GPU MODELS / RTX 5090

RTX 5090 cloud GPU pricing

57 listings across 26 providers. Cheapest published rate: $0.250/GPU/hr at Salad. Sorted by per-GPU hourly price. Every row links to the provider's own published price.

WHAT IS THE RTX 5090

The RTX 5090 is the consumer Blackwell flagship: 32 GB of fast GDDR7 (1.79 TB/s), FP4/FP8 tensor support, and roughly 4090-plus-30% real-world AI throughput. Supply is tight and cloud pricing is still settling - where the panel shows it near 4090 money, it is the better card for everything except multi-GPU linking.

MAKER
NVIDIA
ARCHITECTURE
Blackwell
LAUNCHED
2025
MEMORY
32 GB GDDR7
MEMORY BANDWIDTH
1.79 TB/s
TENSOR COMPUTE
Blackwell tensor cores with FP4; ~3,352 AI TOPS (FP4 sparse)
POWER DRAW
575 W
INTERCONNECT
PCIe 5.0 x16

WHAT IT IS GOOD AT

Small-model inference
13B quantized at high throughput; 32 GB fits what the 4090 cannot.
Fine-tuning
QLoRA up to ~20B-class models.
FP4 experimentation
The cheapest way onto Blackwell FP4.

WHO SHOULD NOT RENT THE RTX 5090

Multi-GPU linked setups - no NVLink on consumer Blackwell.
Budget builds where 4090 rates undercut it for the same 13B-class jobs.
ProviderConfigVRAMLocation$/hr$/GPU/hrSource
SpotGPUsSpotGPUsRTX 5090 × 132 GBNot published$0.190$0.190STALEreceipt
SaladSaladRTX 5090 × 132 GBGlobal (distributed)$0.250$0.250receipt
OpenBayesOpenBayesRTX 5090 × 132 GBChina$0.434$0.434receipt
AutoDLAutoDLRTX 5090 × 132 GBChina$0.438$0.438receipt
FeaturizeFeaturizeRTX 5090 × 132 GBChina$0.449$0.449receipt
SimplePodSimplePodRTX 5090 × 132 GBNot published$0.450$0.450receipt
Theta EdgeCloudTheta EdgeCloudRTX 5090 × 132 GBNot published$0.490$0.490receipt
Suanjia CloudSuanjia CloudRTX 5090 × 132 GBChina$0.491$0.491receipt
LiumLiumRTX 5090 × 132 GBNot published$0.500$0.500receipt
QuickPodQuickPodRTX 5090 × 132 GBNot published$0.500$0.500STALEreceipt
Novita AINovita AIRTX 5090 × 132 GBNot published$0.560$0.560receipt
gpu.aigpu.aiRTX 5090 × 132 GBNot published$0.560$0.560receipt
Gitee AIGitee AIRTX 5090 × 132 GBChina$0.598$0.598receipt
Hostnot GPUHostnot GPURTX 5090 × 832 GBUS East$4.940$0.618receipt
Hostnot GPUHostnot GPURTX 5090 × 232 GBAsia-Pacific$1.250$0.625receipt
CloudRiftCloudRiftRTX 5090 × 132 GBNot published$0.650$0.650receipt
Hostnot GPUHostnot GPURTX 5090 × 132 GBAsia-Pacific$0.650$0.650receipt
AkashAkashRTX 5090 × 132 GBGlobal$0.670$0.670STALEreceipt
RunPodRunPodRTX 5090 × 132 GBGlobal$0.690$0.690receipt
Yotta LabsYotta LabsRTX 5090 × 132 GBNot published$0.700$0.700receipt
HydraHostHydraHostRTX 5090 × 132 GBNot published$0.700$0.700receipt
Hostnot GPUHostnot GPURTX 5090 × 432 GBAsia-Pacific$2.820$0.705receipt
Hostnot GPUHostnot GPURTX 5090 × 832 GBAsia-Pacific$5.650$0.706receipt
Hostnot GPUHostnot GPURTX 5090 × 132 GBUS West$0.720$0.720receipt
BeamBeamRTX 5090 (from) × 132 GBNot published$0.720$0.720receipt
Hostnot GPUHostnot GPURTX 5090 × 132 GBUS East$0.730$0.730receipt
NovaGPU (IP ServerOne)NovaGPU (IP ServerOne)RTX 5090 × 132 GBMalaysia$0.750$0.750receipt
NovaGPU (IP ServerOne)NovaGPU (IP ServerOne)RTX 5090 × 4128 GBMalaysia$3.000$0.750receipt
NovaGPU (IP ServerOne)NovaGPU (IP ServerOne)RTX 5090 × 264 GBMalaysia$1.501$0.751receipt
SpheronSpheronRTX 5090 × 132 GBNot published$0.780$0.780receipt
Hostnot GPUHostnot GPURTX 5090 × 132 GBEU East$0.830$0.830receipt
Hostnot GPUHostnot GPURTX 5090 × 232 GBEU East$1.670$0.835receipt
Hostnot GPUHostnot GPURTX 5090 × 132 GBCanada Central$0.860$0.860receipt
Hostnot GPUHostnot GPURTX 5090 × 132 GBEU West$0.880$0.880receipt
Hostnot GPUHostnot GPURTX 5090 × 532 GBEU East$4.420$0.884receipt
Hostnot GPUHostnot GPURTX 5090 × 432 GBEU East$3.540$0.885receipt
GPUBrazilGPUBrazilRTX 5090 × 132 GBAsia-Pacific$0.927$0.927receipt
Hostnot GPUHostnot GPURTX 5090 × 232 GBEU West$1.940$0.970receipt
Hostnot GPUHostnot GPURTX 5090 × 832 GBEU West$8.130$1.016receipt
Hostnot GPUHostnot GPURTX 5090 × 432 GBEU West$4.240$1.060receipt
GPUBrazilGPUBrazilRTX 5090 × 132 GBNorth America$1.126$1.126receipt
GPUBrazilGPUBrazilRTX 5090 × 132 GBEurope$1.192$1.192receipt
Hostnot GPUHostnot GPURTX 5090 × 432 GBCanada Central$4.930$1.232receipt
Hostnot GPUHostnot GPURTX 5090 × 432 GBUS West$5.080$1.270receipt
Hostnot GPUHostnot GPURTX 5090 × 232 GBUS West$2.550$1.275receipt
DaytonaDaytonaRTX 5090 × 132 GBNot published$1.290$1.290receipt
DCPDCPRTX 5090 (from) × 132 GBSaudi Arabia$1.387$1.387receipt
GPUBrazilGPUBrazilRTX 5090 × 132 GBGlobal$1.577$1.577receipt
Hostnot GPUHostnot GPURTX 5090 × 132 GBEU East$1.580$1.580receipt
Hostnot GPUHostnot GPURTX 5090 × 232 GBUS East$3.170$1.585receipt
Hostnot GPUHostnot GPURTX 5090 × 432 GBUS East$6.350$1.587receipt
Hostnot GPUHostnot GPURTX 5090 × 432 GBUS Central$6.630$1.657receipt
Hostnot GPUHostnot GPURTX 5090 × 832 GBEU East$13.840$1.730receipt
Hostnot GPUHostnot GPURTX 5090 × 232 GBCanada Central$3.520$1.760receipt
iRenderiRenderRTX 5090 × 4128 GBVietnam$17.480$4.370receipt
iRenderiRenderRTX 5090 × 264 GBVietnam$8.934$4.467receipt
iRenderiRenderRTX 5090 × 132 GBVietnam$4.661$4.661receipt

THE RTX 5090 MARKET, BY THE NUMBERS

PROVIDERS
26
CONFIGURATIONS
57
MEDIAN $/GPU/HR
$0.83
SPREAD
18.6x

26 providers publish 57 RTX 5090 configurations on the panel today. Published per-GPU hourly rates run from $0.250 (Salad) to $4.66 (iRender), a 19x spread between the cheapest and the most expensive published rate for the same chip. The median listing sits at $0.83/GPU/hr. 31 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 RTX 5090 BY GEOGRAPHY
chinaOpenBayes$0.434/hrunited statesHostnot GPU$0.618/hrmalaysiaNovaGPU (IP ServerOne)$0.750/hrcanadaHostnot GPU$0.860/hrvietnamiRender$4.370/hr
RTX 5090 PRICE TREND, DAILY PANEL SNAPSHOTS

Across every RTX 5090 row on the panel, the daily floor held flat at $0.250/GPU/hr over 53 days of tracking. The median continuously-tracked listing held flat at $0.69. Snapshots run daily; the full history is public in the repo.

RTX 5090 PROVIDER NOTES
Salad · 1 config · 1 region
Sets the panel floor for this model.
from $0.250/hr
OpenBayes · 1 config · 1 region
Published catalog rates.
from $0.434/hr
AutoDL · 1 config · 1 region
Published catalog rates.
from $0.438/hr
Featurize · 1 config · 1 region
Published catalog rates.
from $0.449/hr
SimplePod · 1 config
Published catalog rates.
from $0.450/hr
Theta EdgeCloud · 1 config
Published catalog rates.
from $0.490/hr
Suanjia Cloud · 1 config · 1 region
Published catalog rates.
from $0.491/hr
Lium · 1 config
Published catalog rates.
from $0.500/hr
BUYING RTX 5090: WHAT THE PANEL SAYS

At the current floor, one RTX 5090 running around the clock costs about $183 a month (730 hours of on-demand arithmetic). Node shape matters: the single-GPU floor is $0.250/GPU/hr against $0.618/GPU/hr on an 8-GPU node (+147.0%). Per-GPU floors by config size: 1x $0.25 · 2x $0.63 · 4x $0.70 · 5x $0.88 · 8x $0.62. Compare the alternatives below before committing - the same budget often buys more than one chip class.

RTX 5090 VS THE ALTERNATIVES

RTX 5090 vs RTX 4090
5090: 32 GB vs 24 GB, 1.79 vs 1.0 TB/s, newer tensor cores. The 4090 counters on price and availability - the panel spread decides.

Compare live rates: RTX 5090 vs RTX 4090 · RTX 5090 vs RTX 5080 · RTX 5090 vs RTX 4090D · RTX 5090 vs RTX PRO 5000 · RTX 5090 vs RTX PRO 6000

RTX 5090 PRICING FAQ

What is the cheapest RTX 5090 cloud GPU right now?
The cheapest published RTX 5090 rate on the panel is $0.250 per GPU-hour at Salad (RTX 5090 x 1, Global (distributed)). Every price links to the provider's own published rate.
How much does a RTX 5090 cost per hour?
Across 26 providers tracked today, published per-GPU hourly rates for RTX 5090 run from $0.25 to $4.66, with a median of $0.83. Committed-use and reserved rates are excluded; everything shown is on-demand.
How many providers rent RTX 5090 GPUs?
26 providers publish 57 distinct RTX 5090 configurations on Compute Cafe. 31 of those rows were reported in stock by the provider's live endpoint at last observation.
Where is RTX 5090 cheapest?
By published per-GPU hourly rate, the cheapest geography for RTX 5090 today is china at $0.43/hr (OpenBayes). The spread across 5 tracked geographies runs to $4.37/hr in vietnam.
Are RTX 5090 prices going up or down?
Over the 53 days Compute Cafe has tracked the panel, the RTX 5090 floor is holding flat: $0.250 to $0.250/GPU/hr (+0.0%). The median continuously-tracked listing moved from $0.69 to $0.69 (+0.0%). Daily snapshots are public in the repo, so the trend is auditable.
What does a RTX 5090 cost per month?
At the current floor of $0.250/GPU/hr, a RTX 5090 running 24/7 costs about $183 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 RTX 5090 node cheaper than a single card?
On today's panel: the single-GPU floor is $0.250/GPU/hr and the 8-GPU node floor is $0.618/GPU/hr - more expensive per GPU on a full node (+147.0%). Check the table for the exact configs behind each floor.
RTX 5090 vs 4090 for AI?
The 5090 wins on memory (32 vs 24 GB), bandwidth (1.79 vs 1.0 TB/s) and per-card throughput. The 4090 wins on cloud availability and often price. Check live panel rates for both.

See also alternatives to the RTX 5090 · cheapest GPU-hour per model · inference chips · training chips · by location