NVIDIA held its GTC 2025 keynote at the SAP Center in San Jose, CA, showcasing its evolution from a graphics hardware manufacturer to a leader in AI technology.

During the keynote, CEO Jensen Huang emphasized the company's confidence in its vision, highlighting the AI Omniverse and continuing themes from previous years.

The event underscored NVIDIA's ongoing innovation and market leadership in both AI and graphics technology, setting the stage for competition with AMD.

A major highlight was the launch of the Blackwell Ultra AI platform, which boasts a remarkable 40x performance increase over the previous Hopper architecture, enhancing AI model training and inference efficiency.

NVIDIA also introduced the Blackwell Ultra chip, known as the B300, and unveiled Dynamo as a replacement for its Triton libraries, aimed at improving AI deployment.

The Blackwell Ultra GPUs, now in full production, deliver 1.5x more AI performance than earlier models, creating new revenue opportunities for AI factories.

Additionally, the DGX Personal AI computers, powered by the NVIDIA Grace Blackwell platform, will be available from major manufacturers like ASUS, Dell, and HP, facilitating the development of large AI models.

NVIDIA introduced the Spectrum-X and Quantum-X networking platforms, utilizing silicon photonics to connect multiple GPUs across sites while significantly reducing energy consumption.

Huang also announced the emergence of 'Agentic AI' and 'Physical AI,' emphasizing their transformative potential in robotics and AI capabilities.

Looking ahead, NVIDIA's roadmap includes the Vera Rubin architecture, expected to enhance data transfer speeds between chips, crucial for large AI systems.

The RTX Pro series of GPUs was unveiled, targeting professional graphics markets, with the high-end RTX Pro 6000 offering significant AI performance and CUDA cores.

A collaboration with Google DeepMind and Disney Research resulted in the Newton Physics Engine, an open-source tool designed to improve robots' learning of complex tasks.