Chinese semiconductor industry

[Overbom]

For now yes, but for 2025? the chip will be used widely are not 28nm and 14nm anymore, but perhaps 5nm and 3nm, while 2nm and 1nm will have been mass produced and 0.5nm will have been on trial and

AFAIK currently 28nm is important. By 2025 I estimate that 14nm should be the mainstream.

Also top-end chips are not the most used as there are a lot of industries which use low-tech chips. No way we reach 2nm (very small possibility), 1nm, or 0.5nm (physically impossible?) by 2025. 3nm already requires such large R&D expenditure and investment and it gets so many delays. Just some days ago we were talking about TSMC having to delay its 3nm program.

Any member here has a graph/chart showing China's demand for nodes?

 

[AndrewS]

For now yes, but for 2025? the chip will be used widely are not 28nm and 14nm anymore, but perhaps 5nm and 3nm, while 2nm and 1nm will have been mass produced and 0.5nm will have been on trial and

The development costs get exponential higher the smaller the node.
Plus per transistor costs start creeping upwards once you reach 14/28nm

So unless you have really high volumes PLUS a requirement for the lowest energy consumption - you don't need anything smaller than 14/28nm. That will remain the mainstream.

 

[AndrewS]

AFAIK currently 28nm is important. By 2025 I estimate that 14nm should be the mainstream.

Also top-end chips are not the most used as there are a lot of industries which use low-tech chips. No way we reach 2nm (very small possibility), 1nm, or 0.5nm (physically impossible?) by 2025. 3nm already requires such large R&D expenditure and investment and it gets so many delays. Just some days ago we were talking about TSMC having to delay its 3nm program.

Any member here has a graph/chart showing China's demand for nodes?

Foundry Market by Feature Dimension.
In 2020, 80% of demand was >20nm

 

[horse]

Please, Log in or Register to view URLs content!

some humour

 

[antonius123]

Foundry Market by Feature Dimension.
In 2020, 80% of demand was >20nm


View attachment 76668

According to this graph, in year 2020 16nm or 14nm have the largest market share, 32/28nm come as second. So it still makes sense to say that 2 years ahead the largest market share will shift from 14nm to 7nm or even to 5nm.

 

[antonius123]

The development costs get exponential higher the smaller the node.
Plus per transistor costs start creeping upwards once you reach 14/28nm

So unless you have really high volumes PLUS a requirement for the lowest energy consumption - you don't need anything smaller than 14/28nm. That will remain the mainstream.

It doesn't matter. The high end product such as the flagship smartphones will consume all of those smallest node. It is very expensive at the beginning as it is the smallest node and the scale of production is still very low; but when it is not the smallest anymore the demand/production will increase then the production cost will reduce and when the node become the most consumed chips the cost will be among the cheapest.

 

[tokenanalyst]

Please, Log in or Register to view URLs content!

some humour

I hope that soon Huawei can respond with "Thank you very much but it is too late now, we buy local or we manufacture our own"

 

[AndrewS]

It doesn't matter. The high end product such as the flagship smartphones will consume all of those smallest node. It is very expensive at the beginning as it is the smallest node and the scale of production is still very low; but when it is not the smallest anymore the demand/production will increase then the production cost will reduce and when the node become the most consumed chips the cost will be among the cheapest.

Uh no. It's the upfront design cost which increases significantly as you move to a smaller node.

The production cost for each transistor keeps increasing slightly after 14/28nm.

So the cheapest semiconductor chips should be produced at 14/28nm. But these won't be as energy efficient as a 5/7nm.

Thankfully, smartphone chips are a small subset of overall chip demand.

 

[FairAndUnbiased]

For now yes, but for 2025? the chip will be used widely are not 28nm and 14nm anymore, but perhaps 5nm and 3nm, while 2nm and 1nm will have been mass produced and 0.5nm will have been on trial and

that's not how it works. the vast majority of chips produced are ASICs or mixed signal chips (battery management, ADCs, microcontrollers, power, op amps, etc) where there is no benefit to make them at 5 nm or whatever.

Indeed, sometimes it is harmful to make them at small nodes because the smaller the node, the lower voltage and current it can withstand. These chips cannot go smaller.

 

[antiterror13]

Uh no. It's the upfront design cost which increases significantly as you move to a smaller node.

The production cost for each transistor keeps increasing slightly after 14/28nm.

So the cheapest semiconductor chips should be produced at 14/28nm. But these won't be as energy efficient as a 5/7nm.

Thankfully, smartphone chips are a small subset of overall chip demand.

Correct and smartphone with =< 7nm is very high end, very much only "nice to have" .. not really essential