At the NVIDIA media meeting at the Computex conference in Taipei a few days ago, the company's CEO Jen-Hsun Huang said that there are indeed certain limitations in chip supply, which is still a worrying issue. "We have secured supply for strong growth in all of these systems," he stressed. "We have enough supply for very strong growth, but supply constraints do remain."
Talking about future growth drivers, Huang Renxun pointed out that the company's Vera CPU will be more popular than its GPU because Vera plays a vital role in processing information. "Vera CPU will become our new main growth driver."
Vera is equipped with 88 NVIDIA self-developed Olympus cores, with a memory bandwidth of 1.2TB/s. Single-core performance is 50% higher than the previous generation, and it is suitable for complex scenarios such as orchestration, reinforcement learning, and long context management.
Vera can work together with Rubin GPU, BlueField 4 DPU and other components. Relying on a unified memory architecture, the data supply energy efficiency is doubled compared with traditional facilities, helping large-scale intelligent AI run efficiently.
With the rapid development of agent AI (Agentic AI) that can perform tasks autonomously, market demand is expanding from a single GPU to a complete computing power system in which CPU and GPU collaborate. The official delivery of Vera will not only open up a new growth curve for NVIDIA, but will also accelerate the comprehensive transformation of global AI infrastructure into the era of intelligent agents.