Sourcing request
H100 · 16–32 GPUs
Dedicated capacity for your workload.
Quote on requestRegion, topology and availability confirmed per request.
The GPU market, in view
H100, H200 & B200 GPU prices.
Public retail prices are the benchmark. For 16+ GPUs, request a wholesale allocation or put your existing quote through a price-beat audit.
Rates refresh hourly. Observation times and source links are shown for each rate. Published prices do not reserve capacity.
16+ GPUs · Private sourcing
Wholesale blocks.
Request a matched block below. Pricing, interconnect and start dates are confirmed before an allocation is offered.
Sourcing request
H100 · 64+ GPUs
Multi-node cluster sourcing.
Quote on requestRegion, topology and availability confirmed per request.
Public on-demand pricing
Retail benchmarks.
Compare the public market below, then audit a like-for-like quote for your run.
H100
$1.99–$12.29
per GPU-hour
Median $4.04 · Spread $10.3
14 rates · 4 stale
H200
$0.5–$13.78
per GPU-hour
Median $5.1403 · Spread $13.28
7 rates · 2 stale
B200
$5.98–$14.2416
per GPU-hour
Median $6.99 · Spread $8.2616
9 rates · 1 stale
34 rates
| Provider / source | GPU / configuration | USD / GPU-hour | Region | Observation |
|---|---|---|---|---|
| Runpod | B200NVIDIA B2001 GPUs per instance | $5.98$5.98/instance-hour | Not specified | Live |
| Lambda | B200NVIDIA B200 SXM68 GPUs per instance | $6.69$53.52/instance-hour | Not specified | Live |
| Lambda | B200NVIDIA B200 SXM64 GPUs per instance | $6.79$27.16/instance-hour | Not specified | Live |
| Lambda | B200NVIDIA B200 SXM62 GPUs per instance | $6.89$13.78/instance-hour | Not specified | Live |
| Lambda | B200NVIDIA B200 SXM61 GPUs per instance | $6.99$6.99/instance-hour | Not specified | Live |
| Crusoe | B200NVIDIA B200 | $8.5 | Not specified | Stale |
| Nebius | B200NVIDIA HGX B200 | $8.5 | Not specified | StaleExcluded from range |
| CoreWeave | B200NVIDIA HGX B2008 GPUs per instance | $8.6$68.8/instance-hour | Not specified | Live |
| Vast.ai | B200B200 verified on-demand offers | $9.375Offer median $9.9875 (4 offers) | Multiple regions | Live |
| AWS | B200p6-b200.48xlarge8 GPUs per instance | $14.2416$113.9328/instance-hour | us-east-1 | Live |
| Runpod | H100NVIDIA H100 PCIe1 GPUs per instance | $1.99$1.99/instance-hour | Not specified | Stale |
| Runpod | H100NVIDIA H100 NVL1 GPUs per instance | $2.59$2.59/instance-hour | Not specified | Stale |
| Runpod | H100NVIDIA H100 80GB HBM31 GPUs per instance | $2.69$2.69/instance-hour | Not specified | Live |
| Vast.ai | H100H100 verified on-demand offers | $3.0267Offer median $3.0267 (1 offers) | Multiple regions | Stale |
| Lambda | H100NVIDIA H100 PCIe1 GPUs per instance | $3.29$3.29/instance-hour | Not specified | Live |
| Crusoe | H100NVIDIA H100 | $3.9 | Not specified | Stale |
| Lambda | H100NVIDIA H100 SXM8 GPUs per instance | $3.99$31.92/instance-hour | Not specified | Live |
| Lambda | H100NVIDIA H100 SXM4 GPUs per instance | $4.09$16.36/instance-hour | Not specified | Live |
| Lambda | H100NVIDIA H100 SXM2 GPUs per instance | $4.19$8.38/instance-hour | Not specified | Live |
| Lambda | H100NVIDIA H100 SXM1 GPUs per instance | $4.29$4.29/instance-hour | Not specified | Live |
| Nebius | H100NVIDIA HGX H100 | $4.5 | Not specified | StaleExcluded from range |
| CoreWeave | H100NVIDIA HGX H1008 GPUs per instance | $6.155$49.24/instance-hour | Not specified | Live |
| AWS | H100p5.4xlarge1 GPUs per instance | $6.88$6.88/instance-hour | us-east-1 | Live |
| AWS | H100p5.48xlarge8 GPUs per instance | $6.88$55.04/instance-hour | us-east-1 | Live |
| Azure | H100Standard_ND96isr_H100_v58 GPUs per instance | $12.29$98.32/instance-hour | eastus | Live |
| Runpod | H200NVIDIA H200 NVL1 GPUs per instance | $0.5$0.5/instance-hour | Not specified | Stale |
| Runpod | H200NVIDIA H2001 GPUs per instance | $3.59$3.59/instance-hour | Not specified | Live |
| Crusoe | H200NVIDIA H200 | $4.29 | Not specified | Stale |
| Vast.ai | H200H200 verified on-demand offers | $5.1403Offer median $5.6951 (2 offers) | Multiple regions | Live |
| Nebius | H200NVIDIA HGX H200 | $5.4 | Not specified | StaleExcluded from range |
| CoreWeave | H200NVIDIA HGX H2008 GPUs per instance | $6.305$50.44/instance-hour | Not specified | Live |
| AWS | H200p5en.48xlarge8 GPUs per instance | $7.912$63.296/instance-hour | us-east-1 | Live |
| Azure | H200Standard_ND96isr_H200_v58 GPUs per instance | $13.78$110.24/instance-hour | westus | Live |
| AWS | H200p5e.48xlarge8 GPUs per instance | Unavailable | us-east-1 | UnavailableNo successful reading |
Starting with 1–8 GPUs?
Explore self-serve providers.
Check hardware, regions and current availability directly with each provider.
Some links are affiliate links. Sydrion may earn a commission from qualifying referrals.
Ranges, medians and spreads use successful readings from the last 7 days. A failed source keeps its last good price and original timestamp, marked stale. Older readings remain visible in the table and are excluded from the range. Arrows show the median’s movement since the previous reading.
These are compute rates for different configurations and regions. AWS, Azure and CoreWeave whole-instance prices are divided by their verified GPU count. Lambda publishes per-GPU rates for whole instances. Vast.ai shows the lowest and median across verified, rentable on-demand offers. Regions marked “Not specified” are not published by the source.
Storage, networking, taxes and other charges may be additional. Published pricing does not reserve capacity. Sydrion confirms technical fit, total cost and availability for your requirement before an introduction.
The price-beat audit · 16+ GPUs
Already holding a quote for 16+ GPUs?
Upload or paste your spec—we will return 3 vetted bare-metal bids that beat your rate within 24 hours.
- Share your quote and workload.
- We compare total cost, topology and terms.
- Review matched bids before an introduction.