NVIDIA Prepares for Early Debut of Rubin Architecture

NVIDIA's next-generation Rubin architecture could hit the market almost anytime soon half a year earlier planned due to the accelerated development of HBM4 memory from SK hynix. According to data provided by ZDNet Korea, NVIDIA's key partner SK hynix is ​​already preparing HBM4 samples, which will be shipped to NVIDIA in June 2025. Mass production will begin in the third quarter of 2025, which is almost one quarter ahead of the original plan.

It is noted that SK hynix will become the exclusive supplier of HBM4 to NVIDIA, giving Team Green access to advanced technologies before competitors. Rubin will be focused on artificial intelligence (AI) tasks and will be the first NVIDIA platform to integrate HBM4. Such memory, with speeds up to 6.4 Gbps and 24GB and 32GB layers, will significantly improve performance by eliminating the need for complex packaging technologies. The compact placement of semiconductors and logic will ensure high energy efficiency and performance, which is critical for AI workloads.

Rubin will be based on TSMC 3nm process technology and CoWoS-L packaging. The new memory standard will allow NVIDIA to significantly reduce power consumption while improving performance per watt (perf/W), which is especially important against the backdrop of growing demands of the AI ​​market. NVIDIA is expected to take the lead, as competitors will release similar solutions only a few quarters after the debut of Rubin.

If NVIDIA can bring the Rubin architecture to market as early as the second half of 2025, it will be a major step in strengthening its leadership in AI and computing.