Nvidia CEO Jensen Huang has announced plans to build a next-generation AI platform called “Rubin,” named after astronomer Vera Rubin.
Huang made the announcement in a speech ahead of the COMPUTEX technology convention, which opens in Taipei on June 4.
Huang spoke to about 6,500 industry executives, journalists, entrepreneurs, gamers, inventors and AI enthusiasts gathered in the glass dome of the National Taiwan University Sports Center in Taipei, according to a company blog.
The CEO revealed an ambitious goal of introducing a new “AI accelerator” every year, a rapid refresh cycle aimed at continually improving the performance, efficiency and capabilities of AI systems.
Part of this strategy is the Rubin chip, which is expected to ship in 2026. Interestingly, the announcement of the Rubin platform comes just months after Nvidia unveiled its Blackwell chips for 2025.
About the Rubin Platform
The Rubin Platform is poised to transform the field of artificial intelligence in the same way that astronomers have revolutionized our understanding of the universe.
Vera Rubin pinpointed the existence of the elusive dark matter and led research into the structure and motion of galaxies.
The Rubin platform uses high-bandwidth memory (HBM4) to deliver unprecedented performance, and also incorporates new graphics processing units (GPUs), advanced Arm-based CPUs called Vera, and cutting-edge networking technologies including NVLink 6, CX9 SuperNICs, and the X1600 integrated InfiniBand/Ethernet switch.
“We work in one-year cycles. Our guiding principle is very simple: build an entire data center in one year cycles, take it apart and sell the parts, and push everything to the limits of technology,” he explained.
Not many details have been revealed about the Rubin yet. It is expected that the Rubin Ultra will follow the Rubin in this “one-year cycle.”
Sustainable Computing
Huang emphasized that “accelerated computing is sustainable computing.”
NVIDIA's accelerated computing approach combines the GPU and CPU to dramatically increase performance while maintaining relatively low power consumption, resulting in significant efficiency and up to 25x better performance per watt compared to using the CPU alone. This makes Rubin not only powerful, but also sustainable accelerated computing.
“The future of computing is accelerating,” Huang said. “Our innovations in AI and accelerated computing are pushing the boundaries of what's possible and driving the next wave of technological advancements.”
AI Networking Initiatives
During his speech, Huang also detailed the company's other robotics and networking efforts.
The company also revealed plans to offer “next-generation networks with Spectrum-X.” The CEO emphasized that the company will create new Spectrum-X solutions every year to meet the growing demand for high-performance Ethernet networks designed specifically for AI applications.
NVIDIA Spectrum-X is the first Ethernet fabric designed for AI, delivering 1.6x improvement in network performance compared to typical Ethernet fabrics, according to a company blog post.
Spectrum-X accelerates the processing, analysis and execution of AI operations, including faster data transfer between computing resources, faster data processing and more efficient implementation of AI algorithms.
About the Editor
Mrigakshi Dixit Mrigakshi is a science journalist who enjoys writing about space exploration, biology and technological innovations. She has professional experience in both broadcast and digital media and is trained in various storytelling formats. Her work has been featured in renowned publications like Nature India, Supercluster, Astronomy magazine etc. If you have an idea, feel free to email us.
