ComputeAssay

GPU price check · updated 2026-09-10

H200-SXM rental prices

H200-SXM rents for $3.99–$10.60 per GPU-hour on demand across 9 sellers this week. The spread is not noise — it is mostly the network fabric, which determines what the chip can actually do.

$3.99
lowest on-demand
$6.20
median
6
Grade A/B offerings
0
unverifiable listings

Fair-price band (verifiable training-class, Grade A/B): $4.29–$7.11 per GPU-hour on demand. Below that band, ask what the seller is not telling you about the fabric; above it, you are paying for something the rate card should be able to name.

Every listing we track

ProviderSKU$/GPU-hrTermsGradeFabric (as published)Src
NebiusNVIDIA HGX H200 — preemptible$2.45spotBInfiniBand, 400 Gbps per GPU / 3.2 Tbps per node (per docs.nebius.com)
HyperstackNVIDIA H200 SXM$3.99on-demandCSR-IOV Ethernet up to 350 Gbps inter-VM (default up to 16 Gbps)
DataCrunch (rebranded to Verda, verda.com)H200 SXM5$4.00on-demandCnot published
Crusoeh200-141gb-sxm-ib.8x$4.29on-demandBInfiniBand, 3200 Gbps per 8-GPU instance (per docs VM overview)
NebiusNVIDIA HGX H200 (gpu-h200-sxm platform)$4.50on-demandAInfiniBand, 400 Gbps per GPU / 3.2 Tbps per 8-GPU node, GPUDirect RDMA (per docs.nebius.com)
OVHcloudNvidia H200 (141 GB HBM3e, NVLink, 8-GPU instance)$6.20on-demandCUp to 25 Gbps Ethernet (public/private network)
CoreWeaveNVIDIA HGX H200 (gd-8xh200ib-i128)$6.31on-demandBInfiniBand (instance ID carries 'ib'; CoreWeave HPC interconnect docs state Quantum-2/Quantum-X800 non-blocking fat-tree fabrics, but per-node bandwidth for H200 not stated on fetched pages)
AWSp5en.48xlarge$7.91on-demandAEFAv3 (Nitro v5), 3200 Gbps per instance, 'up to 35% improvement in latency' vs earlier EFA (AWS P5 page)
Oracle Cloud (OCI)BM.GPU.H200.8$10.00on-demand (PAY_AS_YOU_GO)ARDMA cluster network: '8x1x400 Gb/sec RDMA' = 3,200 Gbps per node (Oracle price-list shape table)
Microsoft AzureStandard_ND96isr_H200_v5$10.60on-demandBInfiniBand NDR: dedicated 400 Gbps NVIDIA Quantum-2 CX7 InfiniBand per GPU, 3.2 Tb/s per VM, GPUDirect RDMA (Microsoft Learn)
Google Clouda3-ultragpu-8gCRDMA over Converged Ethernet (RoCE) via ConnectX-7, 3,600 Gbps max per VM, 10 NICs (Google doc table)

Grades summarize deliverability from the seller's own published specs — methodology. A missing fabric field is recorded as unknown, never guessed. Weekly context in the Delivered Compute Report.

Prices move weekly. Get the report.

Headline prices, deliverability grades, and one teardown a week — free.

Subscribe free