Chinese semiconductor thread II

[iewgnem]

This whole thing remind me of 2 years ago when Huawei came out with 7nm Kirin and the west broke their brain convincing themselves yields must be too low to mass produce, then Huawei went on to sell 30 million Mate 60, lol

 

[PopularScience]

How old are these slides.

no idea. slides might be recent, but news might be old.

 

[tamsen_ikard]

Acceptance - to be honest I don't think I've seen this stage be reached yet for any type of technology/system capability

Acceptace stage exists but its usually in the form of "Ofcourse China would be able to do it, the govt gave so much subsidies". Examples include EV, battery and solar panels. Acceptance stage is usually followed by overcapacity stage or security threat stage.

Overcapacity stage usually goes as: "we need to put tariffs or other barriers to stop China from taking over market share in our own market and destroy our companies".

Security threat stage goes as: "we need to sanction and blockade Chinese companies that are too good at something, they are taking over too much market share in the global south and preventing China's domination is a national security goal, otherwize their dominance becomes a security risk". Huawei dominance in 5G was a big example of this. Now they are gunning for DJI and Tiktok.

 

[pbd456]

How many machines will be needed in order to mass produce chips for 2028? This sounds about right timing as N+3 can still be improved!

 

[AndrewS]

Jordan Schneider.... Can't help but feel bad for the man at this point. Being wrong is annoying, but being incredibly wrong is just humiliating.

He still publicly saying that China and the US will be equal with about 25% of global GDP in 10 years, 20 years, and 30 years.
That is just so unlikely.

China is at 30% larger today, and should be about twice the size in 2030, given likely growth trajectories.

---

He also believes AGI is possible within the next few years, and will result in 10% growth for the USA.
But we have Nvidia and Microsoft publicly saying they don't think AGI is possible.

 

[AndrewS]

even when it works, they yield will be bad, and china is losing money making chips.... they said the same with with every Kirin 90*0.
US could have bankrupted Huawei by buying up all Mate phones but they dont dare to do it.

Remember that the chips using the latest processes have huge profit margins built in.

For example, Nividia buys H100 chips from TSMC for $3K, but sells for $25K+

But the cost of TSMC producing each chip is probably about $1.5K

---

Consider that. The actual production cost is $1.5K, but it sells for $25K.

Even if Chinese DUV or EUV machines are less efficient and the chips cost more, they can still be highly profitable.

So all those Mate phones are likely to be profitable

 

[AndrewS]

China planned from a long time ago to make a network of purely domestics fabs to make chips for the government, the military and the space program independently.

In lithography, the goal was to have: I-Line machines for 250nm, KrF machines for 130nm and Dry ArF machines for 65nm, immersion machines for 45nm and EUV machines for 32nm. All Before 2030.

That information is not top secret, was there from before 2015, just because ASML was selling their lithography machine the development of domestic lithography machines including EUV didn't stop or deaccelerated. The goal was not to compete with ASML but to have a alternative, independent, secure supply of chips.

You don't need EUV for 32nm lithography?

 

[Blitzo]

Acceptace stage exists but its usually in the form of "Ofcourse China would be able to do it, the govt gave so much subsidies". Examples include EV, battery and solar panels. Acceptance stage is usually followed by overcapacity stage or security threat stage.

Overcapacity stage usually goes as: "we need to put tariffs or other barriers to stop China from taking over market share in our own market and destroy our companies".

Security threat stage goes as: "we need to sanction and blockade Chinese companies that are too good at something, they are taking over too much market share in the global south and preventing China's domination is a national security goal, otherwize their dominance becomes a security risk". Huawei dominance in 5G was a big example of this. Now they are gunning for DJI and Tiktok.

"Giving too much subsidies" and "overcapacity/security threat" stage by definition would still be part of bargaining.

Acceptance would be knowing that an system/technology/product is real, unable to be changed, and a reflection of actual good and rational policy. To be honest, acceptance is nigh impossible to be reached because by extension would it mean a degree of acceptance of the PRC system.

 

[latenlazy]

"Giving too much subsidies" and "overcapacity/security threat" stage by definition would still be part of bargaining.

Acceptance would be knowing that an system/technology/product is real, unable to be changed, and a reflection of actual good and rational policy. To be honest, acceptance is nigh impossible to be reached because by extension would it mean a degree of acceptance of the PRC system.

Oh my god China is going to make them think there are too many advanced node chips by the end of the decade huh.

 

[gadgetcool5]

China has developed many prototypes over the years. In 2016, they developed a prototype of a 90nm lithography machine. In 2023 they developed a prototype of a 28nm lithography machine. Now they have supposedly developed a prototype of an EUV lithography machine. The problem is that we have never heard proof that any of these prototypes ever made it to the mass production for commercial use stage. The only machines in use for mass production are ASML, and to a lesser extent, Nikon and Canon. At this point, I'd be more impressed with a confirmed report of any Chinese front end lithography machine of any node size in use for mass production on a commercial line, than another prototype of even the most advanced capability.