Paid links may earn Compute Screener a referral fee. They never affect prices, weights or the order.
About NVIDIA RTX PRO 6000
The NVIDIA RTX PRO 6000 rental price today is $2.02 per GPU-hour, unchanged since yesterday, up 4.2% over the past week and up 7.7% over the past month. At that rate one RTX PRO 6000 costs about $1,480 for a month of 730 hours. Over the past year it has ranged from $1.36 to $2.02. The index is the price of one benchmark contract: RTX PRO 6000 Blackwell 96GB, 600 W, Server or Workstation Edition, on-demand, 8-GPU node, US hardware, CPU-neutral. It is standardized across 9 cloud providers and marketplaces, from Vast.ai at $1.61 to CoreWeave at $2.50, and is updated every 2 to 3 hours.
The RTX PRO 6000 Blackwell is NVIDIA's professional GPU for workstations and standard servers, announced on March 18, 2025 in Workstation, Max-Q and Server editions. Unlike the B200 it is a PCIe card with 96 GB of GDDR7 memory rather than HBM, and it keeps graphics and video hardware, so it serves both AI inference and rendering, simulation and virtual workstations. It is the cheapest Blackwell GPU on this site to rent and one of the most widely listed.
Specifications
| Spec | RTX PRO 6000 Blackwell Server Edition |
|---|
| Architecture | Blackwell, 24,064 CUDA cores, 188 RT cores |
| Memory | 96 GB GDDR7 with ECC, 1,597 GB/s |
| FP4 / FP8 / BF16 tensor | 4 / 2 / 1 PFLOPS; NVIDIA does not say whether dense or sparse |
| Interface | PCIe Gen 5, no NVLink |
| Power | up to 600 W, configurable: Workstation Edition 600 W, Max-Q 300 W |
| Multi-instance GPU | up to 4 instances |
| Typical server | 2U or 4U servers with 2, 4 or 8 cards |
Source: NVIDIA RTX PRO 6000 Blackwell Server Edition, accessed October 24, 2026; edition wattages from Puget Systems and ServeTheHome.
Where the RTX PRO 6000 is rented
NVIDIA said 8-GPU RTX PRO servers were available from major server makers in August 2025. Google Cloud's G4 machines became generally available on October 20, 2025 and AWS G7e on January 20, 2026, and Oracle, Azure, CoreWeave, Nebius and RunPod also rent it. GetDeploying tracked 53 providers on October 23, 2026, close to the H100's 57.
What this price measures
The index is the posted on-demand price of one benchmark contract, an RTX PRO 6000 Blackwell 96GB Server or Workstation Edition at 600 W in an 8-GPU node of US hardware, rented by the hour. Prices for other node sizes are converted to that contract, and prices from other regions are counted without a regional adjustment for now. Prices for the 300 W Max-Q edition are left out until enough providers price both editions to estimate an adjustment; none was in use on October 23, 2026. Reserved and spot prices are not counted.
On October 23, 2026, GetDeploying listed it at $1.80 per GPU-hour at Nebius, $2.49 at RunPod, $2.50 at CoreWeave for 8 GPUs, $3.36 at AWS, $4.50 at Google Cloud and $5.50 at Azure for single GPUs, and $4.50 at Oracle for 8 GPUs. The median was $2.38 on demand and $1.41 spot.
What moves RTX PRO 6000 rental prices
- Inference that fits on one card. With 96 GB, one card can serve models up to about 146 billion parameters at 4-bit precision with a 32K context, GetDeploying estimates. Without NVLink, it is not used to train large models, so its demand follows inference and media work rather than training budgets.
- Price per performance against data-center GPUs. NVIDIA claims up to 3 times the price-performance of an H100 for one model, run in 4-bit NVFP4 on the RTX PRO 6000 against FP8 on the H100, and up to 6 times the inference speed of the L40S.
- Supply from standard servers. It runs in air-cooled PCIe servers in ordinary data centers, so new capacity is less limited by power and cooling than 8-GPU Blackwell HGX nodes, which draw about 14 kW each, or liquid-cooled rack systems such as the GB200 NVL72.
- Shortages of larger GPUs. When H100 and B200 capacity is scarce, as in the 2026 GPU shortage, some inference moves to cards like this one.
Price history
The median posted on-demand price was $2.38 per GPU-hour on October 23, 2026. GetDeploying's weekly provider median rose about 14% over the 13 weeks to October 19, and a year earlier that series had only 5 providers, against 42 now. This index's history before October 23, 2026 is GetDeploying's weekly RTX PRO 6000 series, rescaled to join the daily index.
Futures and benchmarks
No futures contract or Bloomberg ticker covers the RTX PRO 6000. The GPU futures announced by ICE and Ornn in May 2026 name the consumer RTX 5090, a different card.
FAQ
How does it compare with the H100 on price? On GetDeploying's October 23, 2026 medians it rented for about 62% of an H100's hourly price, $2.38 against $3.85. It has more memory, 96 GB against 80 GB, but a much lower memory bandwidth of 1.6 TB/s against 3.35 TB/s, and no NVLink, according to NVIDIA.
Workstation, Max-Q or Server Edition? All three have 96 GB. The Workstation and Server editions run at up to 600 W; the Max-Q runs at 300 W and is slower, according to NVIDIA, Puget Systems and ServeTheHome.
How often is this index updated? Every 2 to 3 hours, with one value per day in US Central time that is final when the day closes.
Sources
Text reviewed Oct 24, 2026; the figures in it are as of that date. The figures at the top of this page update every 2 to 3 hours. Not investment advice.
