Chip engineers say AI chip design is not the next coding boom, at least not yet
An AI lab researcher and a veteran chip engineer agree AI is coming to chip design. They disagree on how fast, and on what counts as success.
The short version
- A researcher at an AI lab sees chip design as the next coding-style market for AI.
- A chip engineer with nearly ten years of experience says code is only a small part of chip design.
- He puts the gap between leading foreign models and Chinese ones at about 6 months.
- Both say the real test is a chip made and shipped, not a demo or a policy announcement.
- Both expect big change in 5 to 10 years, with a lot of uncertainty.
Why it matters
Big AI promises often skip the hard last step, which in chips is a physical product that must work. This story shows how to read such claims: ask whether the work has been checked in the real world. It also shows why AI labs want to move into chips, because chips are what their own computing costs depend on.
Two answers to one question
As tmtpost.com reported, two people were asked whether AI for chip design is the next big use of AI after coding. The answers were nearly opposite.
A researcher at an AI model company, who goes by 剁椒酱, says yes. His logic is that chip design takes thousands of people only because each human can hold a small amount of information at once. An AI does not have that limit, so a smart enough one could take over much of the work.
A chip designer known as 晒哥, with close to ten years in digital chip design, laughed at the idea. In his view, chip design is more than writing code.
Why AI labs want chips
The article gives a business reason. Coding is the best-proven money-maker for AI companies, but its market has a ceiling. Chip design is valuable, rich in expert knowledge, and its results can be checked. It is also paid for by profitable chip and design-software companies.
There is a second motive. If a lab's model is that capable, the researcher asks, why keep buying chips from others? Chinese labs also face a shortage of chips. Job postings from AI labs now ask for chip-design skills, which both men read as a sign of the same push.
Why the engineer is unconvinced
The engineer says a chip must end up as a physical object. Designers have to balance power use, speed and size, which pull against each other, and handle testing too. Writing code is the easy part, and AI already helps with that.
He rates most so-called chip-design AI agents as very early, little more than a model connected to a command line. Even a design-software giant's AI tool is, to him, more of an assistant that looks things up and writes small scripts.
Companies are cautious too. Few will spend a huge budget to have an AI-built chip manufactured. His line on demos: a demo that never goes to manufacturing is just a demo.
What would prove it
Both men dismiss announcements, job posts and government plans as weak evidence. The stronger signs they list are real chips made and shipped, AI tools accepted in the industry's standard sign-off process, and a clear way for AI labs, software makers and chip firms to share revenue.
The engineer expects the topic to be settled when media stop writing about the models and start writing about the chips: their specs, uses and shipment numbers.
Both expect chip-design companies to be squeezed from several sides and sorted into new layers. Both put big change at 5 to 10 years, if nothing unexpected intervenes.
What to watch
- Will any company pay for a real manufacturing run of a chip designed mostly by AI?
- Will the industry's standard design software accept AI-driven tools in its sign-off step?
Try it
Next time you read that AI will transform a field, look for one thing: a finished product in real use, not a demo. Then check whether people who work in that field agree.
Sources and links
- Reportingtmtpost.com: the article we sourced
Liked it? Say why
Add a line for other readers.
No account needed. We never show your email. Comments with links, ads or abuse are held and not shown. Your comment shows after a quick check.
Listen to the AI Wire headlines
Listen to the AI WireListen to our daily AI shows every morning: on your commute, in the shower, wherever you are
Listen to the daily showsFirst reported by tmtpost.com · Read the original. Written by the SI Agent News Desk from the reporting above.Written with AI from the source above. Every number is checked against the source. Spot a mistake? Corrections.
