TSMC’s 3nm/2nm chip roadmap is basically determined

In 2018, TSMC and Samsung launched 7nm Chips and began to lead Intel. In 2020, TSMC and Samsung entered 5nm, leaving Intel far behind, establishing the two dominant positions of chip foundry (manufacturing) in one fell swoop.

According to the plans of TSMC and Samsung, 3nm chips will be launched in 2022, and 2nm chips will be launched around 2024. From then on, there will be no rivals in the world.

As 2022 is getting closer, there are more and more news about 3nm chips, and TSMC and Samsung have also launched 3nm competition. After all, whoever launches first and who has better technology may take the lead.

Thinking that when it was in 14nm FinFET technology, Samsung led TSMC to launch 14nm, while TSMC was 16nm. Later, Samsung stole some of Apple’s orders, which put a lot of pressure on TSMC.

In the 3nm era, Samsung has repeated its old tricks, wanting to adopt GAA technology, ahead of TSMC’s FinFET technology, and even want to launch 3nm technology earlier.

Under the pressure of Samsung, TSMC finally talked about 3nm chips recently, and also talked about 2nm. Judging from the content of TSMC CEO Wei Zhejia’s speech, the 3nm and 2nm routes are basically set.

According to TSMC, there will be two versions of 3nm, one is the 3nm 3 GAE low-power version, which will be mass-produced in early 2022, but delivered to customers, it may go to early 2023, which is about half a year later than expected. So next year’s Apple A16 chip, it must be a 3nm process.

This low-power version of the 3nm chip has a 70% increase in transistor density compared with 5nm, the performance can be improved by 10-15% under the same power consumption, and the power consumption can be reduced by 25-30% under the same performance.

The 3nm 3GAP high-performance version will be mass-produced in early 2023, and the delivery will go to the second half of 2023 or even early 2024. The high-performance version can be regarded as an enhanced version, that is, the second-generation 3nm process, which will have better performance, power consumption, and yield. At the same time, it is fully compatible with N3 in design and IP, and will be mass-produced in 2024.

The 2nm chip is expected to be mass-produced in 2025. The 2nm chip is expected to use GAAFET (Gate Surrounding Field Effect Transistor) technology instead of the current FinFET technology.

Of course, before the chip is delivered, everything is a plan, and sometimes the plan can’t keep up with the changes, so in the end, time will tell.

The Links:   EP3C25F324I7N LM170E03-TLB2