Chinese semiconductor industry

hvpc said:

That's a very surprising result given that Mate60 sale has been supposedly so good in September. The revenue barely ticked up at all in Q3. The tick up in net profit could be partially attributed to balance from sale of Honor three years ago.

I was expecting better financial result from Huawei.

Click to expand...

antiterror13 said:

456 B Yuan is roughly US$62B or if extrapolated to 1 year, it would be US$83B .. don't know, somehow I thought it was more than US$100B. The profit is good 73.05 B yuan or roughly US$10B

Please, Log in or Register to view URLs content!

.

Click to expand...

Mate60 is not the primary driver of Huawei's revenue .. Q4 revenue will tell us the real story.

However, the Huawei spokesperson attributed the third-quarter revenue growth to increases in the digital power, cloud and auto parts businesses, while the Mate 60 series was released relatively late in the quarter.

hvpc said:

That's a very surprising result given that Mate60 sale has been supposedly so good in September. The revenue barely ticked up at all in Q3. The tick up in net profit could be partially attributed to balance from sale of Honor three years ago.

I was expecting better financial result from Huawei.

Click to expand...

huawei only made it back to top of phone sales chart in October. i think the initial demand surprised them. Also more importantly, a lot of consumer electronics only got unveiled on Sep 25th. For example, AITO M7 sales are way up in October vs September.

M60 brought in a lot of extra foot traffic to Huawei stores since Sep 25th, likely to show up in Q4.

But more likely next year as new phones get announced and the production ramps up.

Not easy to go from 7million phone a quarter to 15million phones a quarter. The entire supply chain needs to raise production level.

sunnymaxi said:

Mate60 is not the primary driver of Huawei's revenue .. Q4 revenue will tell us the real story.

However, the Huawei spokesperson attributed the third-quarter revenue growth to increases in the digital power, cloud and auto parts businesses, while the Mate 60 series was released relatively late in the quarter.

Click to expand...

well, in the past, phone sales was 190m in peak years. That's close to $100B in revenue a year.

If they sell 35m Mate and P series phones in 2024, that's $35B alone. So yeah, it should be the primary driver of huawei's revenue increase in 2024.

And of course, there are also MatePad sales, which should go up and probably new harmony laptop sales.

Assuming of course, they can make enough SoCs, which is a big question now.

hvpc said:

The industry look at frontend (FE) and backend (BE) litho as distinct products.

Backend steppers could be i/g/h line, with magnification of 1X or 2X, and like you mentioned, larger field size.

FE are typically 4X magnification scanners; BE are typically i/g/h-line steppers with 1X or 2X magnification

FE litho focus is on resolution so targets larger NA projectio lens; BE needs larger Depth of Focus (DoF) so needs larger NA. BE litho supports thicker substrates up to 2mm.

Click to expand...

I just noticed a typo in an old post. It should’ve said, BE steppers need smaller NA objective lens to handle need for a much larger DoF because topography of wafer that goes into BE steppers have much larger variation.

hvpc said:

0.35um technology node FE appliction uses FE iline steppers with 5X magnification. Backend iLine stepper is not capable of supporting any FE litho needs.

SMEE SSB500/40 & SSB500/50, their primary backend litho tools are 1X i/g/h-line stepper with resolution limit of 2um & 1um; Single Machine Overlay (SMO) of 0.3um & MMO of 0.6um. DoF spec is 8um for 1um feature size & 30um for 3um features. and have very slow throughput <100 wafers-per-hour.

BE steppers are not the same and not in the same class as a high-precision, highly efficient FE scanners, and this shows in the unit price of these machines. So no matter you arguing the technical merit that both are litho tools, they are not the same and viewed as seperate and distinct types of machines.

Click to expand...

tonyget said:

You don't need the best AI chip to train LLM，but using the best chip is the most efficient way to do the work

Click to expand...

The "At what Cost?"-ing exercise works favorable in terms of energy costs.

Electricity cost per unit kWh ():

• U.S.A: 12.5 cents per kilo-watt hour (kWh)
• China: 4.3 cents per kilo-watt hour

That represents a 70% less electricity cost in China vs. US, due to abundance of cheap hydroelectricity power, and as low as 92% less if you go to Sichuan hydroelectric dams with 1 cent per kWh, which became haven for GPU Bitcoin mining farms.

An 5nm chip is 15% performance gain vs. 7nm chip at same power draw, or 30% reduced power consumption at the same performance. These are vastly outweigh by lower electricity cost in China compared with U.S. We are talking about a big gap even if you go 3nm chips.

Notwithstanding that at one point, US market electricity prices increased considerably due to Ukraine war sanctions...my electricity bill increased considerably.

Edit: clarified post

Average prices are deceptive.

Realized those numbers are forecast but not able to edit the post anymore. The non-forecast numbers from :

• Between January 2022 and January 2023, the average electricity price for US homeowners increased from 13.72 to 15.47 cents/kWh.
• This represents a 12.76% increase in kWh prices, which is twice the general inflation rate during the same 12-month period: a 6.4% increase in the
  Please, Log in or Register to view URLs content!
  .

Considerable increase? Yes. “skyrocketing”? No. Doubled or tripled? Hell no.

Not to derail thread, I clarified my original post. Tphuang can clean up this post.

Phead128 said:

The "At what Cost?"-ing exercise works favorable in terms of energy costs.

Electricity cost per unit kWh (

Please, Log in or Register to view URLs content!

):

• U.S.A: 12.5 cents per kilo-watt hour (kWh)
• China: 4.3 cents per kilo-watt hour

That represents a 70% less electricity cost in China vs. US, due to abundance of cheap hydroelectricity power, and as low as 92% less if you go to Sichuan hydroelectric dams with 1 cent per kWh, which became haven for GPU Bitcoin mining farms.

An 5nm chip is 15% performance gain vs. 7nm chip at same power draw, or 30% reduced power consumption at the same performance. These are vastly outweigh by lower electricity cost in China compared with U.S. We are talking about a big gap even if you go 3nm chips.

Notwithstanding that US market electricity prices increased 200% - 300% recently due to Ukraine war sanctions...my electricity bill skyrocketed in past year.

Click to expand...

could you explain why you compare 7nm to 5nm chips?

hvpc said:

could you explain why you compare 7nm to 5nm chips?

Click to expand...

IMO he is implying that China's lower electricity prices can offset for 2-3 nodes in disparity when it comes to HPC. I can't readily disagree with him. If differences between nodes are small more racks can be a viable solution.