Nvidia CEO Jensen Huang took center stage at the annual developer conference in San Jose, California, unveiling groundbreaking AI hardware and software innovations. With a focus on meeting the growing demand for efficient AI training and inference, the company introduced its new Blackwell Ultra GPU—designed with expanded memory capacity to support larger AI models. While this powerful chip is slated for release in the second half of 2025, a design flaw has led to manufacturing delays, reflecting an industry shift to deploying AI in real-world applications.
Not stopping there, Nvidia also presented the Vera Rubin computing system, named after the pioneering astronomer. This innovative system, which combines a custom-designed processor with next-generation GPUs, is tailored for hyperscale AI workloads and promises improved chip-to-chip data transfer speeds. Set to launch in late 2026, it will be followed by the Vera Rubin Ultra in 2027 and the breakthrough Feynman architecture in 2028.
Despite the production hurdles, Nvidia's commitment to pushing the boundaries of AI technology remains unwavering. Tech enthusiasts and innovators alike can look forward to these developments as they promise to reshape how AI models are trained and deployed. Stay tuned for more updates on these exciting advancements! 
Reference(s):
Nvidia unveils updated AI solutions amid design flaws, industry shifts
cgtn.com
