Just now, the industry’s first RISC-V AI computing power super node solution made its debut at WAIC 2026

Intelligent Things Author | Yunpeng Editor | Mo Ying Currently, AI is developing rapidly, the wave of Agent intelligence is surging, the demand for large model reasoning has skyrocketed exponentially, and 10,000-ka or even 100,000-ka level intelligent computing clusters have become the standard "arms" for giants to compete in large models. In this context, AI computing power has become the focus of the entire industry, and the "barrel shortcomings" have gradually become apparent: the "computing power wall" and "ecological wall" have obvious constraints. On the one hand, the single-chip physical process based on the traditional mainstream architecture is approaching its limit, with diminishing marginal benefits. Tens of thousands of chips are interconnected, and latency and bandwidth bottlenecks drag down performance. On the other hand, large models are rapidly iterated and algorithm structures are changing with each passing day. However, the development of underlying operators is still in a "manual workshop" style, and the test of the computing power ecosystem has become a difficulty faced by chip manufacturers. Today, the competition for AI chips has shifted from pure “peak computing power competition” to “token economic benefits and software and hardware collaborative competition.” During WAIC 2026, Yixing Intelligent, the domestic RISC-V AI computing power chip unicorn, handed over a milestone answer sheet, which fully demonstrated the system-level solution from Epoch self-developed chips, super nodes to full-stack software, and unveiled the industry's first RISC-V AI computing power super node. ▲RISC-V AI computing power super node based on Epoch chip 1. Hard core "dismantling" the first domestic RISC-V super node, how to break through the Wanka cluster computing power wall? Whether it is large model training or inference, factors such as computing power density, efficiency, energy consumption, cost, and stability have become more important than pure peak computing power. The industry is in urgent need of overall system-level AI infrastructure solutions to achieve deep collaboration of software and hardware covering the entire link of chips, servers, and cluster scheduling. In the early stage of the release of RISC-V AI computing power super node, Zhidongxi had in-depth exchanges with Yixing Intelligence, and went into many technical details of this product and its full-stack software and hardware solutions in advance. It can be said that this solution will become the "better solution" for the AI ​​computing power industry to crack the computing power and ecological wall. Yixing Intelligence mentioned that similar to Linux in the PC computing era and Android in the mobile computing era, RIS