Skip to content
Theme
OnDevice Tools

Tech BusinessChips & Semiconductors

AI agents in chip design target earlier workflow shifts

Semiconductor Engineering reports that agentic AI may move chip design work earlier, though tool integration and expertise remain hard.

Illustration of a chip design lab with AI agents and software code

Key takeaways

  • Agentic AI could automate more of the chip design flow, the article reports.
  • Integrating agents with electronic design automation tools remains difficult.
  • Capturing long engineering experience and proprietary data remains a challenge.
  • Cost and complexity may give larger companies an advantage.
In this article
  1. What AI agents could do
  2. Integration and expertise hurdles
  3. Who may benefit
  4. What it means next

What AI agents could do

The debate over whether AI agents in chip design can really shift work left has gained attention. Semiconductor Engineering reports on the idea that agentic AI could automate more of the chip flow. The report was published on 18 November 2026.

Agentic AI promises to automate more of the chip flow. The article points to design, verification and validation stages. It says orchestration at several stages will need new methods.

Coordination across tools will also be needed. The aim is to move tasks earlier in the design process. The phrase shift left means doing more work earlier. The article asks how much can be shifted.

The article describes an onion of agents. It says there are lifecycle, manufacturing, digital implementation and functional verification flows. Each flow has many sub-components. Functional verification to emulation is one of the five.

An agent may do work in one area. A larger agent may coordinate several areas. A top agent may connect to the front end. The article says vendors have said they have higher levels.

The article says the industry will work on building each level. It will then iterate and improve them. That is seen as the way to get agents to learn.

Integration and expertise hurdles

Illustration: AI, price, market trend, news and the labels Agentic AI, AI Agents, Chip Design

Integrating AI agents with electronic design automation tools is not simple. The article calls this non-trivial. Capturing decades of engineering experience is also hard. Engineering expertise remains essential.

Proprietary data is still siloed. That may limit how useful an agent is. The article says agentic AI companies may lack field experience. Their tools may not match what the industry needs.

The article says the big consumers are building their own tools. They know what they want. They have field experience and data from many years. The big EDA vendors may go to streamline this.

How that will shape the end use is not clear. The article asks where the big EDA vendors will go. The cost and complexity could limit adoption.

Who may benefit

The article suggests smaller chip designs could benefit if costs fall. It says large companies may still have an advantage. The biggest consumers may build their own systems. They know what they want, according to the article.

Cost and complexity could limit adoption. Only larger companies may be able to afford the work. The article asks who will pay for it.

The article says big consumers have been running AI full on. They have not written code. Their work is all agents. The cost of doing all this may be reduced.

Smaller designs that were difficult may benefit. But an Nvidia competitor still could not do this from a garage. That would still be a large company.

What it means next

The article describes levels of agents. It says vendors have talked about higher levels. The industry is still working out how agents combine. No clear method is known today.

The discussion includes emulation, formal methods and simulation. It notes that experience from individuals with long industry careers is missing. The article says AI may handle some challenges, but unique skills remain.

The article says AI is doing a great job on some challenges. It also says experience from individuals with long careers has unique skills. The industry needs to gather those skills. The article asks whether adoption will lead to a coherent story.

The article says information may be fed back to specification and functional stages. That can improve the probability of manufacturability. The article asks how much can be abstracted.

Sources

  1. Can AI Agents Really Shift Chip Design Left?Semiconductor Engineering

Tags

Spotted a mistake? See our corrections policy. Read how News by OnDevice works in our editorial policy.