Nvidia messages, news and more, from GTC 2025
Presented by nvidia
The AI ​​Revolution accelerates the billion -parameter reasoning models needed to develop agent and physical AI. As NVIDIA’s founder and CEO Jensen Huang shared in his main GTC note, the transition from training to full production conclusion leads to the AI ​​Compute demand to jump to the jump as data centers worldwide, where they are transformed into AI factories designed to discard millions of consumer requests efficiently and efficiently. To respond to this opportunity for $ 1 trillion, NVIDIA in GTC has revealed great achievements – from the Blackwell Ultra AI platform and the AI ​​factories operating system to progress in network, robotics and accelerated calculations.
Blackwell is already in full production – provides an amazing increase of 40x performance over the bunker. This architecture redefines the training and the conclusion of AI model, which makes AI applications more efficient and more prominent. And the coming second half 2025 is the next evolution of the Blackwell AI factory platform: Blackwell Ultra-Powerhouse GPU with advanced memory to support the next generation AI models.
Nvidia continues to move fast, engaged in AI architecture annually. Nvidia Vera Rubin is designed to recharge the efficiency and efficiency of the AI ​​Data Center.
Beyond the graphic processors, AI infrastructure undergoes a seismic change in photone innovation, AI-optimized storage and advanced networks. These breakthroughs will dramatically increase the scale, efficiency and consumption of energy at the massive AI data centers.
Meanwhile, the physical AI for robotics and industry is a colossal opportunity for $ 50 trillion, according to Huang. From production and logistics to healthcare and outside, automation with AI power is ready to reshape all industries. Nvidia Isaac and Cosmos platforms are at the forefront, moving the next era of the AI-Move Robotics.
Some of the NVIDIA messages in GTC
NVIDIA ROAD CARD: NVIDIA road card includes Vera Rubin, Tobe launched in the second half of 2026, followed by the launch of Vera Rubin Ultra in 2027. Rubin chips and servers are proud of improved speeds, especially in data transfers between chips, which is a critical feature for large AI systems with many chips. And scheduled for 2028 is FeinmanThe next architecture that will be released using the next -generation HBM memory.
DGX Personal AI computers: Powered by the NVIDIA Grace Blackwell, DGX Spark and DGX Station platform are designed to develop, finely adjust large designs of desktops. They will be manufactured by a number of companies, including Asus, Dell and HP.
Spectrum-X and Quantum-X Network Platforms: These Silicon Photonics network switches help AI factories connect millions of graphics processors to sites and dramatically reduce energy consumption. Quantum-X Photonics Infiniband switches will be available later this year, and Spectrum-X Photonics Ethernet switches will arrive in 2026.
Dynamo software: Issued free of charge, open source Dynamo software helps to speed up the process of multi -stage reasoning, improve efficiency and reduce time to innovation in AI factories.
NVIDIA accelerates the quantum research center: The Boston -based research center will provide avant -garde technologies for the progress of quantum calculations in collaboration with leading hardware and software manufacturers.
Nvidia isaac GR00T N1: The main model for humanoid robots, the GR00T N1 is the world’s first, fully adaptive main model for generalized humanoid reasoning and skills. It has a dual system similar to models for reasoning, both fast and slow thinking.
Newton Physics Engine: Nvidia also announced collaboration with Google Deepmind and Disney Research to develop Newton, an open source physics engine that allows robots to learn how to handle complex tasks with greater accuracy.
These are just the highlights – Don’t miss the full gtc summary, live on NVIDIA blogS
Sponsored articles contain content from a company that either pays for the publication or has a business relationship with Venturebeat and they are always clearly marked. For more information, contact
English