Chinese semiconductor industry
- Thread starter Hendrik_2000
- Start date
- Status
This might work if the Chinese fabs can actually scale up their production fast enough. However other fabs will get choked first before TSMC as around 49% (and increasing) of TSMC current revenue is coming from <14nm manufacturing process. Again, time is the key here. If Chinese fabs taking too much time to build capacity and flood the market with over 14nm chips, then TSMC revenue from <14nm will get bigger thus would not be impacted as much by the flooding of over 14nm chips in the market. It is essentially a moving target for the Chinese.
Sounds like complete bs, just one of those infomercials selling product.
I am definitely not an engineer, but I think there are somethings ingrained into engineers that they do not even bother or need to talk about it.
Some are this:-
1. The widget must work, that is if it works only half the time, it is not reliable and does not work.
2. The widget must be efficient in its design. Suppose a regular family car a sedan has 5 or 6 wheels. Why 5 or 6 wheels? The sedan only needs 4 wheels. Do they sell cars with heated seats in Africa? Probably not.
3. The widget must be cost effective. Apparently a lot of engineering work is to try to design or redesign the widget to make it cheaper in terms of materials used and energy consumption.
This new ASML machine, definitely satisfies point #1 and point #2 but point #3 is a complete unknown.
You're already not getting the same bang for the buck when going from 14nm to 7nm compared to 7nm to 5nm, yet the costs to do that increased a lot more. Going from 5nm to 3nm will even be running into the marginal utility constraints even more.
Therefore, if the engineering does not produce a product that is cost effective, then that is the end of the story for that product, and some departments get axed.
============= ==============
That is the theoretical stuff. Now that is to be applied.
Do you need faster chips? Of course we do.
But, this is where the engineering part comes in, again.
Forget the details of the story, when China built one of their supercomputers, the fastest one at that time, the United States barred the sales of those advanced chips. The Chinese still were able to build the fastest supercomputer, even when there were faster chips on the market, because they had figured out how to cable and link the machines all together for their supercomputer. That system was the fastest for years.
The logic chip in the cell phone, because that is what we are talking about, maybe they can use an older model chip (less expensive) and put in a kick ass graphics chip. It all depends how they engineer the cell phone for maximum output. At this point in the game, I definitely doubt it all depends on chip size.
Then we have the case of the supercomputer, which should use the smallest chips possible. But the quantum computer is already here. It is kind of obvious of trying to get smaller logic chips or go quantum has more potential.
The engineers cannot get away from the cost effectiveness of the widget. That would be like asking the economist to forget about supply and demand and equilibrium or the accountant to forget about debits and credits or the woke people to forget about ...
Oh ... never mind ...
TSMC is in a bind with their delayed 3NM chip as Apple and other customers are waiting for the delivery as promise, while Samsung FAB only major customer is itself being an IDM so the pressure is less and may focus on solving its 3NM chip problem using an advance method GAA architecture. With previous planned production of 50,000 pcs per month this year and 105,000 pcs in 2023 being moved a year, the investment cost is huge and the ROI may take a long time to recover unless the customer is willing to pay a huge price increase. We may see a Apple 14 costing $2,000, at present ASML EUVL is not up to the job cost wise (need double exposure) and may wait until 2026 for a new lithographic machine that will achieved a single exposure and that machine will cost double ,twice the current amount. I think 5nm is alright for China coupled with 3D stacking Chiplet tech it may achieve the same performance as a 3nm chip.
247 views15 minutes ago
247 views15 minutes ago
They mention EUVL Double exposure and not double patterning.
Double exposure might not be that costly compared to Double patterning ( which involves realigning).
Edit : a correction regarding costs. the latest apple chip cost 100 dollars. Expect it to be 150 to 200 dollars. Displays will costs 100 dollars.
Apple will sell the iPhone for much more. Obvious.
Last edited:
Nokia left O-RAN because it didn't want to anger China. AFAIK in the recent bid awards, Nokia won 10% of the contracts in China.
Plus, Software is bs. Hardware is the king. Can you imagine basing your whole digital economy upon a mess of software with a million lines of code, full of vulnerabilities etc?
Cant wait for the US to adopt O-RAN and Chinese hackers having a field day penetrating American networks
Well if they go the software route Biden better increase that infrastructure bill to 5 trillion. Need a lot of data centers to make that work with a lot of edge nodes. So i get why US big tech wants to go the software route it will costs the American people a lot of tax money.
..they say they didn't want to run afoul of US penalties since there are Chinese companies like Inspur on the entity list that are contributors in the O-RAN alliance .It doesnt make sense since there are so many American companies otherwise all should have left .
How Nokia ,Ericsson and ZTE are contributing to an alliance whose sole purpose is to kill their business models is beyond me
@tokenanalyst @Topazchen @FairAndUnbiased @krautmeister and others, bro the video stated that 12-inch wafers FAB is where the major profit can be made, if that is the case then why nobody is planning to go to 18 inch wafer? Can China explore this possibility as major FABS shift from 6 inch , 8 inch to 12 inch to match or overtake the more established players? I'm all ears guys regarding your opinion?
CC
1K views1 hour ago
CC
- Status