NVIDIA is set to bring its latest Blackwell RTX PRO 6000 Server Edition GPUs to enterprise data centers, with systems from Cisco, Dell Technologies, HPE, Lenovo, and Supermicro arriving in various 2U server configurations.
Designed for high-performance AI, simulation, industrial analytics, and advanced graphics, the new GPUs promise significant gains in both speed and energy efficiency. According to NVIDIA, the Server Edition can deliver up to 45 times the performance of CPU-only systems while using 18 times less energy — a leap aimed at meeting the growing demands of AI-powered business operations.
“AI is reinventing computing for the first time in 60 years,” said Jensen Huang, NVIDIA’s founder and CEO. “With the world’s leading server providers, we’re making NVIDIA Blackwell RTX PRO Servers the standard platform for enterprise and industrial AI.”
Eric Revell
A Push for GPU-Accelerated Data Centers
Most of today’s business servers still rely on traditional CPUs, but NVIDIA’s new RTX PRO line is designed to accelerate analytics, rendering, video processing, and complex simulations. The architecture also targets “AI factory” setups where space, power, and cooling are at a premium.
Dell’s updated PowerEdge R7725 servers, for example, will come equipped with two RTX PRO 6000 GPUs, NVIDIA AI Enterprise software, and NVIDIA networking. The new 2U units can house up to eight GPUs and were first previewed at COMPUTEX in May.
Blackwell Architecture Highlights
Built on NVIDIA’s Blackwell architecture, the RTX PRO 6000 brings:
- Fifth-gen Tensor Cores and a second-gen Transformer Engine with FP4 precision, running inference up to six times faster than the L40S GPU.
- Fourth-gen RTX technology for photo rendering with up to four times the performance of the L40S GPU.
- Multi-Instance GPU (MIG) support for running four separate workloads per GPU.
- Enhanced energy efficiency to reduce data center power consumption.
Expanding AI and Robotics Capabilities
The RTX PRO Servers are optimized for NVIDIA’s Omniverse and Cosmos platforms, enabling digital twin simulations, robot training, and large-scale synthetic data generation. Updates include new Omniverse SDKs, compatibility with MuJoCo and OpenUSD, and the NuRec 3D rendering library, now used by companies like Foretellix and Voxel51 for autonomous vehicle and sensor modeling.
Cosmos updates introduce the Transfer-2 model for faster synthetic image generation and Cosmos Reason — a 7-billion-parameter vision language model for advanced robot planning, physics-aware reasoning, and automated dataset curation. Early adopters include Boston Dynamics, Amazon Devices & Services, Uber, and Magna.
Enterprise AI at Scale
The servers also support the newly launched Llama Nemotron Super model. Running with NVFP4 precision on a single RTX PRO 6000 GPU, it delivers up to three times better price performance than FP8 on NVIDIA’s H100 GPUs.
By combining cutting-edge GPU hardware with an expanding suite of AI software tools, NVIDIA is positioning the Blackwell RTX PRO 6000 as a cornerstone for enterprises building the next generation of AI-driven applications — from autonomous vehicles and robotics to industrial analytics and digital twins.
