GPU MODELS / RTX 4090

RTX 4090 cloud GPU pricing

76 listings across 39 providers. Cheapest published rate: $0.160/GPU/hr at Salad. Sorted by per-GPU hourly price. Every row links to the provider's own published price.

WHAT IS THE RTX 4090

The RTX 4090 is the consumer card that built the budget GPU-cloud market: L40S-class FP16 throughput at a fraction of the price, limited by 24 GB of memory, no ECC, and no datacenter software features. For fine-tuning small models and high-throughput small-model inference, nothing on the panel beats its $/FLOP.

MAKER
NVIDIA
ARCHITECTURE
Ada Lovelace
LAUNCHED
2022
MEMORY
24 GB GDDR6X
MEMORY BANDWIDTH
1.0 TB/s
TENSOR COMPUTE
330 TFLOPS FP16 tensor dense (660 sparse); FP8 supported
POWER DRAW
450 W
INTERCONNECT
PCIe 4.0 x16

WHAT IT IS GOOD AT

Small-model inference
The $/token king for models that fit in 24 GB.
Fine-tuning 7B-13B
QLoRA on 13B, LoRA on 7B - the budget-lab default.
Image generation
Diffusion pipelines at very low cost.

WHO SHOULD NOT RENT THE RTX 4090

Models beyond 24 GB without multi-card tricks - no NVLink, no ECC, no datacenter features.
Production fleets needing support contracts or vGPU.
ProviderConfigVRAMLocation$/hr$/GPU/hrSource
SpotGPUsSpotGPUsRTX 4090 × 124 GBNot published$0.090$0.090STALEreceipt
QuickPodQuickPodRTX 4090 × 124 GBNot published$0.100$0.100STALEreceipt
SaladSaladRTX 4090 × 124 GBGlobal (distributed)$0.160$0.160receipt
CloreCloreRTX 4090 × 124 GBRU$0.208$0.208receipt
AI GalaxyAI GalaxyRTX 4090 × 124 GBChina$0.247$0.247receipt
IO.NETIO.NETRTX 4090 × 124 GBNot published$0.250$0.250receipt
Theta EdgeCloudTheta EdgeCloudRTX 4090 × 124 GBNot published$0.260$0.260receipt
FeaturizeFeaturizeRTX 4090 × 125.2 GBChina$0.280$0.280receipt
4090Xianka4090XiankaRTX 4090 × 124 GBChina$0.295$0.295receipt
AutoDLAutoDLRTX 4090 × 124 GBChina$0.296$0.296receipt
LiumLiumRTX 4090 × 124 GBNot published$0.300$0.300receipt
SimplePodSimplePodRTX 4090 × 124 GBNot published$0.300$0.300receipt
Suanjia CloudSuanjia CloudRTX 4090 × 124 GBChina$0.311$0.311receipt
MatPoolMatPoolRTX 4090 × 124 GBChina$0.329$0.329receipt
RunPodRunPodRTX 4090 × 124 GBGlobal$0.340$0.340receipt
PPIOPPIORTX 4090 × 124 GBChina$0.345$0.345receipt
4090Xianka4090XiankaRTX 4090 × 124 GBTaiwan$0.375$0.375receipt
AtmosComputeAtmosComputeRTX 4090 (from) × 124 GBNot published$0.380$0.380receipt
CloudGPUCloudGPURTX 4090 × 124 GBNot published$0.390$0.390receipt
AkashAkashRTX 4090 × 124 GBGlobal$0.390$0.390STALEreceipt
4090Xianka4090XiankaRTX 4090 × 124 GBHong Kong$0.399$0.399receipt
HydraHostHydraHostRTX 4090 × 124 GBNot published$0.400$0.400receipt
TensorDockTensorDockRTX 4090 × 124 GBMultiple regions$0.410$0.410receipt
gpu.aigpu.aiRTX 4090 × 124 GBNot published$0.410$0.410receipt
Gitee AIGitee AIRTX 4090 × 124 GBChina$0.416$0.416receipt
Hostnot GPUHostnot GPURTX 4090 × 124 GBAsia-Pacific$0.440$0.440receipt
BeamBeamRTX 4090 (from) × 124 GBNot published$0.440$0.440receipt
CloudRiftCloudRiftRTX 4090 × 124 GBNot published$0.450$0.450receipt
Hostnot GPUHostnot GPURTX 4090 × 124 GBCanada Central$0.460$0.460receipt
Hostnot GPUHostnot GPURTX 4090 × 124 GBEU East$0.460$0.460receipt
Hostnot GPUHostnot GPURTX 4090 × 124 GBEU West$0.460$0.460receipt
Hostnot GPUHostnot GPURTX 4090 × 124 GBUS Central$0.460$0.460receipt
Hostnot GPUHostnot GPURTX 4090 × 124 GBUS East$0.460$0.460receipt
Hostnot GPUHostnot GPURTX 4090 × 124 GBUS West$0.460$0.460receipt
Hostnot GPUHostnot GPURTX 4090 × 224 GBAsia-Pacific$0.920$0.460receipt
Yotta LabsYotta LabsRTX 4090 × 124 GBNot published$0.480$0.480receipt
SpheronSpheronRTX 4090 × 124 GBNot published$0.480$0.480receipt
Hostnot GPUHostnot GPURTX 4090 × 424 GBUS East$2.050$0.512receipt
GPUBrazilGPUBrazilRTX 4090 × 124 GBEurope$0.517$0.517receipt
GPUBrazilGPUBrazilRTX 4090 × 124 GBGlobal$0.517$0.517receipt
Hostnot GPUHostnot GPURTX 4090 × 224 GBUS East$1.040$0.520receipt
Hostnot GPUHostnot GPURTX 4090 × 424 GBAsia-Pacific$2.130$0.532receipt
ParateraParateraRTX 4090 × 124 GBChina$0.538$0.538receipt
Hostnot GPUHostnot GPURTX 4090 × 224 GBEU West$1.100$0.550receipt
NovaGPU (IP ServerOne)NovaGPU (IP ServerOne)RTX 4090 × 496 GBMalaysia$2.329$0.582receipt
NovaGPU (IP ServerOne)NovaGPU (IP ServerOne)RTX 4090 × 124 GBMalaysia$0.582$0.582receipt
NovaGPU (IP ServerOne)NovaGPU (IP ServerOne)RTX 4090 × 248 GBMalaysia$1.165$0.582receipt
Hostnot GPUHostnot GPURTX 4090 × 424 GBEU East$2.400$0.600receipt
Hostnot GPUHostnot GPURTX 4090 × 224 GBEU East$1.210$0.605receipt
Hostnot GPUHostnot GPURTX 4090 × 824 GBCanada Central$4.930$0.616receipt
Hostnot GPUHostnot GPURTX 4090 × 424 GBCanada Central$2.470$0.618receipt
Hostnot GPUHostnot GPURTX 4090 × 224 GBUS West$1.250$0.625receipt
OblivusOblivusRTX 4090 × 124 GBNot published$0.640$0.640receipt
packet.aipacket.aiRTX 4090 × 124 GBUS Central · Texas$0.660$0.660receipt
Novita AINovita AIRTX 4090 × 124 GBNot published$0.670$0.670receipt
NovaGPU (IP ServerOne)NovaGPU (IP ServerOne)RTX 4090 × 148 GBMalaysia$0.683$0.683receipt
NovaGPU (IP ServerOne)NovaGPU (IP ServerOne)RTX 4090 × 4192 GBMalaysia$2.732$0.683receipt
NovaGPU (IP ServerOne)NovaGPU (IP ServerOne)RTX 4090 × 296 GBMalaysia$1.366$0.683receipt
SynexaSynexaRTX 4090 × 124 GBGlobal$0.690$0.690receipt
Hostnot GPUHostnot GPURTX 4090 × 424 GBEU West$2.830$0.708receipt
Hostnot GPUHostnot GPURTX 4090 × 824 GBEU East$6.340$0.792receipt
Hostnot GPUHostnot GPURTX 4090 × 824 GBUS East$6.690$0.836receipt
GPUBrazilGPUBrazilRTX 4090 × 124 GBNorth America$0.861$0.861receipt
DCPDCPRTX 4090 (from) × 124 GBSaudi Arabia$0.965$0.965receipt
DaytonaDaytonaRTX 4090 × 124 GBNot published$0.990$0.990receipt
SelectelSelectelRTX 4090 × 124 GBRussia$1.011$1.011receipt
GPUBrazilGPUBrazilRTX 4090 × 124 GBAsia-Pacific$1.060$1.060receipt
GPUBrazilGPUBrazilRTX 4090 × 124 GBSouth America$1.060$1.060receipt
Hostnot GPUHostnot GPURTX 4090 × 124 GBEU East$1.180$1.180receipt
Hostnot GPUHostnot GPURTX 4090 × 124 GBEU West$1.180$1.180receipt
Hostnot GPUHostnot GPURTX 4090 × 124 GBUS Central$1.180$1.180receipt
Hostnot GPUHostnot GPURTX 4090 × 124 GBUS West$1.180$1.180receipt
iRenderiRenderRTX 4090 × 8192 GBVietnam$21.753$2.719receipt
iRenderiRenderRTX 4090 × 496 GBVietnam$11.653$2.913receipt
iRenderiRenderRTX 4090 × 248 GBVietnam$6.215$3.108receipt
iRenderiRenderRTX 4090 × 124 GBVietnam$3.496$3.496receipt

THE RTX 4090 MARKET, BY THE NUMBERS

PROVIDERS
39
CONFIGURATIONS
76
MEDIAN $/GPU/HR
$0.52
SPREAD
21.8x

39 providers publish 76 RTX 4090 configurations on the panel today. Published per-GPU hourly rates run from $0.160 (Salad) to $3.50 (iRender), a 22x spread between the cheapest and the most expensive published rate for the same chip. The median listing sits at $0.52/GPU/hr. 32 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 4090 BY GEOGRAPHY
chinaAI Galaxy$0.247/hrcanadaHostnot GPU$0.460/hrunited statesHostnot GPU$0.460/hrmalaysiaNovaGPU (IP ServerOne)$0.582/hrvietnamiRender$2.719/hr
RTX 4090 PRICE TREND, DAILY PANEL SNAPSHOTS

Across every RTX 4090 row on the panel, the daily floor held flat at $0.160/GPU/hr over 53 days of tracking. The median continuously-tracked listing held flat at $0.34. Snapshots run daily; the full history is public in the repo.

RTX 4090 PROVIDER NOTES
Salad · 1 config · 1 region
Sets the panel floor for this model.
from $0.160/hr
Clore · 1 config · 1 region · 1 in stock
1 of 1 configs reported in stock at last check.
from $0.208/hr
AI Galaxy · 1 config · 1 region
Published catalog rates.
from $0.247/hr
IO.NET · 1 config
Published catalog rates.
from $0.250/hr
Theta EdgeCloud · 1 config
Published catalog rates.
from $0.260/hr
Featurize · 1 config · 1 region
Published catalog rates.
from $0.280/hr
4090Xianka · 3 configs · 3 regions
Published catalog rates.
from $0.295/hr
AutoDL · 1 config · 1 region
Published catalog rates.
from $0.296/hr
BUYING RTX 4090: WHAT THE PANEL SAYS

At the current floor, one RTX 4090 running around the clock costs about $117 a month (730 hours of on-demand arithmetic). Node shape matters: the single-GPU floor is $0.160/GPU/hr against $0.616/GPU/hr on an 8-GPU node (+285.2%). Per-GPU floors by config size: 1x $0.16 · 2x $0.46 · 4x $0.51 · 8x $0.62. Compare the alternatives below before committing - the same budget often buys more than one chip class.

RTX 4090 VS THE ALTERNATIVES

RTX 4090 vs L40S
Similar tensor throughput; L40S adds 48 GB ECC, datacenter drivers and support - at 2-4x the hourly rate. If 24 GB fits your job, the 4090 is the value pick.
RTX 4090 vs RTX 5090
The 5090 adds 8 GB of much faster GDDR7 and Blackwell tensor cores. Where it rents near 4090 prices, take it; it usually does not yet.
RTX 4090 vs RTX 3090
Same 24 GB, roughly half the FP16 throughput, but with NVLink pairs - the only consumer option for linking two cards. 4090 is faster per card; 3090 scales to 48 GB per pair.

Compare live rates: RTX 4090 vs L40S · RTX 4090 vs RTX 5090 · RTX 4090 vs RTX 3090 · RTX 4090 vs RTX 5080 · RTX 4090 vs RTX 5070 Ti · RTX 4090 vs RTX 4090D · RTX 4090 vs RTX 4080 · RTX 4090 vs RTX 3090 Ti · RTX 4090 vs RTX 4080 Super · RTX 4090 vs RTX 6000 Ada

RTX 4090 PRICING FAQ

What is the cheapest RTX 4090 cloud GPU right now?
The cheapest published RTX 4090 rate on the panel is $0.160 per GPU-hour at Salad (RTX 4090 x 1, Global (distributed)). Every price links to the provider's own published rate.
How much does a RTX 4090 cost per hour?
Across 39 providers tracked today, published per-GPU hourly rates for RTX 4090 run from $0.16 to $3.50, with a median of $0.52. Committed-use and reserved rates are excluded; everything shown is on-demand.
How many providers rent RTX 4090 GPUs?
39 providers publish 76 distinct RTX 4090 configurations on Compute Cafe. 32 of those rows were reported in stock by the provider's live endpoint at last observation.
Where is RTX 4090 cheapest?
By published per-GPU hourly rate, the cheapest geography for RTX 4090 today is china at $0.25/hr (AI Galaxy). The spread across 5 tracked geographies runs to $2.72/hr in vietnam.
Are RTX 4090 prices going up or down?
Over the 53 days Compute Cafe has tracked the panel, the RTX 4090 floor is holding flat: $0.160 to $0.160/GPU/hr (+0.0%). The median continuously-tracked listing moved from $0.34 to $0.34 (+0.0%). Daily snapshots are public in the repo, so the trend is auditable.
What does a RTX 4090 cost per month?
At the current floor of $0.160/GPU/hr, a RTX 4090 running 24/7 costs about $117 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 4090 node cheaper than a single card?
On today's panel: the single-GPU floor is $0.160/GPU/hr and the 8-GPU node floor is $0.616/GPU/hr - more expensive per GPU on a full node (+285.2%). Check the table for the exact configs behind each floor.
Is the RTX 4090 good for LLMs?
For models up to ~13B quantized, it is the best price-performance on most of the panel. Its 24 GB and lack of fast multi-GPU interconnect cap it there - bigger models need datacenter parts or multi-card tricks.
Why is the 4090 so much cheaper than an L40S?
Consumer licensing, no ECC, no vGPU, no support contract - the L40S price buys datacenter features, not more compute. For most independent workloads those features do not matter.

See also alternatives to the RTX 4090 · cheapest GPU-hour per model · inference chips · training chips · by location