At the GTC conference on Tuesday, Nvidia made significant announcements, including the unveiling of its next-generation AI chips, Blackwell Ultra and Rubin, which promise to push the boundaries of artificial intelligence capabilities.
Nvidia’s CEO Jensen Huang introduced the Blackwell Ultra family, which will ship later this year, and the Vera Rubin GPU, expected to ship in 2026. The Blackwell Ultra chips are designed to significantly enhance the speed of generating AI models, offering more tokens per second and enabling cloud companies to provide faster AI services. These new chips are targeted at high-demand, time-sensitive applications.
Rubin, Nvidia’s next-generation GPU, is expected to double the performance of its predecessor, Blackwell, by supporting up to 50 petaflops for inference and 288 gigabytes of fast memory. The Rubin chips will also support up to four GPU dies, a revolutionary development in GPU design. Rubin, paired with Nvidia’s first custom-designed Vera CPU, marks a major shift in Nvidia’s strategy, with the company saying that the new architecture will revolutionize AI workloads.
Expansion of AI Capabilities
Nvidia also introduced several improvements to its existing lineup, including the release of a new system called DGX Spark, which will enable running large AI models such as Llama or DeepSeek. This is in line with the company’s continued expansion into AI-focused hardware, aiming to provide solutions that push the limits of AI modeling and machine learning.
The company also made strides in AI reasoning, aiming to ensure that their chips are well-suited to handle “agentic AI,” which is a new class of AI that focuses on reasoning to solve problems. Nvidia’s deep learning products and advanced GPU architectures are designed to tackle these cutting-edge AI challenges with increased efficiency and speed.
With the new Blackwell Ultra and Rubin chips, Nvidia is strengthening its grip on the market, particularly for AI-focused cloud services. The new chips, combined with Nvidia’s other offerings such as their advanced networking tools and software packages, position the company as a leader in powering AI-based systems worldwide. In addition, Nvidia’s long-term plans involve significant advancements in AI technology, with upcoming releases aimed at facilitating even more sophisticated and efficient AI models.
As companies such as Microsoft, Google, and Amazon continue to increase their reliance on Nvidia’s GPUs for AI workloads, Blackwell Ultra and Rubin chips may set the stage for a new era of AI technology. Nvidia’s future products, including those planned for 2028 and beyond, signal an ambitious roadmap for AI-driven solutions across multiple industries.
Author’s Opinion
Nvidia’s continued push to innovate with the Blackwell Ultra and Rubin chips cements its leadership in the AI space. While competitors like AMD and Intel are making moves, Nvidia’s focus on optimizing GPUs for AI workloads keeps it ahead of the pack. The combination of high-performance hardware and AI-specific features like reasoning support makes Nvidia’s chips essential for future AI applications, and their ability to cater to both large enterprises and cloud providers strengthens their dominance in the market.
Featured image credit: Forbes Australia
Follow us for more breaking news on DMR