GPU MODELS / H200 SXM
H200 SXM cloud GPU pricing
98 listings across 49 providers. Cheapest published rate: $2.090/GPU/hr at Beam. Sorted by per-GPU hourly price. Every row links to the provider's own published price.
WHAT IS THE H200 SXM
The H200 is an H100 with the memory subsystem rebuilt: 141 GB of HBM3e at 4.8 TB/s versus 80 GB at 3.35. Compute is identical. That makes it the single best Hopper for inference - a 70B model at BF16 fits on one card with KV cache to spare - and a strictly better H100 for memory-bound training. When panel prices put it near H100 money, take the H200.
MAKER
NVIDIA
ARCHITECTURE
Hopper
LAUNCHED
2024
MEMORY
141 GB HBM3e
MEMORY BANDWIDTH
4.8 TB/s
TENSOR COMPUTE
989 TFLOPS FP16/BF16 tensor dense (1,979 sparse); 1,979 TFLOPS FP8 dense
POWER DRAW
700 W
INTERCONNECT
NVLink 4.0 900 GB/s
WHAT IT IS GOOD AT
Large-model inference
The strongest pre-Blackwell serving chip: 141 GB fits 70B BF16 or 34B with very long context on one GPU.
Memory-bound training
Long-sequence and large-batch training benefits directly from 4.8 TB/s.
MoE serving
Mixture-of-experts models with big active-parameter footprints fit where 80 GB parts do not.
WHO SHOULD NOT RENT THE H200 SXM
Workloads that fit comfortably in 80 GB - the memory premium buys nothing an H100 lacks otherwise.
Budget training where A100 rates undercut it heavily.
| Provider | Config | VRAM | Location | $/hr | $/GPU/hr | Source |
|---|---|---|---|---|---|---|
 SpotGPUs | H200 SXM × 1 | 141 GB | Not published | $1.090 | $1.090STALE | receipt |
 QuickPod | H200 × 1 | 141 GB | Not published | $2.000 | $2.000STALE | receipt |
 Beam | H200 (from) × 1 | 141 GB | Not published | $2.090 | $2.090 | receipt |
 Theta EdgeCloud | H200 × 1 | 141 GB | Not published | $2.290 | $2.290 | receipt |
 Baku Compute | H200 SXM (from) × 8 | 141 GB | Azerbaijan (Baku) | $19.200 | $2.400 | receipt |
 CloudRift | H200 × 1 | 141 GB | Not published | $2.500 | $2.500 | receipt |
 Yotta Labs | H200 × 1 | 141 GB | Not published | $2.500 | $2.500 | receipt |
 HydraHost | H200 × 1 | 141 GB | Not published | $2.500 | $2.500 | receipt |
 Applications.kz | H200 (from) × 1 | 141 GB | Kazakhstan | $2.600 | $2.600 | receipt |
 Seeweb | H200 × 1 | 141 GB | Italy | $2.889 | $2.889STALE | receipt |
 Koyeb | H200 × 1 | 141 GB | Not published | $3.000 | $3.000 | receipt |
 InHosted | H200 × 1 | 141 GB | India | $3.195 | $3.195 | receipt |
 GMI Cloud | H200 (from) × 1 | 141 GB | Not published | $3.200 | $3.200 | receipt |
 Civo | H200 SXM × 1 | 141 GB | Not published | $3.490 | $3.490 | receipt |
 Deep Infra | H200 × 1 | 141 GB | Not published | $3.490 | $3.490 | receipt |
| Civo | H200 SXM × 8 | 141 GB | Not published | $27.920 | $3.490 | receipt |
 Lium | H200 × 1 | 141 GB | Not published | $3.500 | $3.500 | receipt |
 Atlas Cloud | H200 × 1 | 141 GB | Not published | $3.500 | $3.500 | receipt |
 Taiga Cloud (Northern Data) | H200 (from) × 8 | 141 GB | Europe | $28.000 | $3.500 | receipt |
 Gcore | H200 (from) × 1 | 141 GB | Not published | $3.534 | $3.534 | receipt |
 Stratos Lab | H200 × 1 | 141 GB | South Africa | $3.580 | $3.580 | receipt |
| Stratos Lab | H200 × 8 | 141 GB | South Africa | $28.670 | $3.584 | receipt |
| Stratos Lab | H200 × 2 | 141 GB | South Africa | $7.170 | $3.585 | receipt |
| Stratos Lab | H200 × 4 | 141 GB | South Africa | $14.340 | $3.585 | receipt |
 RunPod | H200 SXM × 1 | 141 GB | Global | $3.590 | $3.590 | receipt |
 Hyperstack | H200 SXM × 1 | 141 GB | US / EU / CA | $3.990 | $3.990 | receipt |
 Hyperbolic | H200 (from) × 1 | 141 GB | Not published | $3.990 | $3.990 | receipt |
 Core42 | H200 (from) × 1 | 141 GB | UAE | $4.000 | $4.000 | receipt |
 Cerebrium | H200 × 1 | 141 GB | Not published | $4.198 | $4.198 | receipt |
| H200 × 1 | 141 GB | Global | $4.450 | $4.450STALE | receipt | |
 DigitalOcean | H200 × 1 | 141 GB | Not published | $4.470 | $4.470 | receipt |
 Runware | H200 × 1 | 141 GB | Global | $4.500 | $4.500 | receipt |
 Lightning AI | H200 × 1 | 141 GB | US | $4.500 | $4.500STALE | receipt |
 Daytona | H200 × 1 | 141 GB | Not published | $4.540 | $4.540 | receipt |
 JarvisLabs | H200 SXM × 1 | 141 GB | India | $4.590 | $4.590 | receipt |
 Neysa | H200 SXM (from) × 1 | 141 GB | India | $4.720 | $4.720 | receipt |
 Machine0 | H200 × 1 | 141 GB | Not published | $4.917 | $4.917 | receipt |
 GPUBrazil | H200 × 1 | 140 GB | North America | $4.939 | $4.939 | receipt |
| GPUBrazil | H200 × 1 | 140 GB | Europe | $5.004 | $5.004 | receipt |
 E2E Networks | H200 × 1 | 141 GB | India | $5.355 | $5.355 | receipt |
| GPUBrazil | H200 × 1 | 140 GB | Global | $5.460 | $5.460 | receipt |
 Replicate | H200 × 1 | 141 GB | Global | $5.490 | $5.490 | receipt |
| Replicate | H200 × 2 | 141 GB | Global | $10.980 | $5.490 | receipt |
| Replicate | H200 × 4 | 141 GB | Global | $21.960 | $5.490 | receipt |
| Replicate | H200 × 8 | 141 GB | Global | $43.920 | $5.490 | receipt |
 Spheron | H200 × 1 | 141 GB | Not published | $5.520 | $5.520 | receipt |
 DCP | H200 (from) × 1 | 141 GB | Saudi Arabia | $6.147 | $6.147 | receipt |
 OVHcloud | H200 (from) × 8 | 141 GB | Not published | $49.560 | $6.195 | receipt |
 CoreWeave | H200 × 8 | 141 GB | North America / Europe | $50.440 | $6.305 | receipt |
 gpu.ai | H200 SXM × 1 | 141 GB | Not published | $6.320 | $6.320 | receipt |
 Verda (formerly DataCrunch) | H200 SXM5 × 8 | 141 GB | Europe (Finland / Iceland) | $52.030 | $6.504 | receipt |
| Verda (formerly DataCrunch) | H200 SXM5 × 1 | 141 GB | Europe (Finland / Iceland) | $6.504 | $6.504 | receipt |
| Verda (formerly DataCrunch) | H200 SXM5 × 2 | 141 GB | Europe (Finland / Iceland) | $13.010 | $6.505 | receipt |
| Verda (formerly DataCrunch) | H200 SXM5 × 4 | 141 GB | Europe (Finland / Iceland) | $26.020 | $6.505 | receipt |
 Hostnot GPU | H200 × 2 | 141 GB | US East | $13.900 | $6.950 | receipt |
| Hostnot GPU | H200 × 1 | 141 GB | US East | $6.960 | $6.960 | receipt |
| Hostnot GPU | H200 × 1 | 141 GB | Asia-Pacific | $6.990 | $6.990 | receipt |
| Hostnot GPU | H200 × 1 | 141 GB | EU West | $6.990 | $6.990 | receipt |
| Hostnot GPU | H200 × 1 | 141 GB | US East | $6.990 | $6.990 | receipt |
| GPUBrazil | H200 × 1 | 140 GB | Asia-Pacific | $7.783 | $7.783 | receipt |
 Selectel | H200 × 1 | 141 GB | Russia | $7.877 | $7.877 | receipt |
 AWS | H200 × 8 | 141 GB | US East (N. Virginia) | $63.296 | $7.912 | receipt |
 VoltageGPU | H200 × 1 | 141 GB | Not published | $8.080 | $8.080STALE | receipt |
| Hostnot GPU | H200 × 8 | 141 GB | Asia-Pacific | $72.090 | $9.011 | receipt |
 Oracle OCI | H200 × 1 | 141 GB | Not published (global list) | $10.000 | $10.000 | receipt |
 Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | US West 2 | $84.800 | $10.600 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | US East 2 | $84.800 | $10.600 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | US West 3 | $84.800 | $10.600 | receipt |
 Google Cloud | H200 × 8 | 141 GB | Iowa (us-central1) | $84.807 | $10.601STALE | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | US South Central 2 | $93.280 | $11.660 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | MX Central | $93.280 | $11.660 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | US Central | $95.824 | $11.978 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | ID Central | $99.216 | $12.402 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | MY West | $99.216 | $12.402 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | EU North | $101.760 | $12.720 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | US North Central | $101.760 | $12.720 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | US Gov Virginia | $106.000 | $13.250 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | UK South | $106.000 | $13.250 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | US Gov AZ | $106.000 | $13.250 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | KR South | $106.000 | $13.250 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | JA East | $106.000 | $13.250 | receipt |
| Microsoft Azure | H200 [ND96isrf_H200_v5] × 8 | 141 GB | US Gov AZ | $106.000 | $13.250 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | FR Central | $106.000 | $13.250 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | US West | $110.240 | $13.780 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | AP Southeast | $110.240 | $13.780 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | ES Central | $110.240 | $13.780 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | DE West Central | $110.240 | $13.780 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | IT North | $110.240 | $13.780 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | AU Southeast | $110.240 | $13.780 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | SE Central | $110.272 | $13.784 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | UK West | $111.088 | $13.886 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | NO East | $121.264 | $15.158 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | AE North | $121.264 | $15.158 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | AU East | $122.960 | $15.370 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | ZA North | $123.808 | $15.476 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | PL Central | $130.592 | $16.324 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | JA West | $131.440 | $16.430 | receipt |
| Microsoft Azure | H200 [ND96isr_H200_v5] × 8 | 141 GB | BR South | $169.600 | $21.200 | receipt |
THE H200 SXM MARKET, BY THE NUMBERS
PROVIDERS
49
CONFIGURATIONS
98
MEDIAN $/GPU/HR
$6.50
SPREAD
10.1x
49 providers publish 98 H200 SXM configurations on the panel today. Published per-GPU hourly rates run from $2.090 (Beam) to $21.20 (Microsoft Azure), a 10x spread between the cheapest and the most expensive published rate for the same chip. The median listing sits at $6.50/GPU/hr. 11 rows were reported in stock by the provider at last observation; the rest are published catalog rates. All rates are on-demand - committed-use and reserved pricing is excluded.
CHEAPEST H200 SXM BY GEOGRAPHY
indiaInHosted$3.195/hrsouth africaStratos Lab$3.580/hrunited statesHyperstack$3.990/hrcanadaHyperstack$3.990/hruaeCore42$4.000/hrfinlandVerda (formerly DataCrunch)$6.504/hricelandVerda (formerly DataCrunch)$6.504/hrmexicoMicrosoft Azure$11.660/hrH200 SXM PRICE TREND, DAILY PANEL SNAPSHOTS
Across every H200 SXM row on the panel, the daily floor moved from $2.500 to $2.090/GPU/hr over 53 days of tracking (-16.4%). The median continuously-tracked listing moved from $3.99 to $4.59 (+15.0%). Snapshots run daily; the full history is public in the repo.
H200 SXM PROVIDER NOTES
from $2.090/hr
from $2.290/hr
from $2.400/hr
from $2.500/hr
from $2.500/hr
from $2.500/hr
from $2.600/hr
from $3.000/hr
BUYING H200 SXM: WHAT THE PANEL SAYS
At the current floor, one H200 SXM running around the clock costs about $1,526 a month (730 hours of on-demand arithmetic). Node shape matters: the single-GPU floor is $2.090/GPU/hr against $2.400/GPU/hr on an 8-GPU node (+14.8%). Per-GPU floors by config size: 1x $2.09 · 2x $3.58 · 4x $3.58 · 8x $2.40. Compare the alternatives below before committing - the same budget often buys more than one chip class.
H200 SXM VS THE ALTERNATIVES
H200 SXM vs H100 SXM
Identical compute; 76% more memory and 43% more bandwidth. Training jobs that fit in 80 GB see little difference - pick on price. Inference and memory-bound work favor the H200 outright.
H200 SXM vs B200
B200 adds ~2.3x dense compute and 8 TB/s, at 1000 W and early-cycle prices. H200 is the mature, available, usually cheaper option.
Compare live rates: H200 SXM vs H100 SXM · H200 SXM vs B200 · H200 SXM vs MI300X · H200 SXM vs MI325X · H200 SXM vs A100 80GB SXM
H200 SXM PRICING FAQ
What is the cheapest H200 SXM cloud GPU right now?
The cheapest published H200 SXM rate on the panel is $2.090 per GPU-hour at Beam (H200 (from) x 1, Not published). Every price links to the provider's own published rate.
How much does a H200 SXM cost per hour?
Across 49 providers tracked today, published per-GPU hourly rates for H200 SXM run from $2.09 to $21.20, with a median of $6.50. Committed-use and reserved rates are excluded; everything shown is on-demand.
How many providers rent H200 SXM GPUs?
49 providers publish 98 distinct H200 SXM configurations on Compute Cafe. 11 of those rows were reported in stock by the provider's live endpoint at last observation.
Where is H200 SXM cheapest?
By published per-GPU hourly rate, the cheapest geography for H200 SXM today is india at $3.19/hr (InHosted). The spread across 22 tracked geographies runs to $21.20/hr in brazil.
Are H200 SXM prices going up or down?
Over the 53 days Compute Cafe has tracked the panel, the H200 SXM floor is falling: $2.500 to $2.090/GPU/hr (-16.4%). The median continuously-tracked listing moved from $3.99 to $4.59 (+15.0%). Daily snapshots are public in the repo, so the trend is auditable.
What does a H200 SXM cost per month?
At the current floor of $2.090/GPU/hr, a H200 SXM running 24/7 costs about $1,526 per month (730 hours of on-demand arithmetic on the cheapest published rate). Committed-use contracts can price lower; the panel tracks on-demand rates only.
Is a full 8-GPU H200 SXM node cheaper than a single card?
On today's panel: the single-GPU floor is $2.090/GPU/hr and the 8-GPU node floor is $2.400/GPU/hr - more expensive per GPU on a full node (+14.8%). Check the table for the exact configs behind each floor.
H200 vs H100 - which should I rent?
For inference of models over ~40B or with long context, the H200's 141 GB and 4.8 TB/s win outright. For training that fits in 80 GB, compute is identical, so the cheaper hourly rate on the panel is the right answer.
See also alternatives to the H200 SXM · cheapest GPU-hour per model · inference chips · training chips · by location