About NVIDIA RTX 4090
The NVIDIA RTX 4090 rental price today is $0.55 per GPU-hour, down 0.4% since yesterday, down 0.4% over the past week and down 2.9% over the past month. At that rate one RTX 4090 costs about $400 for a month of 730 hours. Over the past year it has ranged from $0.50 to $0.71. The index is the price of one benchmark contract: GeForce RTX 4090 24GB, on-demand, 8-GPU node, US hardware, CPU-neutral. It is standardized across 6 cloud providers and marketplaces, from Akash at $0.34 to RunPod at $0.62, and is updated every minute.
The GeForce RTX 4090 is the flagship of NVIDIA's consumer GeForce RTX 40 Series, announced on September 20, 2022 for sale from October 12, starting at $1,599. It is an Ada Lovelace PCIe card with 24 GB of GDDR6X memory and no NVLink, on the same architecture as the data-center L40S. Many GPU clouds and marketplaces rent it: on October 30, 2026 GetDeploying tracked 19 RTX 4090 providers, and of the consumer GeForce cards it lists, only the RTX 5090 had more, 20.
Specifications
| Spec | RTX 4090 |
|---|
| Architecture | Ada Lovelace AD102, TSMC 4N, 76.3 billion transistors |
| Cores | 16,384 CUDA, 512 fourth-generation Tensor, 128 third-generation RT |
| Memory | 24 GB GDDR6X, 384-bit, 1,008 GB/s |
| FP8 tensor | 330.3 TFLOPS dense, 660.6 with sparsity, with FP32 accumulate; twice that with FP16 accumulate |
| BF16 / FP16 tensor | 165.2 TFLOPS dense, 330.4 with sparsity, with FP32 accumulate; FP16 twice that with FP16 accumulate |
| FP32 | 82.6 TFLOPS |
| Interface | PCIe Gen 4; no NVLink |
| Power | 450 W total graphics power; Founders Edition 3 slots, 304 mm long |
| Video | 2 eighth-generation encoders and 1 fifth-generation decoder, with AV1 |
Source: NVIDIA GeForce RTX 4090, accessed October 30, 2026; chip, core, tensor, memory bandwidth and video figures from NVIDIA's Ada whitepaper.
Where the RTX 4090 is rented
GetDeploying tracked 19 RTX 4090 providers on October 30, 2026, among them RunPod, Novita, Salad, Lium, TensorDock and Oblivus, and its on-demand median covered 14; none was AWS, Azure, Google Cloud or Oracle. Seven of them listed machines of up to 8 GPUs, and ten listed single GPUs only, among 118 listings in all.
What this price measures
The index is the posted on-demand price of one benchmark contract, an RTX 4090 24GB in an 8-GPU node of US hardware, rented by the hour from GPU clouds and marketplaces. Prices for other node sizes are converted to that contract with the index's node-size factors, and prices from other regions are counted without a regional adjustment for now. Reserved and spot prices are not counted, and RTX 4090 cards modified to 48 GB are a different product and are not counted.
On October 30, 2026, GetDeploying listed 8-GPU RTX 4090 machines on demand at $0.44 per GPU-hour at Sesterce, out of stock, and $2.22 at Daytona, in stock; GPUhub's 8-GPU machines, at $0.46, had 48 GB per card. Single GPUs started at $0.33 in stock at Novita and Salad, against $0.89 on RunPod's Secure Cloud. Its medians were $0.47 on demand, $0.52 reserved and $0.51 spot.
What moves RTX 4090 rental prices
- Inference that fits on one card. One RTX 4090 can run models up to about 28 billion parameters at 4-bit precision, or 6 billion at 16-bit, with a 32K context, GetDeploying estimates. With no NVLink, cards in one machine exchange data over PCIe Gen 4.
- Newer cards. Its Blackwell successor, the RTX 5090, announced on January 6, 2025 for sale from January 30 at $1,999, has 32 GB of GDDR7 at 1,792 GB/s, about 27% more dense BF16 tensor throughput and FP4 support, according to NVIDIA's Blackwell whitepaper. Its GetDeploying median was $0.70 on October 30, 2026, against $0.47 for the RTX 4090.
- Export controls. NVIDIA listed the RTX 4090 on October 17, 2023 among the products whose exports to China and some other countries would need a US license, under a rule in force from November 17, 2023. For China it sells the RTX 4090 D, with 14,592 CUDA cores against 16,384, which it said was "designed to fully comply" with US export controls.
- Capacity. Where a source shows capacity, this index counts a provider only on days when one of its offers had capacity; its posted prices then all count, so an offer going out of stock does not read as a price change (methodology).
Price history
- October 2025 to October 2026: GetDeploying's weekly on-demand provider median fell from $0.52 in the week of October 27, 2025, when its public series starts, to $0.46 in the week of October 19, 2026, about 12% lower. It peaked at $0.59 in the weeks of December 15 and 22, 2025 and was lowest, $0.42, in the week of August 31, 2026. Its provider count rose from 9 to 12, with a peak of 17 in September 2026.
- October 30, 2026: the median posted on-demand price was $0.47 per GPU-hour across 14 providers, about 5% higher than 90 days earlier.
This index's history before October 30, 2026, its first day, is GetDeploying's weekly RTX 4090 series, rescaled to join the daily index, counting only weeks with at least 5 providers; every week had at least 9.
Futures and benchmarks
No futures contract or Bloomberg ticker covers the RTX 4090. CME's filed contracts and the contracts ICE says it will list, with no date set yet, cover the H100 and B200. The GPU futures ICE announced in May 2026 may cover the H100, H200, B200 and RTX 5090 and other GPUs "as the market develops", and the contracts Nodal Exchange plans are for "Hopper, Blackwell and beyond".
FAQ
How does it compare with the L40S? The L40S, NVIDIA's Ada data-center card, has twice the memory, 48 GB with ECC against 24 GB, and about 2.2 times the dense BF16 tensor throughput, 362 TFLOPS against 165.2, but 14% less memory bandwidth, 864 GB/s against 1,008 GB/s, according to NVIDIA's L40S page and Ada whitepaper. On GetDeploying's October 30, 2026 medians the RTX 4090 rented for about 30% of an L40S's hourly price, $0.47 against $1.57.
Why is a consumer card rented in clouds? It has the same generation of Tensor Cores, with FP8, as Ada data-center cards such as the L40S, at a lower hourly price (above). NVIDIA designed the RTX 40 Series "for gamers and creators", and its driver license, in the version of February 25, 2025, says GeForce software "is not licensed for datacenter deployment"; none of AWS, Azure, Google Cloud or Oracle was among the card's providers on GetDeploying (above).
How often is this index updated? It is recalculated every minute, with one value per day in US Central time that is final when the day ends.
Sources
- NVIDIA: GeForce RTX 4090 product page and specifications
- NVIDIA: GeForce RTX 40 Series announcement, September 20, 2022
- NVIDIA: Ada GPU architecture whitepaper
- NVIDIA: GeForce RTX 50 Series announcement, January 6, 2025
- NVIDIA: driver license agreement, version of February 25, 2025
- GetDeploying: NVIDIA RTX 4090 prices
- GetDeploying: GPU price dataset, weekly price history, CC BY 4.0
- NVIDIA: export licensing, Form 8-K, October 17, 2023
Text reviewed Oct 30, 2026; the figures in it are as of that date. The figures at the top of this page update every minute.
Cite as: Compute Screener, “NVIDIA RTX 4090 rental price index”, computescreener.com/gpu/rtx-4090. The figures are free to reuse under CC BY 4.0 with credit to Compute Screener.
