Chinese Economics Thread

[2handedswordsman]

What I don't understand is why has there not been a major stimulus in China for tech?

• China has surplus tech talent, some of which is unable to find jobs, or is under/suboptimally employed.
• The wage growth is limited and there's need to expand that.
• There's major tech containment going on, and China needs urgent breakthroughs in dozens of technologies (EUV, Semi Equipment, chips, GPUs, materials, space launch reusability, megaconstellations etc. etc.)
• Major deflation as a problem, need for inflation.

All of these are addressed if there's a big tech stimulus that is undertaken.

Will absorb talent, lead to wage growth, lead to greater demand stimulus, and result in higher rate of tech breakthroughs.

What explains this reticence though?

The leadership is acting like fiscal conservative Germans, however as Germans will themselves tell you, too much fiscal conservativeness is suboptimal. Fiat Money is not something real, so personal finance advice being applied to fiat money at a country level seems puzzling.

Why spend money when the market itself says that it is saturated? Better experiment on other fields. Like encourage people to work on lower skilled jobs. Better cultivating grapes and send them by an app than working 996 on bigtech like it's somekind of a fetish

 

[Wrought]

That is a plan to deploy that some over many years. And 1 trillion yuan, is what 142 billion usd over I believe many many years. I actually think China can do a LOT more.



I am not arguing there's no financial support. I am saying it is insufficient specially when the government can spend a LOT more. How much is the government spending on all of these projects? The combined R&D budget of the economy is just around 500 billion usd.

Strategically, I would say the government should spend 200 billion usd more every year, and do it in discrete (without talking about it). It is very easily doable, and would generate many jobs and have a good effect.

1. We have no concrete idea where Chinese EUV is currently. Granted, it is a very tightly controlled project, but this means that apart from some whispers, we don't know where the project really is. And the whispers in China tech are not as reliable as China military, where China military whisperers have proven track record over decades. Case in point, half of this forum was expecting 5nm chip in Huawei phones last year, but that was not the case.

2. The money spent by private Chinese SME firms on R&D is far less than what western companies spend even today. And it is the Chinese companies that need to catch up. Granted money probably stretches further in China, but I am talking about stimulus kind of somes. Maybe I am just more ambitious.

2(a). Let's talk about EDA. Chinese firms are still heavily reliant on Synopsys, and others. Why is that? Despite getting a shock in 2018 with EDA controls on Huawei, the US was still able to play the EDA card this April again. I would have liked if they spent hundreds of billions just on EDA. In fact, most of the talent in these EDA firms is also Chinese. EDA will be solved in literally few years, if there is a huge stimulus.

3. Reusability is in fact the exact example that illustrates this. That China is running a decade behind SpaceX should be alarming. It is only now that these companies are getting serious money, but even this is not that high. Government should probably spend 10s of billions every year to first gain advantage. And space was one sector where China was gaining and getting on par with US, then SpaceX just pulled through. Government needs to fund risky, moonshot projects like SpaceX. Hundreds of them.

4. Look at aircrafts now. C919 has been missing its delivery targets for 3-4 years now. I am just saying to accelerate. Talent is present, you need money now. Accelerate.



How is that believable when China is still importing upto 50% of the equipment that these western firms supply.

Perhaps I am too ambitious, but I think China is in a position to, and should be spending 200 billion usd of additional funding (every year) for the next 10 years. All of the challenges will then gradually dissipate.

This is American-level finance brain. Developing this kind of sophisticated technology simply takes time. No, you can't just wave money around and make physics or chemistry or biology "accelerate" like some magic spell. Nine women can't deliver a baby in one month.

Look at how long it took for the current leaders to get where they are now, and you'll appreciate how fast Chinese efforts are moving.

 

[vincent]

This is American-level finance brain. Developing this kind of sophisticated technology simply takes time. No, you can't just wave money around and make physics or chemistry or biology "accelerate" like some magic spell. Nine women can't deliver a baby in one month.

Look at how long it took for the current leaders to get where they are now, and you'll appreciate how fast Chinese efforts are moving.

I can make nine women deliver nine babies in nine months though.

 

[Blitzo]

I can make nine women deliver nine babies in nine months though.

Jli88's post was about spending more money to accelerate the speed of development, which up to a point will not speed things up because of realities of research and development.

You are talking about spending more money to explore multiple avenues of development... which actually China is doing for most of the important domains of technology where it makes sense to.

 

[jli88]

Jli88's post was about spending more money to accelerate the speed of development, which up to a point will not speed things up because of realities of research and development.

You are talking about spending more money to explore multiple avenues of development... which actually China is doing for most of the important domains of technology where it makes sense to.

I would argue exploring multiple avenues of development is in some ways crucial to speed up research and development. A lot of things are about trial and error, and keep running experiments until optimal conditions or configurations are found.

While I am sure putting more money in will not hasten many processes, I am nevertheless certain that things would still move much faster on various fronts. If nothing, you can hire even more Chinese in other countries to come back and contribute. Getting more money in, also helps in talent generation in niche sectors.

Other than that, my stimulus hypothesis is not purely about tech, it's also a way to absorb surplus tech talent, give a demand boost to economy, and inflate the economy.

 

[vincent]

I would argue exploring multiple avenues of development is in some ways crucial to speed up research and development. A lot of things are about trial and error, and keep running experiments until optimal conditions or configurations are found.

While I am sure putting more money in will not hasten many processes, I am nevertheless certain that things would still move much faster on various fronts. If nothing, you can hire even more Chinese in other countries to come back and contribute. Getting more money in, also helps in talent generation in niche sectors.

Other than that, my stimulus hypothesis is not purely about tech, it's also a way to absorb surplus tech talent, give a demand boost to economy, and inflate the economy.

You don’t work in research and development, do you?

 

[jli88]

Please, Log in or Register to view URLs content!

Huawei has its own FinFet EDA since 2022. most likely SMIC using this already

I just read the piece. Empyrean has a puny market share in China, it itself acknowledges that. If Empyrean was even half as good, the government would have insisted local companies to use it. In fact I don't see a sustained push in EDA at all. Where is the large money flowing into EDA? Only a person working with actual EDA tools might know the answer.

This reminds me of Huawei's ascend 910b chips. Going by this forum, it was already good enough to replace at least until Nvidia's A100 chips. However, I happened to speak some time back with someone who tried Huawei's training stack (CANN,Mindspore) and he told me that the training stack is extremely unstable and almost unusable by lay people (i.e. not directly involved in Huawei's chip and framework development). I am sure things must have improved specially by standardization brought by TileLang and the time Huawei has had since then.

I do think more money will allow to have more R&D resources to both do more things and do them faster.

They NEED to spend at least 200 bn usd+ every year over and above normal R&D budget growth.

You don’t work in research and development, do you?

I actually do. While a lot of R&D is indeed not parallelizable, some significant chunk is.

 

[Blitzo]

I would argue exploring multiple avenues of development is in some ways crucial to speed up research and development. A lot of things are about trial and error, and keep running experiments until optimal conditions or configurations are found.

While I am sure putting more money in will not hasten many processes, I am nevertheless certain that things would still move much faster on various fronts. If nothing, you can hire even more Chinese in other countries to come back and contribute. Getting more money in, also helps in talent generation in niche sectors.

Other than that, my stimulus hypothesis is not purely about tech, it's also a way to absorb surplus tech talent, give a demand boost to economy, and inflate the economy.

I said that China is exploring multiple avenues of development and funding is occurring in relevant domains where multiple avenues of development makes sense. Multiple EUV routes, multiple reusable rocket pursuits, are examples.

However that doesn't mean simply putting in more money will help to actually make things go faster.


In other words, what we are seeing in terms of speed is probably about as good as you'll get, and adding more money will likely not introduce commensurate benefits in speed (i.e.: diminishing returns).

 

[ENTED64]

I am not arguing there's no financial support. I am saying it is insufficient specially when the government can spend a LOT more. How much is the government spending on all of these projects? The combined R&D budget of the economy is just around 500 billion usd.

I do think more money will allow to have more R&D resources to both do more things and do them faster.

They NEED to spend at least 200 bn usd+ every year over and above normal R&D budget growth.

I actually do. While a lot of R&D is indeed not parallelizable, some significant chunk is.

I think the main disagreement here is one of scale not of fundamental ideas. Everybody agrees that the Chinese government should be investing and trying to help these companies along. Everybody agrees that the Chinese government has in fact provided significant investment over a sustained period of time to doing just that. So your argument is mostly a matter of how much investment is the right amount.

Obviously too little investment will slow things but there is surely a sweet spot, if you invest 50% of your GDP into R&D things will go poorly. At some point you've reached the point where further investments are unlikely to produce results faster and may actually be counterproductive as excessive investment that has nowhere to go but is forced to be invested can result in graft, waste, and many unsound companies being created purely for the sake of soaking up the endless investment in this sector. The Chinese government, balancing all its responsibilities, clearly believes that this amount of investment, which is certainly not a trivial amount, is the right balance. While you might argue a bit more is reasonable, I'm skeptical of your assertion that they should double or more the R&D going into this sector.

 

[FairAndUnbiased]

I just read the piece. Empyrean has a puny market share in China, it itself acknowledges that. If Empyrean was even half as good, the government would have insisted local companies to use it. In fact I don't see a sustained push in EDA at all. Where is the large money flowing into EDA? Only a person working with actual EDA tools might know the answer.

This reminds me of Huawei's ascend 910b chips. Going by this forum, it was already good enough to replace at least until Nvidia's A100 chips. However, I happened to speak some time back with someone who tried Huawei's training stack (CANN,Mindspore) and he told me that the training stack is extremely unstable and almost unusable by lay people (i.e. not directly involved in Huawei's chip and framework development). I am sure things must have improved specially by standardization brought by TileLang and the time Huawei has had since then.

I do think more money will allow to have more R&D resources to both do more things and do them faster.

They NEED to spend at least 200 bn usd+ every year over and above normal R&D budget growth.



I actually do. While a lot of R&D is indeed not parallelizable, some significant chunk is.

alot of software tool use is due to institutional inertia. Empyrean is good enough for Samsung.

Please, Log in or Register to view URLs content!

According to the cooperation of two parties, Empyrean's SPICE circuit simulator—

Please, Log in or Register to view URLs content!

had been evaluated under Samsung Foundry's EDA tool certification program (SAFE-QEDA) and confirmed to meet the requirements for certification in Samsung Foundry 14nm and 8nm technology.

Empyrean ALPS is the next-generation high-performance high-accuracy true SPICE simulator. ALPS delivers 100% SPICE accuracy on circuits with over 100M elements. ALPS proprietary Smart Matrix Solving technology provides efficient matrix solving and helps to deliver significant speedup especially in post-layout simulation comparing with other parallel SPICE simulators.

More money would help Empyrean scale up by paying higher wages and hiring more, but really its more about consistent investment. A large, unsustainable surge of funding leads to uncertainty and companies would just want to pocket the short term gains.