I don't think it's just the cost, but just the idea of competing with NOCs and IOCs to build an oil refinery is incredibly stupid
Chinese semiconductor industry
- Thread starter Hendrik_2000
- Start date
- Status
Probably because there never really was any pressure to create third or fourth player.
Now that there is a national security pressure for domestic software maybe not even for China alone but for other core SCO or BRI nations too.
You left out that Unisoc's chips are made at TSMC and thus their success is due to US permission. HiSilicon dropped 77%, Qualcomm's shipments grew 32%. Its stock is up 7% this morning as its earnings and revenue beat expectations.
There is not going around TSMC for now. That's why I earlier dismissed Alibaba's chip.
They are all helpful developments, but the biggest obstacles is domestic manufacturing of advanced nodes
Hard disagree. IC design is very important and if Chinese companies are able to design cutting-edge products, then it is great and very useful for China.
What's the point of having 5 nm capabilities if you cannot design anything and would end up as a factory for American companies (AMD, Intel, NVIDIA, etc.)? Another thing is that high-tech fabs are the epitome of "economics of scale" - if there's not enough demand, then those fabs will be essentially useless because the production costs would be too high compared to the competitors. That's why the fact that Huawei, Alibaba, and others are catching up in IC design & producing world-class products is extremely important if China wants to have a localized semiconductor supply chain. The absence of demand for high-tech nodes is precisely the reason why I'm very skeptical on EU, India, Japan trying to localize the supply chains on their shores as well (mimicking moves by the US and China).
Partly agree and partly disagree. For sure there is no black magic that makes EDA too complex to catch up or some bullshit like this, but it won't be easy either. I think it will take years before Chinese EDA companies will be able to provide the same speed, precision, or whatever when compared to the companies that have been long in the business. He is not exactly wrong about the perks of "built-up" knowledge.
It is true that software is software but experience matters a lot in certain branches - you won't see random EDA, computational optimization, etc. startups precisely because of this. For example, I used to work in one of the companies that produce the current SOTA optimization software (think - Gurobi, CPLEX, etc.) and while even a student can produce a simple optimizer, it takes thousands of hours of work (not only of programmers but also mathematicians) to fine-tune and optimize the software to be as fast and as accurate as possible. It is especially important when we move to mixed-integer optimization or non-linear problems - you can take an open-source solver like COIN-OR ones (or GLPK for linear cases) and compare them to the proprietary ones like Gurobi. The former is literally a hundred times slower and often cannot even produce the solution for more complex problems which are easily solved by the likes of Gurobi. The reason is that there are a lot of unobvious mathematical and programming tricks that are applied to achieve that result, while a good chunk of them are "proprietary" - i.e. you cannot just Google them. Plus the number of computational optimization specialists and of people who are adept in "squeezing out" the performance in software is way lower than the general software developer population, where the majority are mostly doing very mundane stuff. I admit that I do not know much about the EDA sphere but I think the same thing should apply to that sphere, otherwise it would not be essentially monopolized by the likes of Synopsis.
There are also some pros that could be utilized by Chinese EDA companies if they are just starting out - no need to worry about some old legacy and backward compatibility shit that often severely restricts the development process & drains resources. Use good industry practices and architectures from the start, optimize as much as possible, and all that without worrying about potentially breaking some 1.X.X legacy crap.
If still using TSMC fab, it changes nothing. Look at what happened to Hisilicon. US can sanction anytime they want. What China needs now are manufacturing capabilities.
True, but that's all a matter of degree - not of have or not have. With existing EDA software companies like Empyrean there exists a sufficient foundation for improvement with internal experience alone.
The logic or rational of the least financial cost doesn't come into operation or domination here, the United States has clearly already demonstrated its capability in preventing others from selling what they'd like to sell to China. That being the case China cannot leave itself at the mercy of foreign supply chains and must develop the entire domestic supply chains. That is what it is in the process of doing. It does not matter how much anyone in China or elsewhere complains about the US messing global supply chains and adding to costs and its actions contributing to shortages of chips and other goods, the US will continue its actions as there is full bipartisan support among its elites to constrain China.I do not agree completely.
What makes the IC worldwide supply chain unique is its interconnections.
Through this web, the Americans believe they can exert influence, such as a ban on chips sold to Chinese companies.
However, through this same web, the Chinese exert a lot of power too.
The most advanced chips are still sold to China. Given the astronomical R&D put into the technical know-how to produce such a chip, those monies will never be recouped if the China market is off limits. China is the biggest market for chips. Do not sell this 5nm chip to China, then the entire IC current model and supply chain will start breaking down.
The most cutting edge chips make the highest profits. Cut those sales, and the pain will reverberate through the entire industry.
This interconnected web of IC supply chains and the end product, it works both ways, and all ways.
The Americans do not view the world that way. That is why I labelled the article American centric. That is why Afghanistan collapsed in 3 weeks. The only people truly surprised where the Americans.
If the Americans start banning 5nm chips to China, a blanket ban will hurt the entire industry, including American companies. Then the Chinese could bring the rare earth card onto the table. Along with other restrictions (formal or informal) on US companies operating in China. American companies made a lot of money in China in the past. They better smarten up if they still want to do the same in the future.
Without the China market, can the most cutting edge chips be commercially viable? Probably not. No one wants to find out. Bluff city.
The Americans are not strategic geniuses. That is why Afghanistan collapsed in 3 weeks.
What chance do they really have here in IC against the Chinese in a few years? What is going to be left?
- Status