On Sunday, Nvidia CEO Jensen Huang gave a speech on the eve of Computex 2024 at the Taipei International Computer Show in Taiwan. In his speech, Huang stated that the company will launch Blackwell Ultra chips in 2025 and plans to launch the next generation platform called Rubin under development in 2026.

In his speech, Huang disclosed Nvidia's new semiconductor technology roadmap, which is updated annually. Huang said, "Our company's update pace is on an annual basis. Our basic philosophy is very simple: to build a data center scale, implement it step by step at an annual pace, and deliver products to customers, achieving technological breakthroughs in various fields."

Previously, Huang stated in a conference call on May 22 for the first quarter of the 2025 fiscal year that Nvidia will design new chips annually. Huang stated in his speech that Blackwell Ultra will be launched in 2025.

Huang also introduced the Rubin platform for the first time, which is the next generation platform after Blackwell. It is reported that Rubin will be equipped with a new ARM based CPU - Vera, as well as an advanced network platform that utilizes NVLink 6, CX9 SuperNIC, and X1600, and integrates InfiniBand/Ethernet switches.

Huang stated that the upcoming Rubin AI platform will use the next generation of high bandwidth memory - HBM4. This is currently the bottleneck in AI chip production, and it is reported that SK Hynix's HBM4 has been sold out by 2025.

Nvidia traditionally launches a new platform every two years: Ampere in 2020, Hopper in 2022, and Blackwell in 2024. The latest generation Rubin platform is expected to be launched in 2026.

In his speech, Huang also mentioned that companies in various industries that do not embrace artificial intelligence will be eliminated. Only through artificial intelligence chips can companies control the "computational inflation" brought about by the growing demand for data.

Huang stated that as the amount of data to be processed grows exponentially, traditional calculation methods cannot keep up. Only through Nvidia's accelerated calculation method can costs be reduced. He claims that using Nvidia's technology can save 98% of costs and reduce energy demand by 97%.

He also introduced how the combination of GPU and CPU can achieve up to 100 times acceleration, while only increasing power consumption by three times, resulting in a 25 times improvement in performance per watt compared to using CPU alone. "The more you buy, the more you save!" said Huang.

"Today, we are at the forefront of a significant transformation in the field of computing," said Huang. "The intersection of artificial intelligence and accelerated computing will redefine the future."

In addition, Huang also announced other plans related to artificial intelligence.

For example, in order to meet the growing demand of AI for high-performance Ethernet network platforms, Huang announced plans to launch new Spectra-X products every year. Spectrum X is the world's first Ethernet network platform specifically designed for AI, which can improve network performance by 1.6 times compared to traditional Ethernet network platforms. Spectrum X can accelerate the processing, analysis, and execution speed of AI workloads, thereby accelerating the development and deployment speed of AI solutions.

For example, Nvidia stated that NVIDIA RTX AI PCs equipped with RTX technology will completely change the consumer experience through over 200 RTX AI laptops and over 500 applications and games using AI technology. The RTX suite and the latest PC based NIM inference microservice for the NVIDIA ACE digital human platform will further enhance the accessibility of AI. In addition, Nvidia has also released a demonstration G-Assist project of AI assistant technology equipped with RTX, showcasing context aware assistance features for PC games and applications.

Nvidia mentioned that Microsoft is working with them to help developers incorporate new generative AI features into their Windows native and web applications, allowing developers to easily access RTX accelerated SLM through APIs.

It is worth noting that in addition to Nvidia, several executives related to the chip industry will give speeches during Computex 2024.

During his speech at Computex 2024, AMD CEO Su Zifeng focused on introducing the company's AI chips for artificial intelligence computers. She also vigorously promoted the new CPU and game related products, and revealed the cooperation between AMD and hardware manufacturers such as Lenovo.