Chinese semiconductor thread II

[gelgoog]

looks like revenue flat at 55/65nm node, but down everywhere else
That's why HH needs to move up in node. The really legacy nodes are in a bloodbath

A lot of this production at the Hua Hong Wuxi fab was expected to bank on sales of display driver chips. But it competes in that with Nexchip. I don't know what is Nexchip's current production, but they were ramping up really fast. Over a year ago they announced they had hit 100,000 wpm. And that is with a fully modern fab with 300mm diameter wafers. Nexchip is located in Hefei right next to BOE. Nexchip were initially doing display driver chips which require lower resolution processes for lower resolution displays. But they have been developing capabilities to do improved ones all the time as well.

If you think it is a blood bath for Hua Hong, the South Korean display driver industry is basically in total meltdown mode right now. This after the South Korean government prevented the takeover of one of their display driver chip companies by Chinese commercial interests not that long ago. Well, now they will just face a total wipeout.

Remember that SMIC is building 340,000 wpm of 28nm capacity. This is basically as much capacity as TSMC has. At the same time Nexchip is moving to higher resolutions as well. They have 150mm-55nm products in volume production and are working on 40nm and 28nm display driver chips as well.
Thus Hua Hong moving to 28nm would not be a panacea. They are being squeezed in between Nexchip and SMIC. But yeah I think they could further increase production of 28nm. They could build like a 100,000 wpm 28nm fab. And convert their existing 28nm fab to FinFET. I think they really need to get their FinFET project back on track. Still, given the current sanctions on semiconductor equipment they kind of lost their window of opportunity. But if SMEE does get their immersion lithography equipment available this might be possible to ramp up again.

DDIC/TDDI/AMOLED drivers all doing well. Guess what, domestic display suppliers are gaining and fabbing at SMIC
again, fully booked in 28-55nm nodes. a lot of demand for SMIC in these nodes.

SMIC is doing well in display drivers because it is basically doing the nodes which Nexchip cannot produce. I doubt the 55nm node will be successful for long since Nexchip also can produce that in volume already. As for 40nm and 28nm that is used in higher resolution displays it will last for longer as a client base but Nexchip will eventually enter that segment as well.

 

[gelgoog]

Please, Log in or Register to view URLs content!

"4 January 2024
...
Based in Hefei, Anhui Province, Nexchip was founded in 2015 as a joint venture between Taiwan-based Powerchip Technology Corp. and Hefei City Construction Investment Group and was also the first 300mm (12-inch) wafer foundry in the province.
...
According to Nexchip, its 150nm and 55nm nodes are already in volume production, and the R&D on 40nm and 28nm platforms is underway. However, due to the industrial downturn, the company's 2023 half-year report revealed CNY2.97 billion in revenue, a year-on-year decline of 50.4%. Among them, DDIC accounted for 87.84% of sales, followed by CMOS Image Sensors (4.08%), PMIC (5.77%), and MCU (1.35%). In terms of process nodes, Nexchip's 90nm platform accounted for 49.92% of sales, followed by 110nm (31.65%), 150nm (13.6%), and 55nm (4.83%).

As pointed out by the company, overconcentration on the DDIC sector could be a potential risk, subjecting Nexchip's performance to order fluctuation and its bargaining power. Therefore, it becomes necessary to further diversify into CIS, MCU, and PMIC platforms. The five largest customers of Nexchip, in particular, accounted for 66.47% of total sales, rendering Nexchip overdependent on the performances of these major customers."

 

[AndrewS]

SMIC is doing well in display drivers because it is basically doing the nodes which Nexchip cannot produce. I doubt the 55nm node will be successful for long since Nexchip also can produce that in volume already. As for 40nm and 28nm that is used in higher resolution displays it will last for longer as a client base but Nexchip will eventually enter that segment as well.

I'm assuming 55nm requires immersion DUV?

Also, if Nexchip start entering 55nm, 40nm and 28nm display driver field where SMIC is currently, wouldn't it be logical for SMIC to redeploy its immersion DUV machines to logic nodes <14nm?

 

[gelgoog]

I'm assuming 55nm requires immersion DUV?

Not necessarily. You can get up to 32nm with dry lithography using double patterning. But it was originally used to do 90/65nm in one pass I think.

Also, if Nexchip start entering 55nm, 40nm and 28nm display driver field where SMIC is currently, wouldn't it be logical for SMIC to redeploy its immersion DUV machines to logic nodes <14nm?

The thing is those processes require a lot more lithography machines to achieve the same wafer output. So there is a tradeoff. SMIC also probably has much better processes to manufacture, for example, MCUs than Nexchip does. So they could just reallocate production to other kinds of chips.

China today has several MCU vendors like Gigadevice. They have FPGA vendors like GOWIN. These are just a couple of examples. SMIC could try to get these kinds of customers.

 

[Clark Gap]

i don't believe Havok said 20 DUV, he said 20 DUV/DUVi total. I think we are far from knowing how many are actually getting put into production this year, but it's safe to say the production is ramping up now that Guowang, U-Precision are up supplying a lot more to SMEE.

anyhow, SMIC results

Please, Log in or Register to view URLs content!

Please, Log in or Register to view URLs content!

bright spot domestically are the phone suppliers, at least for SMIC
what does it mean when CIS or ISP are up 60% YoY? Well, Omnivision has increased it's share big time. That's what this means


DDIC/TDDI/AMOLED drivers all doing well. Guess what, domestic display suppliers are gaining and fabbing at SMIC

again, fully booked in 28-55nm nodes. a lot of demand for SMIC in these nodes.

You posted the news about SMIC N+3 process on X. Does it equivalent to TSMC N7+?

 

[measuredingabens]

Steep slope VTFET 2D material transistor research. This and other 2D material FETs are the expected successors to CMOS transistors past the next decade.

Please, Log in or Register to view URLs content!

 

[pbd456]

Has havok stopped posting for almost 1 year now?

 

[generalmeng]

Imagine a nation that has spent two decades trying to make an ordinary gas-powered car but keeps failing, making a barely functional product.

Then imagine that that nation suddenly tries to develop a cutting edge state of the art electric vehicle in half that amount of time. Not likel-wait... Nevermind...

And this is an aside to how wrong and oversimplified your analogy was in the first place. China is renowned for making slow progress until forced by Western sanctions, causing them to explode forth with decades of advances in years.

Every major Ice auto maker had about 100 years to perfect their engine. Ev is a level playing ground, China is already a head.

 

[Odin]

If I look at what Tinrobert has written, you would need 10 Immersion DUV for monthly capacity of 50K wafers.
This is with 300mm wafers @ 28nm with a single pattern.
That sounds like the minimum size you would want for a 300mm fab. But ideally, it looks like a capacity of 100K is optimal.

---
As for Dry DUV, my understanding is that ASML is still selling these without restriction.
So there isn't an urgent need to replace these.
Presumably you would also need 10 machines for a nameplate capacity of 50K wafers, but of course, this is at a larger node.

---

And given that ASML operates with an average gross profit margin of around 100% (selling for double what it costs), there is a lot of space for a Chinese competitor. Even if the initial Chinese equipment from SMEE is worse and more expensive, there's still enough margin to be profitable.

On the flip side, it implies that ASML has a sizeable war chest and they may opt to cut margins to compete for international market share. The domestic Chinese market is where SMEE can make some interesting moves.

 

[manqiangrexue]

Every major Ice auto maker had about 100 years to perfect their engine. Ev is a level playing ground, China is already a head.

LOLOL I know that. That was to show how wrong Tacoburger's analogy was, especially when applied to China.