NVIDIA Unveils Blackwell RTX Pro Servers and Advanced AI Capabilities for Enterprise Transformation NVIDIA Corporation has unveiled plans to expand its AI portfolio with next-generation intelligent models, robotics technology powered by physical reasoning, and a new line of enterprise-grade AI servers. At the ACM SIGGRAPH 2025 conference in Vancouver, NVIDIA introduced the RTX Pro 6000 Blackwell server GPU designed specifically for enterprise infrastructure. This 2U rack-mounted solution leverages the Blackwell architecture to deliver high-performance AI inference capabilities for large language models, enabling organizations to process complex workloads at accelerated speeds. "Artificial intelligence is redefining computing fundamentals for the first time in six decades - starting in the cloud and now transforming on-premises data center architectures," stated NVIDIA founder and CEO Jensen Huang during the announcement. The Blackwell RTX Pro servers integrate GPU acceleration into traditional CPU-centric workflows such as data analytics, simulation, video processing, and graphics rendering. NVIDIA claims these systems achieve 45x performance improvements over CPU-only solutions while delivering 18x better energy efficiency and significantly lower operational costs. Strategic partnerships with Cisco Systems, Dell Technologies, HP Enterprise, Lenovo Group, and Supermicro will ensure multi-configuration deployment options for global enterprises. "Through collaboration with leading server providers, we're establishing the NVIDIA Blackwell RTX Pro as the standard platform for industrial and enterprise AI," Huang added.

Smarter Agent AI Models

As autonomous AI agents become foundational infrastructure for modern organizations, NVIDIA is expanding its Nemotron model family with two new advanced reasoning capabilities: Nemotron Nano 2 and Llama Nemotron Super 1.5. These models deliver precision in complex scientific reasoning, code generation, tool utilization, instruction following, and conversational contexts while enhancing agent cognition through decision-making frameworks. Nemotron Nano 2 demonstrates 6x higher token generation throughput compared to similar models, while Llama Nemotron Super 1.5 establishes industry-leading accuracy for enterprise workloads. Enterprise deployments are already underway with partners including Zoom Video Communications integrating Nemotron into AI assistants, CrowdStrike testing models for cybersecurity query generation, and EY leveraging Nano 2 for business analytics automation.

Physical AI and Enhanced Reasoning Robotics

NVIDIA is advancing robotics through the Cosmos Reason framework - an open-source 7 billion parameter visual language model (VLM) that combines perception with physical reasoning capabilities. Unlike traditional VLMs that only enable computer vision, Cosmos Reason allows robots to process visual information and apply human-like reasoning using physical knowledge and common-sense logic. This breakthrough technology accelerates AI model development by automating dataset curation and annotation processes. The system functions as a sophisticated reasoning engine for robotic planning, translating complex instructions into actionable steps for vision-language-action (VLA) models even in novel environments. NVIDIA's physical AI initiatives are already transforming smart cities and industrial operations through partnerships with Accenture, Belden, DeepHow, Milestone Systems, and Telit Cinterion. The NVIDIA Metropolis platform further enables deployment of video analysis agents for campus security and facility management applications.