGX SXM | reserved | 12 months or longer, floor rate on the longest term | 2.95 | from $2.95 per GPU-hour | medium | 2026-11-02 | source |
| Together AI | H100 | HGX | reserved | 91 to 180 days | 3.19 | $3.19 | high | 2026-11-02 | source |
| DigitalOcean | H100 | HGX | reserved | 12 months | 3.26 | $3.26/GPU/hour | high | 2026-11-02 | source |
| Together AI | H100 | HGX | reserved | 31 to 90 days | 3.45 | $3.45 | high | 2026-11-02 | source |
| Together AI | H100 | HGX | reserved | 3 to 30 days | 3.69 | $3.69 | high | 2026-11-02 | source |
| Lambda | H100 | reserved | 2 weeks to 1 year | 5.54 | $5.54 per GPU-hr at 256 GPUs, 2 weeks to 1 year | high | 2026-11-02 | source | |
| Lambda | H100 | reserved | 2 weeks to 1 year | 5.85 | $5.85 per GPU-hr at 64 GPUs, 2 weeks to 1 year | high | 2026-11-02 | source | |
| Lambda | H100 | reserved | 2 weeks to 1 year | 6.16 | $6.16 per GPU-hr at 16 GPUs, 2 weeks to 1 year | high | 2026-11-02 | source | |
| Nebius | H100 | HGX | preemptible | none | 0.79 | from $0.79 per GPU-hr (preemptible) | medium | 2026-11-02 | source |
| Microsoft Azure | H100 | NVL 94GB, 1x VM | spot | none | 1.4183 | $1.418336/hr for Standard_NC40ads_H100_v5 (Linux) | high | 2026-11-02 | source |
| Together AI | H100 | HGX | preemptible | none | 1.99 | $1.99 | high | 2026-11-02 | source |
| Hyperstack | H100 | PCIe | spot | none | 2.00 | $2.00/hr (spot) | medium | 2026-11-02 | source |
| Microsoft Azure | H100 | SXM 80GB, 8x VM | spot | none | 2.2712 | $18.169536/hr for Standard_ND96isr_H100_v5 (Linux) | high | 2026-11-02 | source |
| Verda | H100 | SXM5 80GB | spot | none | 2.421 | $2.421/h (spot) | high | 2026-11-02 | source |
| Fluidstack | H100 | contact_sales | No public pricing page found (/pricing returns 404) | medium | 2026-11-01 | source | |||
| Together AI | H100 | HGX | contact_sales | 181+ days | Contact us | high | 2026-11-02 | source | |
| Voltage Park | H100 | HGX | contact_sales | 6+ months | Contact for pricing | high | 2026-11-02 | source | |
| Google Cloud | H100 | not_public | Pricing page is rendered by script; static fetch returns no rates | high | 2026-11-01 | source | |||
| Hyperbolic | H100 | not_public | On-demand, reserved and private cloud offered. No rate in the home page text | medium | 2026-11-01 | source | |||
| Paperspace (DigitalOcean) | H100 | not_public | Page shows $3.09/hour on the H100 card and a footnote saying $5.95/hour on-demand promo for H100 | low | 2026-11-01 | source | |||
| SF Compute | H100 | not_public | Calculator on site, no static rate in page text | low | 2026-11-01 | source | |||
| Vast.ai | H100 | not_public | Market-set. No static rate on the pricing page | high | 2026-11-01 | source | |||
| Civo | H200 | 141GB | on_demand | none | 3.49 | $3.49 per hour for 1x | high | 2026-11-02 | source |
| Massed Compute | H200 | NVL 141GB | on_demand | none | 3.62 | $3.62/hr for 1x | medium | 2026-11-02 | source |
| Hyperstack | H200 | SXM | on_demand | none | 3.99 | $3.99/hr | high | 2026-11-02 | source |
| Crusoe | H200 | 141GB HGX | on_demand | none | 4.29 | $4.29/GPU-hr | medium | 2026-11-02 | source |
| DigitalOcean | H200 | HGX | on_demand | none | 4.47 | $4.47/GPU/hour | high | 2026-11-02 | source |
| RunPod | H200 | on_demand | none | 5.29 | $5.29/hr | medium | 2026-11-02 | source | |
| Nebius | H200 | HGX | on_demand | none | 5.40 | $5.40 per GPU-hr | high | 2026-11-02 | source |
| Together AI | H200 | HGX | on_demand | none | 5.99 | $5.99 | high | 2026-11-02 | source |
| CoreWeave | H200 | HGX | on_demand | none | 6.305 | $50.44/hr per 8-GPU instance | high | 2026-11-02 | source |
| Verda | H200 | SXM5 141GB | on_demand | none | 6.313 | $6.313/h | high | 2026-11-02 | source |
| Amazon Web Services | H200 | SXM 141GB, 8x | on_demand | none | 7.912 | $63.296/hr for p5en.48xlarge (Linux, shared tenancy) | high | 2026-11-01 | source |
| Oracle Cloud (OCI) | H200 | on_demand | none | 10.00 | $10 GPU Per Hour (B110519) | high | 2026-11-02 | source | |
| Modal | H200 | SXM | serverless_per_second | none | 4.5396 | $0.001261/sec | high | 2026-11-02 | source |
| Hyperstack | H200 | SXM | reserved | reservation, term not stated on page | 2.79 | $2.79/hr (starting from) | medium | 2026-11-02 | source |
| Civo | H200 | 141GB | reserved | 36 months | 2.99 | $2.99 per hour for 1x, 36 months commitment | high | 2026-11-02 | source |
| Civo | H200 | 141GB | reserved | 24 months | 3.09 | $3.09 per hour for 1x, 24 months commitment | high | 2026-11-02 | source |
| Civo | H200 | 141GB | reserved | 12 months | 3.19 | $3.19 per hour for 1x, 12 months commitment | high | 2026-11-02 | source |
| Civo | H200 | 141GB | reserved | 6 months | 3.29 | $3.29 per hour for 1x, 6 months commitment | high | 2026-11-02 | source |
| FlexiCloud (broker) | H200 | HGX SXM | reserved | 12 months or longer, floor rate on the longest term | 3.35 | from $3.35 per GPU-hour | medium | 2026-11-02 | source |
| DigitalOcean | H200 | HGX | reserved | 12 months | 3.40 | $3.40/GPU/hour | high | 2026-11-02 | source |
| Together AI | H200 | HGX | reserved | 91 to 180 days | 3.99 | $3.99 | high | 2026-11-02 | source |
| Together AI | H200 | HGX | reserved | 31 to 90 days | 4.15 | $4.15 | high | 2026-11-02 | source |
| Together AI | H200 | HGX | reserved | 3 to 30 days | 4.99 | $4.99 | high | 2026-11-02 | source |
| Nebius | H200 | HGX | preemptible | none | 0.79 | from $0.79 per GPU-hr (preemptible) | medium | 2026-11-02 | source |
| Together AI | H200 | HGX | preemptible | none | 2.99 | $2.99 | high | 2026-11-02 | source |
| Verda | H200 | SXM5 141GB | spot | none | 3.157 | $3.157/h (spot) | high | 2026-11-02 | source |
| Together AI | H200 | HGX | contact_sales | 181+ days | Contact us | high | 2026-11-02 | source | |
| Hyperbolic | H200 | not_public | No rate in the home page text | medium | 2026-11-01 | source | |||
| Massed Compute | B200 | SXM6 | on_demand | none | 5.4325 | $43.46/hr for 8x | medium | 2026-11-02 | source |
| Hyperstack | B200 | on_demand | none | 6.00 | $6.00/hr | high | 2026-11-02 | source | |
| Lambda | B200 | SXM6 180GB | on_demand | none | 6.69 | $6.69 per GPU-hr (8x instance) | high | 2026-11-02 | source |
| Lambda | B200 | SXM6 180GB | on_demand | none | 6.79 | $6.79 per GPU-hr (4x instance) | high | 2026-11-02 | source |
| Lambda | B200 | SXM6 180GB | on_demand | none | 6.89 | $6.89 per GPU-hr (2x instance) | high | 2026-11-02 | source |
| Lambda | B200 | SXM6 180GB | on_demand | none | 6.99 | $6.99 per GPU-hr (1x instance) | high | 2026-11-02 | source |
| RunPod | B200 | on_demand | none | 7.99 | $7.99/hr | medium | 2026-11-02 | source | |
| Together AI | B200 | HGX | on_demand | none | 8.19 | $8.19 | high | 2026-11-02 | source |
| Nebius | B200 | HGX | on_demand | none | 8.50 | $8.50 per GPU-hr | high | 2026-11-02 | source |
| CoreWeave | B200 | HGX | on_demand | none | 8.60 | $68.80/hr per 8-GPU instance | high | 2026-11-02 | source |
| Verda | B200 | SXM6 180GB | on_demand | none | 8.958 | $8.958/h | high | 2026-11-02 | source |
| Oracle Cloud (OCI) | B200 | on_demand | none | 14.00 | $14 GPU Per Hour (B110978) | high | 2026-11-02 | source | |
| Amazon Web Services | B200 | SXM 8x | on_demand | none | 14.2416 | $113.9328/hr for p6-b200.48xlarge (Linux, shared tenancy) | high | 2026-11-01 | source |
| Modal | B200 | serverless_per_second | none | 6.2496 | $0.001736/sec | high | 2026-11-02 | source | |
| FlexiCloud (broker) | B200 | HGX SXM | reserved | 12 months or longer, floor rate on the longest term | 3.75 | from $3.75 per GPU-hour | medium | 2026-11-02 | source |
| Civo | B200 | 180GB | reserved | 36 months | 3.79 | $30.32 per hour for 8x, 36 months commitment | high | 2026-11-02 | source |
| Civo | B200 | 180GB | reserved | 24 months | 3.99 | $31.92 per hour for 8x, 24 months commitment | high | 2026-11-02 | source |
| Civo | B200 | 180GB | reserved | 12 months | 4.49 | $35.92 per hour for 8x, 12 months commitment | high | 2026-11-02 | source |
| Hyperstack | B200 | reserved | reservation, term not stated on page | 5.10 | $5.10/hr (starting from) | medium | 2026-11-02 | source | |
| Together AI | B200 | HGX | reserved | 91 to 180 days | 6.79 | $6.79 | high | 2026-11-02 | source |
| Together AI | B200 | HGX | reserved | 31 to 90 days | 7.79 | $7.79 | high | 2026-11-02 | source |
| Together AI | B200 | HGX | reserved | 3 to 30 days | 7.99 | $7.99 | high | 2026-11-02 | source |
| Lambda | B200 | HGX | reserved | 2 weeks to 1 year | 8.87 | $8.87 per GPU-hr at 256+ GPUs, 2 weeks to 1 year | high | 2026-11-02 | source |
| Lambda | B200 | HGX | reserved | 2 weeks to 1 year | 9.36 | $9.36 per GPU-hr at 64 GPUs, 2 weeks to 1 year | high | 2026-11-02 | source |
| Lambda | B200 | HGX | reserved | 2 weeks to 1 year | 9.86 | $9.86 per GPU-hr at 16 GPUs, 2 weeks to 1 year | high | 2026-11-02 | source |
| Nebius | B200 | HGX | preemptible | none | 0.99 | from $0.99 per GPU-hr (preemptible) | medium | 2026-11-02 | source |
| Together AI | B200 | HGX | preemptible | none | 4.09 | $4.09 | high | 2026-11-02 | source |
| Verda | B200 | SXM6 180GB | spot | none | 4.479 | $4.479/h (spot) | high | 2026-11-02 | source |
| Crusoe | B200 | 180GB HGX | contact_sales | Contact sales | high | 2026-11-02 | source | ||
| Together AI | B200 | HGX | contact_sales | 181+ days | Contact us | high | 2026-11-02 | source | |
| Google Cloud | B200 | not_public | Pricing page is rendered by script; static fetch returns no rates | high | 2026-11-01 | source | |||
| Hyperbolic | B200 | not_public | No rate in the home page text | medium | 2026-11-01 | source | |||
| Vast.ai | B200 | not_public | Market-set. No static rate on the pricing page | high | 2026-11-01 | source | |||
| Hyperstack | B300 | on_demand | none | 7.40 | $7.40/hr | high | 2026-11-02 | source | |
| Civo | B300 | on_demand | none | 7.49 | $59.92 per hour for 8x | high | 2026-11-02 | source | |
| Crusoe | B300 | 288GB HGX | on_demand | none | 8.96 | $8.96/GPU-hr | medium | 2026-11-02 | source |
| RunPod | B300 | on_demand | none | 8.99 | $8.99/hr | medium | 2026-11-02 | source | |
| Nebius | B300 | HGX | on_demand | none | 9.50 | $9.50 per GPU-hr | high | 2026-11-02 | source |
| Verda | B300 | SXM6 268GB | on_demand | none | 11.46 | $11.46/h | high | 2026-11-02 | source |
| Oracle Cloud (OCI) | B300 | on_demand | none | 15.00 | $15 GPU Per Hour (B112237) | high | 2026-11-02 | source | |
| Amazon Web Services | B300 | SXM 8x | on_demand | none | 17.802 | $142.416/hr for p6-b300.48xlarge (Linux, shared tenancy) | high | 2026-11-01 | source |
| Modal | B300 | serverless_per_second | none | 7.0992 | $0.001972/sec | high | 2026-11-02 | source | |
| FlexiCloud (broker) | B300 | HGX SXM | reserved | 12 months or longer, floor rate on the longest term | 4.50 | from $4.50 per GPU-hour | medium | 2026-11-02 | source |
| Civo | B300 | reserved | 36 months | 5.49 | $43.92 per hour for 8x, 36 months commitment | high | 2026-11-02 | source | |
| Civo | B300 | reserved | 24 months | 5.99 | $47.92 per hour for 8x, 24 months commitment | high | 2026-11-02 | source | |
| Civo | B300 | reserved | 12 months | 6.49 | $51.92 per hour for 8x, 12 months commitment | high | 2026-11-02 | source | |
| DigitalOcean | B300 | HGX | reserved | 12 months | 7.94 | $7.94/GPU/hour | high | 2026-11-02 | source |
| Nebius | B300 | HGX | preemptible | none | 0.99 | from $0.99 per GPU-hr (preemptible) | medium | 2026-11-02 | source |
| Verda | B300 | SXM6 268GB | spot | none | 5.73 | $5.730/h (spot) | high | 2026-11-02 | source |
| DigitalOcean | B300 | HGX | spot | none | 6.17 | $6.17/GPU/hour | high | 2026-11-02 | source |
| CoreWeave | B300 | HGX | contact_sales | Contact sales | high | 2026-11-02 | source | ||
| CoreWeave | GB200 | NVL72 | on_demand | none | 10.50 | $42.00/hr per 4-GPU instance | high | 2026-11-02 | source |
| Oracle Cloud (OCI) | GB200 | on_demand | none | 16.00 | $16 GPU Per Hour (B110979) | high | 2026-11-02 | source | |
| Microsoft Azure | GB200 | NVL72, 4x VM | on_demand | none | 27.04 | $108.16/hr for Standard_ND128isr_NDR_GB200_v6 (Linux) | medium | 2026-11-02 | source |
| Crusoe | GB200 | 186GB NVL72 | contact_sales | Contact sales | high | 2026-11-02 | source | ||
| Nebius | GB200 | NVL72 | contact_sales | Contact us | high | 2026-11-02 | source | ||
| Verda | GB300 | SXM6 288GB | on_demand | none | 13.18 | $13.18/h | high | 2026-11-02 | source |
| Oracle Cloud (OCI) | GB300 | on_demand | none | 18.00 | $18 GPU Per Hour (B112140) | high | 2026-11-02 | source | |
| Verda | GB300 | SXM6 288GB | spot | none | 6.59 | $6.590/h (spot) | high | 2026-11-02 | source |
| CoreWeave | GB300 | NVL72 | contact_sales | Contact sales | high | 2026-11-02 | source | ||
| Crusoe | GB300 | 288GB NVL72 | contact_sales | Contact sales | high | 2026-11-02 | source | ||
| Nebius | GB300 | NVL72 | contact_sales | Contact us | high | 2026-11-02 | source | ||
| Civo | A100 | 40GB | on_demand | none | 1.09 | $1.09 per hour for 1x | high | 2026-11-02 | source |
| Thunder Compute | A100 | 80GB | on_demand | none | 1.09 | $1.09/hr | high | 2026-11-02 | source |
| Hyperstack | A100 | on_demand | none | 1.35 | $1.35/hr | high | 2026-11-02 | source | |
| Massed Compute | A100 | 80GB | on_demand | none | 1.35 | $1.35/hr for 1x | medium | 2026-11-02 | source |
| Massed Compute | A100 | SXM4 80GB | on_demand | none | 1.38 | $1.38/hr for 1x | medium | 2026-11-02 | source |
| Hyperstack | A100 | NVL | on_demand | none | 1.40 | $1.40/hr | high | 2026-11-02 | source |
| Hyperstack | A100 | SXM | on_demand | none | 1.60 | $1.60/hr | high | 2026-11-02 | source |
| Verda | A100 | SXM4 40GB | on_demand | none | 1.625 | $1.625/h | high | 2026-11-02 | source |
| RunPod | A100 | PCIe | on_demand | none | 1.79 | $1.79/hr | medium | 2026-11-02 | source |
| RunPod | A100 | SXM | on_demand | none | 1.79 | $1.79/hr | medium | 2026-11-02 | source |
| Lambda | A100 | SXM 40GB | on_demand | none | 1.99 | $1.99 per GPU-hr (8x instance) | high | 2026-11-02 | source |
| Lambda | A100 | PCIe 40GB | on_demand | none | 1.99 | $1.99 per GPU-hr (4x instance) | high | 2026-11-02 | source |
| Lambda | A100 | PCIe 40GB | on_demand | none | 1.99 | $1.99 per GPU-hr (2x instance) | high | 2026-11-02 | source |
| Lambda | A100 | SXM 40GB | on_demand | none | 1.99 | $1.99 per GPU-hr (1x instance) | high | 2026-11-02 | source |
| Lambda | A100 | PCIe 40GB | on_demand | none | 1.99 | $1.99 per GPU-hr (1x instance) | high | 2026-11-02 | source |
| Crusoe | A100 | 80GB PCIe | on_demand | none | 2.00 | $2.00/GPU-hr | medium | 2026-11-02 | source |
| Verda | A100 | SXM4 80GB | on_demand | none | 2.27 | $2.270/h | high | 2026-11-02 | source |
| Crusoe | A100 | 80GB SXM | on_demand | none | 2.30 | $2.30/GPU-hr | medium | 2026-11-02 | source |
| CoreWeave | A100 | on_demand | none | 2.70 | $21.60/hr per 8-GPU instance | high | 2026-11-02 | source | |
| Amazon Web Services | A100 | SXM 40GB, 8x | on_demand | none | 2.7447 | $21.957642/hr for p4d.24xlarge (Linux, shared tenancy) | high | 2026-11-01 | source |
| Lambda | A100 | SXM 80GB | on_demand | none | 2.79 | $2.79 per GPU-hr (8x instance) | high | 2026-11-02 | source |
| Microsoft Azure | A100 | SXM 40GB, 8x VM | on_demand | none | 3.3996 | $27.197/hr for Standard_ND96asr_v4 (Linux) | high | 2026-11-02 | source |
| Amazon Web Services | A100 | SXM 80GB, 8x | on_demand | none | 3.4309 | $27.44705/hr for p4de.24xlarge (Linux, shared tenancy) | high | 2026-11-01 | source |
| Oracle Cloud (OCI) | A100 | on_demand | none | 4.00 | $4 GPU Per Hour (B95907) | high | 2026-11-02 | source | |
| Microsoft Azure | A100 | SXM 80GB, 8x VM | on_demand | none | 4.0963 | $32.77/hr for Standard_ND96amsr_A100_v4 (Linux) | high | 2026-11-02 | source |
| Modal | A100 | 40GB | serverless_per_second | none | 2.0988 | $0.000583/sec | high | 2026-11-02 | source |
| Modal | A100 | 80GB | serverless_per_second | none | 2.4984 | $0.000694/sec | high | 2026-11-02 | source |
| Civo | A100 | 40GB | reserved | 36 months | 0.69 | $0.69 per hour for 1x, 36 months commitment | high | 2026-11-02 | source |
| Civo | A100 | 40GB | reserved | 24 months | 0.79 | $0.79 per hour for 1x, 24 months commitment | high | 2026-11-02 | source |
| Civo | A100 | 40GB | reserved | 12 months | 0.89 | $0.89 per hour for 1x, 12 months commitment | high | 2026-11-02 | source |
| Hyperstack | A100 | reserved | reservation, term not stated on page | 0.95 | $0.95/hr (starting from) | medium | 2026-11-02 | source | |
| Hyperstack | A100 | NVL | reserved | reservation, term not stated on page | 0.98 | $0.98/hr (starting from) | medium | 2026-11-02 | source |
| Civo | A100 | 40GB | reserved | 6 months | 0.99 | $0.99 per hour for 1x, 6 months commitment | high | 2026-11-02 | source |
| Hyperstack | A100 | SXM | reserved | reservation, term not stated on page | 1.36 | $1.36/hr (starting from) | medium | 2026-11-02 | source |
| Microsoft Azure | A100 | SXM 40GB, 8x VM | spot | none | 0.7381 | $5.904469/hr for Standard_ND96asr_v4 (Linux) | high | 2026-11-02 | source |
| Verda | A100 | SXM4 40GB | spot | none | 0.8125 | $0.8125/h (spot) | high | 2026-11-02 | source |
| Microsoft Azure | A100 | SXM 80GB, 8x VM | spot | none | 1.0552 | $8.441552/hr for Standard_ND96amsr_A100_v4 (Linux) | high | 2026-11-02 | source |
| Hyperstack | A100 | spot | none | 1.08 | $1.08/hr (spot) | medium | 2026-11-02 | source | |
| Verda | A100 | SXM4 80GB | spot | none | 1.135 | $1.135/h (spot) | high | 2026-11-02 | source |
| Massed Compute | L40S | 48GB | on_demand | none | 0.97 | $0.97/hr for 1x | medium | 2026-11-02 | source |
| RunPod | L40S | on_demand | none | 1.09 | $1.09/hr | medium | 2026-11-02 | source | |
| Civo | L40S | 48GB | on_demand | none | 1.29 | $1.29 per hour for 1x | high | 2026-11-02 | source |
| Crusoe | L40S | 48GB | on_demand | none | 1.50 | $1.50/GPU-hr | medium | 2026-11-02 | source |
| Nebius | L40S | Intel CPU | on_demand | none | 1.55 | from $1.55 per GPU-hr | high | 2026-11-02 | source |
| DigitalOcean | L40S | on_demand | none | 1.57 | $1.57/GPU/hour | high | 2026-11-02 | source | |
| Nebius | L40S | AMD CPU | on_demand | none | 1.82 | from $1.82 per GPU-hr | high | 2026-11-02 | source |
| Amazon Web Services | L40S | 48GB, 1x | on_demand | none | 1.861 | $1.861/hr for g6e.xlarge (Linux, shared tenancy) | high | 2026-11-01 | source |
| Verda | L40S | 48GB | on_demand | none | 1.92 | $1.920/h | high | 2026-11-02 | source |
| CoreWeave | L40S | on_demand | none | 2.25 | $18.00/hr per 8-GPU instance | high | 2026-11-02 | source | |
| Oracle Cloud (OCI) | L40S | on_demand | none | 3.50 | $3.5 GPU Per Hour (B109479) | high | 2026-11-02 | source | |
| Amazon Web Services | L40S | 48GB, 8x | on_demand | none | 3.7664 | $30.13118/hr for g6e.48xlarge (Linux, shared tenancy) | high | 2026-11-01 | source |
| Modal | L40S | serverless_per_second | none | 1.9512 | $0.000542/sec | high | 2026-11-02 | source | |
| Civo | L40S | 48GB | reserved | 36 months | 0.89 | $0.89 per hour for 1x, 36 months commitment | high | 2026-11-02 | source |
| Civo | L40S | 48GB | reserved | 24 months | 0.99 | $0.99 per hour for 1x, 24 months commitment | high | 2026-11-02 | source |
| Civo | L40S | 48GB | reserved | 12 months | 1.09 | $1.09 per hour for 1x, 12 months commitment | high | 2026-11-02 | source |
| Civo | L40S | 48GB | reserved | 6 months | 1.19 | $1.19 per hour for 1x, 6 months commitment | high | 2026-11-02 | source |
| Nebius | L40S | Intel CPU | preemptible | none | 0.74 | from $0.74** per GPU-hr (preemptible) | medium | 2026-11-02 | source |
| Nebius | L40S | AMD CPU | preemptible | none | 0.90 | from $0.90** per GPU-hr (preemptible) | medium | 2026-11-02 | source |
| Verda | L40S | 48GB | spot | none | 0.96 | $0.9600/h (spot) | high | 2026-11-02 | source |
| Crusoe | MI300X | 192GB | on_demand | none | 3.45 | $3.45/GPU-hr | medium | 2026-11-02 | source |
| Oracle Cloud (OCI) | MI300X | on_demand | none | 6.00 | $6 GPU Per Hour (B109485) | high | 2026-11-02 | source | |
| Oracle Cloud (OCI) | MI355X | on_demand | none | 8.60 | $8.6 GPU Per Hour (B111758) | high | 2026-11-02 | source | |
| Crusoe | MI355X | 288GB | contact_sales | Contact sales | high | 2026-11-02 | source |
How this is built
- Public pricing pages and official public price APIs only (Azure Retail Prices, Oracle price list, AWS price list file). No logins.
- A script reads each page and writes the dataset. If a page changes and cannot be read, the old rows stay and are marked stale. Nothing is estimated.
- Confidence is high where the page states the number plainly, medium where we had to match a name to a price by position, low where the number appears in passing text.
- The free GPU price and checklist bundle is on Gumroad.