NVIDIA has officially confirmed the general availability of a new 72GB configuration for its RTX PRO 5000 Blackwell GPU, expanding the memory options for professional workstations focused on AI development, simulation, and high-end content creation.
This confirmation follows updates to NVIDIA’s official product listings and a company blog announcement, which now show the RTX PRO 5000 shipping in both 48GB and 72GB GDDR7 ECC variants. Initial availability is limited to select distribution partners, with broader system availability expected to begin in early 2026.
Availability Confirmed Through NVIDIA’s Distribution Partners
According to NVIDIA, the RTX PRO 5000 72GB Blackwell GPU is now available through partners such as Ingram Micro, Leadtek, Unisplendour, and xFusion. Wider availability through global system integrators and OEM workstation builds is expected early next year.
NVIDIA has not yet disclosed the official price of the 72GB model. Based on current market listings, the 48GB RTX PRO 5000 is significantly less expensive than the RTX PRO 6000 Blackwell card, suggesting the final price of the 72GB variant will fall somewhere between the two. Currently, NVIDIA has not clarified how the new model’s price will compare to existing workstation options.
Same GPU, Higher Memory Capacity
NVIDIA’s official specifications make it clear that the 72GB RTX PRO 5000 is not a new GPU design, but a higher-memory configuration of the existing RTX PRO 5000 Blackwell.
Both the 48GB and 72GB models share the same core specifications, including:
- 14,080 CUDA cores
- Blackwell architecture
- 300W total board power
The primary difference is the increase to 72GB of GDDR7 memory with ECC, representing a 50 percent capacity increase over the original 48GB configuration. NVIDIA positions this upgrade for professional workloads where GPU memory capacity, rather than raw compute throughput, becomes the primary constraint.
Memory Interface Details Remain Unclear
A technical detail in NVIDIA’s official documentation remains unresolved. The current datasheet specifies a 512-bit memory interface, along with a memory bandwidth of 1,344 GB/s.
This bandwidth figure perfectly matches that of 28 Gbps GDDR7 memory running on a 384-bit bus, consistent with figures reported in third-party analyses. A true 512-bit interface at the same bandwidth would imply significantly lower memory speeds.
NVIDIA has not yet clarified this discrepancy, and no updated documentation has been released to reconcile the stated interface width with the published bandwidth figures.
Designed for Memory-Bound AI and Professional Workloads
NVIDIA positions the RTX PRO 5000 72GB as a workstation-class solution aimed at memory-intensive local workloads rather than maximum compute density.
According to the company, the expanded memory capacity enables:
- Larger local AI models and extended context windows
- Agentic and multimodal AI workflows
- Complex 3D scenes and large simulation datasets
- Reduced reliance on cloud or data-center infrastructure
The additional VRAM allows developers and professionals to keep more models, assets, and datasets resident in GPU memory at the same time, which can be critical for workloads that scale poorly when data must be frequently swapped or streamed.
By increasing VRAM capacity without changing the underlying GPU or its 300W power envelope, NVIDIA is targeting professionals who need additional memory headroom rather than higher core counts. This positioning places the 72GB model between the existing 48GB RTX PRO 5000 and RTX PRO 6000 Blackwell cards, which carry higher power and system requirements.
Performance and Feature Set
NVIDIA’s published specifications list up to 2,064 AI TOPS for the RTX PRO 5000 Blackwell, supported by fifth-generation Tensor Cores, fourth-generation ray tracing cores, and ninth-generation NVENC along with sixth-generation NVDEC engines.
The GPU supports PCIe Gen 5, DisplayPort 2.1, and Multi-Instance GPU (MIG) functionality, allowing the card to be partitioned into multiple isolated instances for parallel workloads or multi-user environments.
While NVIDIA claims significant performance gains over prior-generation workstation GPUs in AI inference, rendering, and engineering tasks, independent benchmarks have not yet been published. Real-world performance will likely vary depending on whether workloads are compute-bound or memory-bound.
Pricing Still Undisclosed
Despite confirming availability, NVIDIA has not announced an official MSRP for the RTX PRO 5000 72GB. Existing pricing data suggests a wide gap between 48GB RTX PRO 5000 cards and RTX PRO 6000 Blackwell models, but where the 72GB version will land remains unclear.
Until official pricing or OEM configurations are announced, workstation buyers will need to weigh memory requirements against system cost, cooling, and power constraints.
What Remains Unknown
Several key details remain unconfirmed, including:
- Final MSRP or partner pricing
- Definitive confirmation of memory interface width
- Regional rollout timelines beyond early 2026 system builds
Further clarification is expected as more OEM workstations ship and NVIDIA updates its technical documentation.
Positioning Within NVIDIA’s Workstation Lineup
The RTX PRO 5000 72GB Blackwell is a notable addition to NVIDIA’s professional GPU lineup, focusing on increasing memory capacity without altering the underlying architecture or power envelope. This new configuration sits between the existing 48GB RTX PRO 5000 and the higher-tier RTX PRO 6000 Blackwell models in terms of memory capacity and system requirements.
More details are expected as system manufacturers begin shipping workstation configurations and NVIDIA updates its pricing and platform documentation.
Source: NVIDIA Blog
