Is China's chip technology only 3 years behind TSMC? Chairman of the National Science Council: more than 10 years

[Daniel Nenni]

Wu Chengwen, chairman of the National Science Council. United Daily News file photo/Photography by reporter Chen Zhengxing

Some media reported that China's chip technology strength is only three years behind TSMC . Wu Chengwen, chairman of the National Science Council, said today that he has doubts about this statement. The technology gap between the two sides should be more than 10 years, especially when TSMC's advanced process will enter 2 nanometers .

The Education and Culture Committee of the Legislative Yuan today invited Wu Chengwen to give a special report on "Sustainable Scientific Research and Policy Communication" and to prepare questions.

Democratic Progressive Party legislator Wu Peiyi pointed out during the inquiry that according to recent Japanese media reports, after a Japanese semiconductor survey company dismantled Huawei mobile phones, it said that China has caught up and its chip technology strength is only three years behind TSMC. She asked Wu Chengwen whether this statement is true.

Wu Chengwen responded, "I actually have doubts." The gap should be more than 10 years, especially when TSMC's advanced process will enter 2 nanometers.

According to information disclosed by TSMC at a corporate briefing in July, the research and development of the 2-nanometer advanced process is progressing smoothly. The device performance and yield are as planned or better than expected, and mass production will be launched in 2025 as scheduled.

In China, market research agency TrendForce pointed out that China is the most active in expanding production of 28nm and more mature processes. It is estimated that China’s mature process production capacity will account for 39% of the world’s share in 2027, with applications covering automotive and consumer applications. Type, server and industrial control and other fields.

China's Ministry of Industry and Information Technology recently announced a catalog of guidance for the promotion and application of major technical equipment. Two of the exposure machines have been interpreted by the outside world as China has made a major technological breakthrough and can produce wafers of 8 nanometers and below. According to Deutsche Welle, Shanghai Microelectronics Applying for a series of EUV technology patents to the National Intellectual Property Administration of China may mean that it has mastered the key capabilities of manufacturing 7nm and below wafers.

中國晶片技術只落後台積3年？ 國科會主委：10年以上 | 聯合新聞網

有媒體報導稱中國晶片技術實力僅落後台積電3年，國科會主委吳誠文今天表示，他對此說法存疑，雙方技術落差應是10年以上，尤其...

udn.com

 

[chenyj]

Both 3-year (~1 advanced node) and 10-year (~ 4 nodes) gap (behind TSMC) are under/over-estimating the current Chinese (mainland) semiconductor status. There is no doubt SMIC+Huawei already can make 7-nm node chips, and probably grasped the 6-nm node (min. metal pitch ~ 32nm) but with much lower yield, and are working on 5-nm node (pathfinding). It seems to me they are lagging ~ 2.5 nodes behind TSMC, but definitely not 10 years. And these are all based on their access to 193i DUV scanners. Today, the real focus of Chinese semiconductor is lithography, and they are not worrying much about other processing/device technologies. Last month they just disclosed their 193nm dry scanner (resolution ~ 65nm, overlay ~ 8nm) and it's about 20 years behind ASML's EUV scanners. Some media reports misunderstood the 8-nm overlay accuracy as the resolution capability. To my knowledge, scanner technology is the real concern of Chinese central government, rather than TSMC's processing/device technology advancement. They do have plans to get around the mainstream EUV source (e.g., so-called "EUV factory") and develop maskless EUV direct writer (so called "digital" EUV scanner) to replace the mask based EUVL. However, all these will take very long time (much more than 10 years) to develop.

 

[Xebec]

Both 3-year (~1 advanced node) and 10-year (~ 4 nodes) gap (behind TSMC) are under/over-estimating the current Chinese (mainland) semiconductor status. There is no doubt SMIC+Huawei already can make 7-nm node chips, and probably grasped the 6-nm node (min. metal pitch ~ 32nm) but with much lower yield, and are working on 5-nm node (pathfinding). It seems to me they are lagging ~ 2.5 nodes behind TSMC, but definitely not 10 years. And these are all based on their access to 193i DUV scanners. Today, the real focus of Chinese semiconductor is lithography, and they are not worrying much about other processing/device technologies. Last month they just disclosed their 193nm dry scanner (resolution ~ 65nm, overlay ~ 8nm) and it's about 20 years behind ASML's EUV scanners. Some media reports misunderstood the 8-nm overlay accuracy as the resolution capability. To my knowledge, scanner technology is the real concern of Chinese central government, rather than TSMC's processing/device technology advancement. They do have plans to get around the mainstream EUV source (e.g., so-called "EUV factory") and develop maskless EUV direct writer (so called "digital" EUV scanner) to replace the mask based EUVL. However, all these will take very long time (much more than 10 years) to develop.

It's variable years depending on which metric you use.

If we're talking leading edge node, at scale, at a good cost - it's certainly more in the 10+ year range.

If we're talking about 'can produce a transistor of X dimension' then maybe it's less.

 

[Fred Chen]

The 7nm was 5 years after TSMC, so they need to release 5nm this year to keep the same gap. Otherwise, they will start falling behind.

 

[samwilde]

10 years? SMIC's N+2 has similar transistor density to the TSMC N7+ process which came out in 2019.
The Huawei Mate 60 with SMIC N+2 process came out last year in 2023.

SMIC is like 4 years behind TSMC. And that is with the sanctions. Had they been allowed to buy EUV machines the gap would be even narrower.

How can you even claim that SMIC lacks scale at making their 7nm process when tens of millions of Huawei phones with it have been sold in a year? And this is just one product using the 7nm process. Huawei also sell millions of baseband towers which likely use such chips. And SMIC has other customers with FinFET process, plus other Huawei products like the Ascend AI processor, or the Huawei self-driving chips.

Phytium, Hygon, and Loongson are currently in the Entity List. They cannot fab chips at TSMC. And yet they continue to operate. So where are they making their CPUs?

 

[Barnsley]

10 years? SMIC's N+2 has similar transistor density to the TSMC N7+ process which came out in 2019.
The Huawei Mate 60 with SMIC N+2 process came out last year in 2023.

SMIC is like 4 years behind TSMC. And that is with the sanctions. Had they been allowed to buy EUV machines the gap would be even narrower.

How can you even claim that SMIC lacks scale at making their 7nm process when tens of millions of Huawei phones with it have been sold in a year? And this is just one product using the 7nm process. Huawei also sell millions of baseband towers which likely use such chips. And SMIC has other customers with FinFET process, plus other Huawei products like the Ascend AI processor, or the Huawei self-driving chips.

Phytium, Hygon, and Loongson are currently in the Entity List. They cannot fab chips at TSMC. And yet they continue to operate. So where are they making their CPUs?

How was your golden week?

I avoided the crowds by going during mid-autumn festival.

 

[count]

I think the gap is probably around 4-6 years. It would probably be more like 1 year if it wasn't for EUV sanctions.

I think the gap will widen a little before closing. So they may stall out for the next few years while China builds it's in house alternatives to EUV, and as soon as they do that the gap will rapidly close. But by that time there might have been enough progress in quantum that that becomes the next thing, and China is very strong in quantum.