Presented by NVIDIA
The AI revolution accelerates the parameters of reasoning models required by billions needed to develop agency and physical artificial intelligence. Like the founder and general director of NVIDIA, Jensen Huang in his GTC speech, the transition from training to full production means that the demand for AI increases growth when data centers around the world transform into AI factories designed to effectively and effectively pay millions of user questions. To achieve this opportunity worth $ 1 trillion, NVIDIA in GTC presented significant progress – from the Blackwell Ultra AI platform and the operating system for AI factories to progress in the creation of networks, robotics and accelerated calculations.
Blackwell is already in full production – it provides an amazing 40x enhance in efficiency over Hopper. This architecture again defines the training and inference of the AI model, thanks to which AI applications are more competent and more scalable. And the upcoming in the second half of 2025 is another evolution of the Blackwell AI factory platform: Blackwell Ultra-Patty graphics processor with extended memory to support the next generation of AI models.
Nvidia still moves quickly, undertakes to refresh the architecture of AI annual. Nvidia Vera Rubin has been designed to supplement the performance and performance of the AI data center.
In addition to GPU, the AI infrastructure undergoes a seismic change with innovations in photonics, optimized by the treatment remembering and advanced contacts. These breaks will significantly enhance scalability, efficiency and energy consumption in massive AI data centers.
Meanwhile, physical artificial intelligence for robotics and industry is a colossal opportunity of $ 50 trillion, according to Huang. From production and logistics to health care and not only, automation driven by artificial intelligence is ready to transform entire industries. The Nvidia Isaac and Cosmos platforms are on the first platform, driving another era of robotics powered by AI.
Some of the NVIDIA ads in GTC
NVIDIA road map: The NVIDIA road map includes Vera Rubin, Tobe released in the second half of 2026, followed by the introduction of Vera Rubin Ultra in 2027. Chips and Rubin servers offer better speeds, especially in data transfers between systems, which is a key function for vast AI systems with many systems. And planned for 2028 is FeynmanThe next architecture that will be released using the novel generation HBM memory.
DGX Personal computers AI: Powered by the NVIDIA Grace Blackwell, DGX Spark and DGX Station platform are designed to develop, adapt and inference vast models on desktop computers. They will be produced by many companies, including ASUS, Dell and HP.
Spectrum-X and Quantum-X Networking Platforms: These silicone photonic network switches lend a hand AI factories to combine millions of GPU in various places and dramatically reduce energy consumption. The Infiniband Quantum-X Photonics switches will be available at the end of this year, and Ethernet Photonnet Photonics will appear in 2026.
Dynamo software: Free Dynamo Open-Source software has been released to accelerate the process of multi-stage reasoning, improve performance and shortening time to innovation in AI factories.
Nvidia accelerated quantum research center: The Research Center based in Boston will provide the latest technologies to enhance quantum calculations in cooperation with leading hardware and software manufacturers.
Nvidia isaac gr00t n1: The basic model of humanoid robots, the GR00T N1 is the world’s first open, fully configurable model of the foundation of generalized reasoning and humanoid skills. It has a double system similar to reasoning models, both for quick and tardy thinking.
Newton physics engine: NVIDIA also announced cooperation with Google Deepmind and Disney Research to develop Newton, an open source physics engine, which allows robots to learn to handle intricate tasks with greater precision.
This is just the most essential information – Do not miss the full GTC summary, live on the NVIDIA blog.
The articles are sponsored with the content produced by a company that either pays for the position or has a business relationship with Venturebeat and is always clearly marked. For more information, please contact