尊敬的各位嘉宾、亲爱的业内同仁们:
大家好!欢迎参加 2026 年 DVCon China 大会!作为本次大会的主席,我非常荣幸能够在这里与大家共同探讨 设计验证领域的最新动态与前沿技术。
过去一年,全球半导体产业在波动中复苏、在变革中前行,进入了以 AI 算力、先进封装、车规电子、自主可控 为主线的全新发展周期。全球市场规模持续扩张, AI 驱动的数据中心芯片、高带宽内存、车规级半导体成为增长核心, 产业格局加速重构,多极化趋势日益明显。与此同时,中国半导体设计产业顶住外部压力、抓住内需机遇,呈现出 结构性突破、系统性升级、生态化成长的鲜明特征。
MORE >
刘斌
AI-Driven Hardware Engineering: From Automation to Autonomy
Chip complexity has made traditional design and verification workflows unsustainable — verification alone consumes over 60% of project schedules and carries a 76% risk of costly respins. The path forward is AI that doesn't just automate tasks, but reasons through them.
This talk presents a production multi-agent system for hardware engineering, covering AI agents for design and verification — deployed at enterprise scale and achieving over 70% efficiency gains. We discuss the architectural principles that make this work: structured design representations as a shared memory layer, tool-augmented reasoning with EDA integration, and orchestration frameworks that coordinate agents across long-horizon tasks. We also share real experiences with fine-tuning language models for hardware and the nightly infrastructure we built to measure agent quality rigorously.
The talk closes with a look at the broader vision: collaborative multi-agent systems that give every hardware engineer a team of AI specialists — amplifying human judgment across design, verification, and physical design, and compressing timelines without sacrificing engineering control.
IC Design and Verification Transformation in the Era of Agentic AI
With the profound advancement of artificial intelligence, Agentic AI has brought a paradigm shift to the field of chip design. Different from the traditional "AI + EDA", the EDA Agent breaks away from the conventional human-led design model supported by EDA tools, freeing engineers from tedious implementation details and enabling them to focus on higher-value work such as architectural innovation, strategic planning and complex decision-making. The next generation of competition in EDA technology will be determined by AI-centric technological breakthroughs. Targeting the real efficiency pain points in IC design, UniVista focuses on reshaping the EDA toolchain with Agentic AI. It analyzes industrial challenges and proposes corresponding solutions around key scientific issues including the evolution of full-flow Agents for digital IC design, the transformation of EDA underlying architecture for Agents, AI hallucination mitigation, and Agent-enabled EDA tool development. This drives the leap of EDA from tool automation to engineer automation, supports domestic EDA in achieving technological breakthroughs and building an independent, controllable ecosystem, and significantly improves chip R&D efficiency.
Copyright © 2026 DVCon all rights reserved. Site designed, developed and maintained by MCI China.