Chinese semiconductor industry

Status
Not open for further replies.
9dashline

9dashline

Captain
Registered Member
Overbom said:
Why even bother when big brother, US, will harvest it anyway lol
View attachment 78361
Click to expand...

In the Chinese econ thread Anlsvrthng was stating how its humiliating that China metrics don't hold candle compared to American productivity... " Chinese work more hours a week, and receive less of the result of they work than the USA/european counterpart." he doesn't say is that a huge chuck of that is because of dividends from USD still being the reserve currency and all the compounding harvests America has reaped over the decades that makes it not even comparable... But when Anglo colonizers continually overharvest, eventually they end up killing the buffalos, just ask the native Americans.

In any case, AI is the final frontier and the rate limiting factor is no longer population nor that of human laborers...
 
T

Topazchen

Junior Member
Registered Member
supersnoop said:
I'm flattered that I give the impression of being an insider, however I do not work in the industry. I just like to read and have some basic technical and business knowledge. In fact, I would say you seem better informed than me.

If you still want to hear my opinion, here it is
If we are talking about overall capacity, then yes there is a need for more fab capacity. Since smartphones/IOT is a continually growing segment, you see TSMC is concentrating on the sub 7nm (also the most profitable).

However, the problems are more business than technical (sure they have been mentioned previously).
- 2024 initial production is very aggressive. How many construction projects get finished in Western countries in 2 years? One of the limitations is simply labour supply. If you need some kind of skilled construction labour, there is not enough.

- Supply location is far from customers. As mentioned, Globalfoundries is a big money loser. However, their process technology is quite advanced. One of the few foundries with FinFET and SOI capability. The issue becomes it doesn't make a lot of sense to ship chips from Western New York and Germany back to Asia for packaging. Actually Intel is doing this, but at least they are on the west coast and they had a captive market for a long time. Now that AMD's performance is competitive again, you see that their margins are threatened.

As a corollary to the above:
- Who are the customers for 5nm in the USA? 5nm by 2024 will not be the leading edge process for TSMC if things are going according to schedule. The risk is ending up with the same problem as Globalfoundries. If you are not the best, then the margin becomes lower, lower margin products are going to favor being close to Asia (To put it another way, MediaTek is not going to pay those shipping costs).

- Finally, to the best of my knowledge, the supply constraints are primarily because TSMC has been very good at supporting their customers on leading (or close to leading) edge processes in addition to actually having the best process. As a result, everyone is fighting to get their high-end chips made at TSMC (which also means TSMC has less capacity for low profit chips like 28nm)

nVidia and Qualcomm are using Samsung 8nm in addition to TSMC and there have been many reports of overheating because Samsung 8nm is simply inferior. 8nm is just the smallest feature size, if other parts of the process are not engineered well, then "Xnm" is just meaningless marketing. This was also the problem with Intel's first 10nm process (I think it was used on Cannon Lake, eventually the whole process design was thrown out). Intel is finally able to fab 10nm at volume, but only lower power chips right now (15W Laptop). Overall, Intel 10nm is probably still superior to Samsung despite the naming, which is probably why they recently announced a name change. Really, Intel is only one step behind. However, since the TSMC-AZ fab will be one step behind, then Intel and TSMC would be equals in stateside manufacturing (as long as Intel doesn't suffer further delays). As such, the issue would become less technical and more about customer support.

This was a lot to go over, so let's cut to the chase. Is the AZ Fab a Potemkin village? If you consider from a corporate standpoint, then it is really a waste of money. Is it viable? Yes, probably, but it would not be maximizing the ROI.

Ironically, Cold War ideology would favour Control economy (Communism) over Capitalism. You have empty car lots and store shelves all over the country to fight the trade war with China. It is like the old Soviet system where it was difficult to get basic consumer goods like cars, bicycles, and TVs. National champion system is inefficient and prone to corruption, otherwise everyone in Malaysia would be driving Protons.

I don't think the US is wrong to ban Huawei backend equipment domestically, USA is the expert in telecom snooping, they know what the possibilities are. Where they went wrong was their attempt to destroy the company. I get why they did it, it is like the economic version of "Shock and Awe". How well did that work out though?
Click to expand...
They are 'killing' the company rhetorically and in public but in private, they are milking it.
See the latest Reuters scoop
https://twitter.com/i/web/status/1451273781486538752
 
O

Overbom

Brigadier
Registered Member
horse

horse

Colonel
Registered Member
Topazchen said:
They are 'killing' the company rhetorically and in public but in private, they are milking it.
See the latest Reuters scoop
https://twitter.com/i/web/status/1451273781486538752
Click to expand...

The numbers quoted in the article are on the fake side.

IIRC the previous year before the pandemic, Huawei had $109 USD billion in revenue and the ordering budget of like $90-$100 billion for procurement of supplies.

What I do remember is that Huawei bought $11 billion worth of stuff from companies in the United States. Certainly that is a lot lower now with the restrictions, and likely never to reach those levels again.

As for SMIC, I think they only do about $5 billion USD in revenue. The Reuters article claims there are licenses granted to sell SMIC about $42 billion dollars worth of stuff from the US.

The only way those numbers, for actual goods sold with the licenses, can be true is if they use Monopoly money.

:D
 
ZeEa5KPul

ZeEa5KPul

Colonel
Registered Member
dfrtyhgj said:
That's just cope.
Click to expand...
It doesn't even make sense as a cope. If a license has been granted but not used, that's the customer's choice not to use it. They're implicitly admitting that Huawei and SMIC are cutting their reliance on US companies so well that they don't even use the licenses they've got to their full capacity.

The authors of this piece are too stupid to understand the own goal they just scored.
 
Status
Not open for further replies.
Top