Home » CFET transistors, before 2030?

CFET transistors, before 2030?

The technology is amazing. Not a single chip with GAA Transistors, also called Nanosheet depending on the company we are looking at and we already have the first to name them with a date included. TSMC again seems to be ahead of everyone else by re-launching its chips with CFET transistors in a very narrow timeframe, which as we saw last week is a very big move for any processor or graphics card .

It was in a symposium for a few journalists, but in this type of more personal and perhaps a little more informal event than a big presentation, that’s where TSMC dropped the bomb that will be repeated this summer in several conferences around the world. Are they really that advanced in this area?

CFET or Complementary FET, the next step of GAAFET for TSMC


We have already spoken briefly about this type of transistors that promise to revolutionize technology and make it more expensive almost equally, on devices N-type and P-type and how TSMC, Samsung and Intel will stack them vertically to reduce power consumption per transistor. Well, what we talked about that day left a period of about 10 years ahead, where several problems had to be overcome that currently have no solution.

Now Zhang comes along and drops a hard-to-digest bombshell for Samsung and Intel that has tungsten disulfide as the protagonist:

This (CFET) is still in the research phase. It’s just a transistor option. I don’t think I can give you a timeline for putting this transistor technology into production. We believe that all these new devices and materials will help us advance technology. Likewise, we think 3nm will be a long knot.

We will continue to see high volume demand on this node. But in terms of the transition from 3nm to 2nm… Nanosheet has a unique advantage in terms of improved computational and power efficiency due to the transistor architecture. We would expect customers whose products require greater power efficiency in terms of compute requirements to first move to 2nm.

Is Samsung getting ahead of itself? Is Intel still lagging behind?


Well, Zhang talks about it in some very interesting statements since CFET is still far from being a star transistor:

Samsung is first and is now adopting Nanosheet, but this has instead scared off customers like Qualcomm and NVIDIA and brought them to TSMC because those customers are worried about runtime risks. Intel’s plan also predates TSMC’s, probably by about a year, but again, whether Intel can execute it is another matter.

Bringing new technology to market with predictable quality, performance, and therefore cost and volume, requires careful judgment about technology readiness and execution capabilities, and this is where TSMC really shines. And it is that the 3nm and 2nm will overlap [y] They will coexist for some time.

The truth is that recent history certainly favors TSMC, but at the same time the competitive landscape has changed with Intel ready to invest far more than it has in the past 20 years, and also has the backing of the government, which hadn’t happened on that scale, so he can go along with TSMC.


In any case, CFET will be a very important change and at the same time a challenge that must first be met and then work to overcome it. The estimates speak of 2030 for CFETbut TSMC will have their 2nm with Nanosheet in 2025 and as we have seen they are already working on it, so they can move forward and outperform Intel and Samsung again.

About the author


Add Comment

Click here to post a comment

Your email address will not be published. Required fields are marked *