NVIDIA has announced the next evolution of its Blackwell AI factory platform, known as Blackwell Ultra, designed to accelerate applications such as AI reasoning, agentic AI, and physical AI. Built on the groundbreaking Blackwell architecture introduced last year, Blackwell Ultra includes the NVIDIA GB300 NVL72 rack-scale solution and the NVIDIA HGX B300 NVL16 system.
The new platform boasts 1.5x more AI performance than its predecessor and increases revenue opportunities for AI factories by 50x compared to those built with NVIDIA Hopper. According to Jensen Huang, founder and CEO of NVIDIA, “AI has made a giant leap — reasoning and agentic AI demand orders of magnitude more computing performance.”
Blackwell Ultra enables AI models to access the platform’s increased compute capacity, allowing them to explore different solutions to problems and break down complex requests into multiple steps. The system is also expected to be available on NVIDIA DGX Cloud, an end-to-end, fully managed AI platform.
The new platform features 11x faster inference on large language models, 7x more compute, and 4x larger memory compared with the Hopper generation. It’s ideal for applications such as agentic AI, which uses sophisticated reasoning and iterative planning to autonomously solve complex problems, and physical AI, which enables companies to generate synthetic videos in real time.
Advanced scale-out networking is a critical component of Blackwell Ultra, delivering top performance while reducing latency and jitter. The system seamlessly integrates with NVIDIA Spectrum-X Ethernet and NVIDIA Quantum-X800 InfiniBand platforms, as well as NVIDIA BlueField-3 DPUs, which enable multi-tenant networking, GPU compute elasticity, accelerated data access, and real-time cybersecurity threat detection.
Blackwell Ultra-based products are expected to be available from partners starting in the second half of 2025. Cloud service providers such as Amazon Web Services, Google Cloud, Microsoft Azure, and Oracle Cloud Infrastructure will offer Blackwell Ultra-powered instances, along with GPU cloud providers like CoreWeave, Crusoe, Lambda, Nebius, Nscale, Yotta, and YTL.
The entire NVIDIA Blackwell product portfolio is supported by the full-stack NVIDIA AI platform. The new AI inference-serving software, NVIDIA Dynamo, scales up reasoning AI services, delivering leaps in throughput while reducing response times and model serving costs.
Source: https://nvidianews.nvidia.com/news/nvidia-blackwell-ultra-ai-factory-platform-paves-way-for-age-of-ai-reasoning