
The NVIDIA RTX PRO 6000 Blackwell costs $16,000 to buy and $1.70-$5.50/GPU-hr to rent, which makes rental the practical choice for most projects. The RTX PRO 6000 is NVIDIA's most powerful workstation graphics card. It packs 96GB of GDDR7 VRAM and 5th-generation Tensor Cores for 3D rendering, simulation, and local AI development.
Takeaways
- The purchase price starts at $16,000, up 87% from the $8,565 launch price.
- Cloud rental costs $1.70-$5.50/GPU-hr on-demand, with a $3.03 median across 9 providers.
- Specialist clouds cost about a third of Azure, the most expensive hyperscaler for the RTX PRO 6000.
- Buying breaks even after 7-13 months of 24/7 use, before power and host hardware costs.
- 96GB of VRAM and native FP4 suit single-GPU inference, but the missing NVLink limits multi-GPU training.
On-demand GPUs
Starting at $0.35/hr, billed per minute.
NVIDIA RTX PRO 6000 Blackwell Price
NVIDIA's official marketplace lists the card at $16,000 as of November 2026, an 87% increase since being launched in March 2025 at a retail price of $8,565.
Third-party retailers charge $16,399-$19,999 for a new RTX PRO 6000:
- NVIDIA Marketplace: $16,000
- Newegg: From $16,399
- B&H Photo: $19,999 (Workstation Edition)
- Refurbished/Used: $17,464–$30,895 (limited availability)
Why Is the RTX PRO 6000 Price Rising?
The memory shortage is the main cause of the RTX PRO 6000 price increase. The card uses 96GB of GDDR7, making it sensitive to limited supply in the memory market.
There is no near-term relief; with no direct workstation competitor at this memory capacity (AMD's Radeon PRO line tops out at 48GB), and AI-driven demand likely to keep rising.
NoteCloud rental rates have stayed comparatively stable. Renting reduces exposure to hardware price volatility and depreciation.
How Much Does it Cost to Rent?
Renting an RTX PRO 6000 costs $1.70-$5.50/GPU-hr on-demand, with a $3.03 median across 9 cloud providers. Below are current on-demand rates by provider:
| Provider | GPU / Instance | On-Demand $/GPU-hr | Notes |
|---|---|---|---|
| Vast.ai | RTX PRO 6000 Blackwell | $1.70 | Median of 21 verified US/CA hosts; ranges $1.48-$6.69/GPU-hr. Priced as compute plus 100GB storage, with bandwidth billed separately. |
| Nebius | RTX PRO 6000 Blackwell | $1.80 | |
| CoreWeave | 8 x RTX PRO 6000 Blackwell | $2.50 | Normalized from an 8-GPU node. |
| Runpod | RTX PRO 6000 Blackwell | $2.69 | Raised from $2.49 since Nov 9, 2026. |
| Modal | RTX PRO 6000 Blackwell | $3.03 | |
| AWS | g7e.2xlarge | $3.36 | Cheapest US region (us-east-1); ranges $3.36-$5.27/GPU-hr across US regions and sizes. |
| Google Cloud | g4-standard-48 | $4.50 | Cheapest US region (us-central1); ranges $4.50-$4.95/GPU-hr across US regions. |
| Oracle Cloud | BM.GPU.RTXPro.8 | $4.50 | Normalized from an 8-GPU bare-metal shape. |
| Azure | Standard_NC288ds_xl_RTXPRO6000BSE_v6 | $5.50 | Normalized from a 2-GPU VM in the cheapest US region (centralus); ranges $5.50-$7.15/GPU-hr across US regions. |
| Last reviewed on November 13, 2026. | |||
RTX PRO 6000 Price Trend
The mean on-demand RTX PRO 6000 price across 9 providers rose 2.4%, from $3.19/GPU-hr on Aug 5 to $3.27 on Nov 9, 2026.
Recent RTX PRO 6000 Price Changes
| Date | Provider | On-demand $/GPU-hr | Change |
|---|---|---|---|
| Oct 9, 2026 | Runpod | $2.09 → $2.49 | +19% |
| On-demand list-price changes from Aug 5 to Nov 9, 2026. | |||
AWS G7e Instances
AWS offers NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs through its EC2 G7e family, launched in January 2026.
AWS G7e instances include:
- NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs (96GB GDDR7 VRAM per GPU)
- 8–192 vCPUs
- 64–2,048GB RAM
- Up to 15.2 TB local NVMe SSD storage
AWS bills per-second (60-second minimum). Egress fees apply at $0.09/GB after the first free 100GB per month.
| SKU | RTX PRO 6000 GPUs | vCPUs | RAM (GB) | Hourly price | Price per GPU |
|---|---|---|---|---|---|
| g7e.2xlarge | 1 | 8 | 64 | $3.36 | |
| g7e.4xlarge | 1 | 16 | 128 | $4.00 | |
| g7e.8xlarge | 1 | 32 | 256 | $5.27 | |
| g7e.12xlarge | 2 | 48 | 512 | $8.29 | $4.14 |
| g7e.24xlarge | 4 | 96 | 1,024 | $16.57 | $4.14 |
| g7e.48xlarge | 8 | 192 | 2,048 | $33.14 | $4.14 |
Read nextEC2 GPU Instances – A full breakdown of G-series and P-series pricing, spot discounts, and cost optimization strategies.
Google Cloud G4 Instances
GCP's accelerator-optimized G4 machine family includes:
- NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs
- 5th Gen AMD EPYC (Turin)
- 3.75GB of RAM per vCPU on all
g4-standard-shapes - One RTX PRO 6000 GPU per 48 vCPUs, scaling to 8 GPUs
| Provider | SKU | RTX 6000 GPUs | VCPUs | RAM | Hourly price | Price per-GPU |
|---|---|---|---|---|---|---|
| GCP | g4-standard-6 | 0.125 | 6 | 22 | $0.56 | $4.50 |
| GCP | g4-standard-12 | 0.25 | 12 | 45 | $1.12 | $4.50 |
| GCP | g4-standard-24 | 0.5 | 24 | 90 | $2.25 | $4.50 |
| GCP | g4-standard-48 | 1 | 48 | 180 | $4.50 | |
| GCP | g4-standard-96 | 2 | 96 | 360 | $9.00 | $4.50 |
| GCP | g4-standard-192 | 4 | 192 | 720 | $18.00 | $4.50 |
| GCP | g4-standard-384 | 8 | 384 | 1440 | $36.00 | $4.50 |
Read nextGoogle Cloud GPU Instances – A comparison of A4, A3, A2, G2, and N1 series with pricing tables and cost tips.
Azure NC RTX PRO 6000 BSE v6 Instances
Microsoft Azure offers the RTX PRO 6000 Blackwell Server Edition through its NC RTX PRO 6000 BSE v6 series (NCv6). NCv6 VMs run on Intel Xeon Granite Rapids CPUs (up to 4.2 GHz all-core turbo).
Azure NCv6 instances include:
- NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs (96GB GDDR7 per GPU)
- Fractional GPU sizes (down to 1/4 GPU) via SR-IOV
- Up to 288 vCPUs
- Up to 1,032GB DDR5 RAM
- Not currently supported on AKS or Azure Batch
Azure NCv6 costs $5.50/GPU-hr at minimum, more than 3x the cheapest specialist clouds. Azure's cheapest single H100 (the NC40ads H100 v5) costs $6.98/hr, versus $3.20/hr on Thunder Compute. Azure also bills egress and storage separately from the compute rate.
| SKU | GPUs | vCPUs | RAM (GB) | On-Demand ($/hr)* |
|---|---|---|---|---|
| Standard_NC36ds_xl_RTXPRO6000BSE_v6 | 1/4 | 36 | 132 | $1.44 |
| Standard_NC72ds_xl_RTXPRO6000BSE_v6 | 1/2 | 72 | 264 | $2.83 |
| Standard_NC144ds_xl_RTXPRO6000BSE_v6 | 1 | 144 | 516 | $6.38 |
| Standard_NC288ds_xl_RTXPRO6000BSE_v6 | 2 | 288 | 1,032 | $11.00 |
| * On-demand (pay-as-you-go) rate for Linux in East US; pricing varies by region and OS. | ||||
Read nextAzure GPU Instances – A breakdown of on-demand, reserved, and spot pricing, with a cost comparison against AWS and GCP.
RTX PRO 6000 Workstation Edition vs Max-Q vs Server Edition
The three RTX PRO 6000 Blackwell variants differ in cooling design, power envelope, and form factor. While sharing the Blackwell architecture, 96GB of GDDR7 ECC memory, and the same core counts.
The RTX PRO 6000 Blackwell Workstation Edition is built for maximum performance in single-GPU workstations. Its dual-flow-through cooler sustains peak performance under a full 600W load, but a high heat output makes multi-card setups impractical in a standard office.
The RTX PRO 6000 Blackwell Max-Q suits multi-GPU workstation builds where thermal management is a concern. The Max-Q runs at a reduced 300W TDP, with a blower-style cooler that exhausts heat out the rear of the chassis. NVIDIA rates the Max-Q at 3,511 AI TOPS versus 4,000 on the Workstation Edition, roughly 12% lower.
The RTX PRO 6000 Blackwell Server Edition targets passive-cooled, rack-mounted data center deployments. The Server Edition is optimized for multi-GPU servers, enterprise inference, fine-tuning, HPC, and virtual workstations. Its memory bandwidth is slightly lower at 1.6 TB/s versus 1.8 TB/s on the workstation variants, a marginal gap for most AI workloads.
For AI performance, all three versions are close but not identical. The 600W Workstation and Server Editions deliver the most throughput, while the 300W Max-Q trades some performance for less power draw. Form factor and thermal environment usually decide the choice.
| Workstation Edition | Max-Q Edition | Server Edition | |
|---|---|---|---|
| Target environment | Single-GPU desktop workstations | Multi-GPU workstations | Data center racks |
| Cooling | Active dual-flow-through | Blower (rear exhaust) | Passive |
| TDP | 600W | 300W | 400–600W |
| Multi-GPU suitability | Limited (high heat output) | Good | Best |
| Memory bandwidth | 1.8 TB/s | 1.8 TB/s | 1.6 TB/s |
| Memory | 96GB GDDR7 ECC | ||
| Architecture | Blackwell | ||
Hardware Specifications and AI Capabilities
The RTX Pro 6000 is a generational leap over the RTX 6000 Ada and RTX A6000.
NVIDIA RTX PRO 6000 Blackwell VRAM
96GB of GDDR7 is the RTX PRO 6000's standout feature. A 70B-parameter model like Llama 3 70B fits on a single RTX PRO 6000 at 8-bit precision, with room for the KV cache. Memory bandwidth reaches 1,792GB/s, nearly double the RTX 6000 Ada's 960GB/s.
NVIDIA RTX PRO 6000 Blackwell CUDA Cores
The NVIDIA RTX PRO 6000 features 24,064 CUDA cores, delivering 125 TFLOPS of single-precision (FP32) compute.
NVIDIA RTX PRO 6000 Blackwell Tensor Cores
The NVIDIA RTX PRO 6000 has 752 tensor cores rated at up to 4,000 AI TOPS (FP4 with sparsity). The 5th-generation tensor cores add native FP4 precision support, which no earlier NVIDIA generation offers.
FP4 inference can roughly double throughput compared to FP8 by packing neural network weights at 4x the density of FP16. Inference frameworks like vLLM increasingly use FP4 and FP8 modes to maximize tokens/second with minimal quality loss for most models. having native FP4 support, the RTX PRO 6000 can outrun the H100 PCIe on FP4-quantized single-node inference.

RTX Pro 6000 Blackwell NVLink Support
The RTX Pro 6000 doesn't support NVLink. The RTX A6000 was the last NVIDIA workstation card with an NVLink bridge. NVIDIA dropped NVLink starting with the RTX 6000 Ada.
RTX PRO 6000 multi-GPU communication runs over the PCIe Gen 5 x16 bus. PCIe cannot match the low-latency GPU-to-GPU bandwidth of NVLink-equipped data-center GPUs. Workloads that need model parallelism across 4 or 8 GPUs will hit a bandwidth bottleneck.
RTX PRO 6000 vs RTX 6000 Ada vs RTX A6000
The RTX PRO 6000, RTX 6000 Ada, and RTX A6000 belong to different NVIDIA architectures: Blackwell, Ada Lovelace, and Ampere.
| GPU | Architecture | CUDA Cores | Tensor Cores | VRAM | Memory Type | Memory Bandwidth | TDP |
|---|---|---|---|---|---|---|---|
| RTX PRO 6000 | Blackwell | 24,064 | 752 (5th Gen) | 96GB | GDDR7 ECC | 1.8 TB/s | 600W |
| RTX 6000 Ada | Ada Lovelace | 18,176 | 568 (4th Gen) | 48GB | GDDR6 ECC | 960GB/s | 300W |
| RTX A6000 | Ampere | 10,752 | 336 (3rd Gen) | 48GB | GDDR6 ECC | 768GB/s | 300W |
The Best Alternative for AI: NVIDIA H100
The NVIDIA H100 outperforms the RTX Pro 6000 Blackwell for large-scale AI training, even though the RTX PRO 6000 is the top workstation GPU available.
Why the H100 Wins for AI
- Memory Bandwidth: The H100 delivers more memory bandwidth than the RTX PRO 6000: 2.0 TB/s (PCIe) and 3.35 TB/s (SXM) vs. 1.8 TB/s. Higher bandwidth speeds up memory-bound training and large-batch inference.
- True NVLink Scaling: The H100 SXM supports the NVLink Switch System, enabling up to 256 GPUs to communicate at 900GB/s, the key advantage for large-scale distributed training.
- Price: An RTX Pro 6000 now lists at ~$16,000, while a new H100 PCIe costs $24,000 to $45,000 through authorized resellers. Rental nearly erases the gap: Thunder Compute offers on-demand H100s at $3.20/hr, close to the $3.03 RTX PRO 6000 median, with no hardware depreciation risk.
Comparison Snapshot: RTX Pro 6000 Blackwell vs. H100
| Feature | RTX Pro 6000 Blackwell | NVIDIA H100 (PCIe) |
|---|---|---|
| Architecture | Blackwell | Hopper |
| VRAM | 96GB GDDR7 | 80GB HBM2e |
| Memory Bandwidth | 1.8 TB/s | 2.0 TB/s |
| Interconnect | PCIe Gen 5 | 600GB/s NVLink |
| Target Workload | Design & Development | Foundation Model Training |
NVIDIA RTX PRO 6000 Blackwell Release Date
The NVIDIA RTX PRO 6000 Blackwell release date was March 18, 2025, at NVIDIA's GTC conference. It has since become the benchmark for high-end workstation workloads.
Last Thoughts on NVIDIA RTX PRO 6000 Pricing
The RTX Pro 6000 Blackwell is uncontested for professional visualization, but its $16,000 price, 600W draw, and GDDR7 price pressure make it a poor buy for most AI teams. Renting avoids hardware maintenance, depreciation, and price volatility. Thunder Compute provides instant access to NVIDIA H100s at $3.20/hr, with no long-term commitments.
Read nextAI GPU rental market trends - See how RTX Pro 6000 costs fit broader industry shifts.
FAQ
What is the NVIDIA RTX PRO 6000 Blackwell Price?
The RTX PRO 6000 Blackwell costs $16,000 on NVIDIA's marketplace, up 87% from its $8,565 launch price in March 2025. Newegg starts at $16,399 and B&H at $19,999. Used listings range from $17,464-$30,895.
How much VRAM does the RTX Pro 6000 Blackwell have?
The RTX Pro 6000 Blackwell has 96GB of GDDR7 ECC VRAM with 1,792GB/s of memory bandwidth.
How much does it cost to rent an RTX PRO 6000 in the cloud?
Renting an RTX PRO 6000 costs $1.70-$5.50/GPU-hr on-demand, with a $3.03 median across 9 providers (November 13, 2026). Vast.ai ($1.70), Nebius ($1.80), CoreWeave ($2.50), and Runpod ($2.69) are cheapest. AWS ($3.36), Google Cloud and Oracle ($4.50), and Azure ($5.50) charge more.
How much does it cost to run an RTX PRO 6000 for a month?
Running one RTX PRO 6000 24/7 (720 hours) costs about $1,220/month at $1.70/hr and $2,180/month at the $3.03 median. Buying the card at $16,000 breaks even after roughly 7-13 months of continuous use, before power, host hardware, and maintenance.
Are RTX PRO 6000 rental prices going up?
Yes, slightly. The mean on-demand price across 9 providers rose 2.4%, from $3.19 to $3.27/GPU-hr, between August 5 and November 9, 2026. Runpod made the only list-price change, from $2.09 to $2.49 on October 9.