EUV is not easy.
Chinese semiconductor industry
- Thread starter Hendrik_2000
- Start date
- Status
Of course, I am not arguing that silicon based development is dead now.
However we are soon approaching the physical limits of how low we can go with silicon (approx 2nm). Now they may find some new techniques and architecure to extend the limit but it wont do much.
So at 2nm, the current silicon based chips cant advance any more by shrinking the process size
Thats its is good that they are researching other ways to get more performance gains as silicon is at its last legs and it is time to utilise nee materials. I also agree that it will take a lot of.time to replace it (~10 yrs)
Node shrinkage can stop purely because of commercial reasons too. For most of the chip industry, profit margins are extremely low. Each new process used to almost half the costs for the same performance. Now it is around 10%. This why most of the chips don't use below 10 nm processes. Nobody used 5 nm except Apple and Huawei initially.
Two years ago, I asked in another forum with some semiconductor experts the question of what potential large-scale application could drive the demand for even smaller processing nodes after smartphone.
Smartphone is what's been driving the current wave of going for 7nm and below. It has the need and the scale to justify the increasingly huge cost for a fab for advanced nodes. IoT chips probably have the potential volume, but not the complexity.
Nobody could come up with a good answer, but some suggested that the trend is more likely for innovation in architecture, not smaller nodes as it hits the physical limits.
True. Intel recently got a 20%+ increase in performance by improving their architecture. This more than the performance difference between the first TSMC 7 nm and the recent 5 nm.
5nm is still new, so it will likely get cheaper as time goes on.
So 5nm chips will be somewhat cheaper than their 7nm equivalents.
Plus 5nm will save a little on space and also has some improvement in performance and power consumption.
See below.
However chip volumes have to be large given the added development cost. See below
China main strategic semiconductor opportunity lies at electric vehicle sector not cellphone, PC, AI or Server. It's estimated by 2025 , there will annual electric car ouput of 5 million in China market alone and each electric car take about 400 to 500 chips. It will go up more in the future. Maybe needs 10,000 chip in each electric vehicle further in future.
So do the math, by 2025, 5million cars *500=2.5billion chip each year just for electric car in China market.
And 28nm process is sufficient for 90% of 2.5billion chips annually.
That's expansion of 28nm capacity is greatly needed. Beijing and Shenzhen currently expands at this node. More will come online to meet the great demand for electric vehicle.
So do the math, by 2025, 5million cars *500=2.5billion chip each year just for electric car in China market.
And 28nm process is sufficient for 90% of 2.5billion chips annually.
That's expansion of 28nm capacity is greatly needed. Beijing and Shenzhen currently expands at this node. More will come online to meet the great demand for electric vehicle.
broadsword
Brigadier
I would think "cellphone, PC, AI or Server" are as important as electric cars.
China would encounter technical difficulties.
SMIC has a record revenue year in 2020 over $4 billion and majority of that from mature process above 28nm.
Last edited:
- Status