CEO of Intel: Is this really the case that China's core technology is 10 years behind the world's top level?

At the recent World Economic Forum in Davos, the CEO of a major chip factory in the United States publicly stated that under the joint restrictions of the United States, Japan and the Netherlands, the technological gap between China and the world's top wafer factories is 10 years.

I am not convinced by his statement. I don't know how this guy's so-called 10-year gap was calculated.

First of all, TSMC's 7nm process was mass-produced in 2018, and Apple's A12, Huawei Kirin 980, and Qualcomm Snapdragon 855 all use this process.

So in order to better tell you about today's video,When I encounter key places later,I'll use S instead.,I hope to understand more.,I also believe everyone can understand.。

With the release of Huawei Mate60 in 2023, the chip S has also been born, and the recent hot Nova12 has begun to be supplied in batches, which shows that this chip has already achieved large-scale mass production. And, because of the excellent performance in terms of performance and the fact that we don't have EUV in our hands, it is likely to be a 7nm process.

Of course, we can't talk nonsense, we have to have a basis.

According to a report by TechInsights, a well-known research institution, after electron microscopy scanning, its analysis team believes that S has 7nm characteristics.

In addition, it is also reported that the transistor density of the S has reached 1 million per square millimeter, which exceeds TSMC's N7P process and is between N7P and N6.

According to a report by TechInsights, the area of the S is about 107 square millimeters, and the number of transistors in the S is about 10.7 billion.

For the above data on density, I haven't found the original source or authoritative media reports for the time being, but I found a video of a great god and picked the key points to tell you.

Specifically, first, the five cores of B1, A1, A2, A3, and A4 on the chip were found for analysis.

5处密度分别是:B1的111.64Mmtr(HD Sram),A1的82.19Mmtr(HP Logic),A2的77.09Mmtr(HP Sram),A3的119.61Mmtr(HD Sram),A4的115.992Mmtr(HD Logic)。 那么平均下来的密度大概为HP 80mtr,HD 115mtr。

It may not be easy to understand, in fact, the transistor density is different in different regions, so the final combination of area and other parameters can be calculated, you can get a total of 11,402.87 million transistors in S, and then considering the error of 5% up and down, the number of transistors in S is about 11.973-10.832 billion. This figure is still relatively close to the above 10.7 billion.

So, we already have 7nm capability. So let's compare it like this, in 2018 and 2023, the gap is 5 years. But don't forget, this gap was made with all the power to suppress it, and if there were no restrictions, I believe that the gap would be narrowed to the limit.

In addition, TSMC's 5nm process will be mass-produced in 2020. Previously, a well-known blogger said that our EUV has been debugged online, and it is 3nm, and it will be mass-produced in the next 2 years. If it is 2025, it will still be 5 years compared to 2020. However, it was also achieved with full force. Of course, we must admit that such statements are naturally unverifiable. However, some netizens said that the blogger's revelations have always been quite accurate. However, this evaluation cannot be used as a criterion. In short, I hope that what he said will come true as soon as possible.

However, the development of the semiconductor industry actually has inertia, that is, once it breaks through in a certain link, it seems to be completely opened, and rapid iteration will lead to a full flowering of the industrial chain. Just like TSMC's 7nm is 2018, it will be 5nm after only 2 years. Of course, they have EUV, we don't have it yet, and admittedly, it's going to be harder for us to iterate.

In addition, the 10-year gap mentioned by the previous executive may have been speculated by him according to the development route of the West, which absolutely does not apply to us in China. For example, CCTV previously reported that the vice chairman of TechInsights believes that the chip of the Mate 60 Pro is still 2-2.5 nodes away from the most advanced technology. In this regard, Lu Tingjie, a professor at Beijing University of Posts and Telecommunications and executive vice president of the China Information Economics Society, said that this means that there is still a gap of 3 to 5 years between us and the advanced process, but this is judged by Western countries with their technological progress speed, and we can often use China's speed to complete the surpass.

So why is this? Let's put aside the technical level, to put it simply, the office building of Dachang H, every night, is brightly lit, and everyone is working overtime. Can the West do this alone?

In other words, doing the same thing, in Chinese, there is always a miracle. For example, in the domestic memory chip manufacturers, hasn't it already been staged?

In short, after the large-scale commercial application of S, many people may be a little panicked, and they have also started the old routine - dumping strategy, that is, price reduction. For example, TSMC has reduced the price of mature processes above 28nm by 2-5%, while the price reduction of 7nm processes is even greater, at about 5%-10%.

This move also shows that the pressure on TSMC is increasing.

On the one hand, when I was asked to build a factory in the United States, one would not work, and I had to build two. Tens of billions of dollars have been invested, but subsidies have not been put in place, and the financial pressure can be imagined. In addition, during the construction of the factory, there were many troubles. What's more, the cost of building a factory in the United States is high, including the operating costs in the later stage, and even if it is mass-produced in the future, the chips produced may not have any advantages in terms of price. In particular, it is very likely that it will face the impact of chips from the mainland in the future.

On the other hand, in 2023, TSMC's revenue will be about NT$2,161.74 billion, equivalent to RMB 492.66 billion, a decrease of 4.5% compared with 2022, and in US dollar terms, revenue will be US$69.298 billion, down more year-on-year, reaching 8.7%. The net profit margin was 38.8%, down 6.1% from 44.9% in 2022.

In recent years, the performance that has been soaring all the way has suddenly stopped, which is somewhat unexpected. The same decline in TSMC's performance is the decline in the scale of China's chip imports for two consecutive years, which fell by 15.4% this year to $349.4 billion.

So, why is this happening? Behind it, of course, is the rise of China's semiconductor industry chain.

Before 2018, China's local chip self-sufficiency rate was about 5%. According to IC Insights, the self-sufficiency rate of China's local chips will be about 16.6% in 2020, about 17.6% in 2021, and about 18.3% in 2022.

In 2023, there are two sets of data, one is given by IBS, an internationally renowned research institution, with a self-sufficiency rate of 25.61%. The other is 23.3% given by TechInsights. Although there is a slight gap between the two, it is not much, and it may be related to statistical methods.

But these are enough to show that in 5 years, our self-sufficiency rate has been significantly improved, and the production capacity of domestic chips continues to rise.

Just like Huawei's mobile phone, using one more Kirin means that one less one can be used, and it also means that TSMC may have to produce one less chip, and even all links in the global semiconductor industry chain will be implicated. On the other hand, the use of one more Kirin means that the domestic semiconductor industry chain needs to produce one more chip, which is too big and will promote the further development of the entire industry.

And, once we achieve a real breakthrough in high-end technology one day, coupled with the development and growth of the underlying operating system HarmonyOS, Kirin may be able to open supply, and domestic manufacturers will have more choices, even if they do not use Kirin, they will also have more say in the procurement of other chips. The purchase price of foreign chips is cheaper, the cost of the product is low, and the selling price will also decrease, and in the end, we, as ordinary users, will also benefit. Isn't that bad, you say?

To be honest, that scene must have been beautiful, which I really didn't dare to think about before, but now, everyone can really be whimsical.