In the White House, in the room to the left of the Oval Office, the White House Science and Technology Committee is making the last report of his presidency to Obama.
After explaining the contents of the report in detail, the chief scientific and technological adviser, Professor Holdren from the Department of Physics of Harvard University, warned the people, and he emphasized that the development of China's chip technology has threatened the security of the United States.
However, because it was at a critical stage at the time, neither Obama nor Secretary of State and candidate Hillary Clinton took the warning to heart.
Only Biden, as the mascot, listened with relish, and nodded in approval from time to time.
After five years, when Biden stepped into this room again to listen to the report, nothing else had changed, except for two places where the most obvious changes had occurred. One is that he has completed a gorgeous transformation from deputy ** to the owner of the White House, and the other is that he is stronger than China's chip manufacturing industry five years ago.
According to the Boston Consulting Group**:
By 2030, Chinese mainland plans to increase the global chip production capacity by 40 and is expected to become the world's largest chip production base.In contrast, the share of U.S. chip manufacturing in the global market fell from 37% in 1990 to 12% in 2020, making it the weakest link among all high-end industries.
Biden knows very well that the chip war between China and the United States seems to be a one-sided ** of the United States, but in fact it is a life-and-death race in which both sides have half the advantage. China has the advantage of chip production capacity, and it needs to make up for the shortcomings at the technical level, which is a bottom-up breakthrough; The U.S. has the advantage of technology patents, but lacks the industrial chain to turn technology into physical objects, which is a top-down reconstruction.
In the next ten years, China can win; The United States can play, and the United States wins.
one
On January 31, 2024, the U.S. Department of Defense added more than a dozen Chinese companies engaged in chips and artificial intelligence to the list of Chinese companies on the grounds of cooperation with the People's Liberation Army, and declared that it reserved the right to take further measures against these entities. Before that, Huawei was also added to the list.
The Pentagon's actions seem to be high-flying, but they are full of anxiety inside, and the source of this anxiety is to learn from the past of others: China has embarked on the road of the United States as a god.
The development of American chips to this point is largely due to the careful nurturing of the military in its infancy.
In 1947, when Bell Labs discovered semiconductors by accident, the scientific advisory group, which had just experienced the baptism of World War II, was keenly aware that in time, this gadget would inevitably change the world, but the problem was how to make it survive until that time.
As a result, under the strong advice of the scientific advisory group, the Pentagon soon listed semiconductors as a strategic project to be supported by military-civilian integration, and at the same time applied to the second source mechanism in the contract competition law.
The second source mechanism, specifically refers to the practice of the Ministry of Defense using at least two different suppliers when procuring key components, in order to prevent over-reliance on a single supplier.
Thanks to the principle that eggs are not put in one basket, any chips procured by the Ministry of Defense are produced by at least two companies.
On the other hand, the second source also connects procurement and technology transfer, requiring large R&D departments such as Bell Labs to publish technical details according to the R&D process, and licensing other companies to use these new technologies in accordance with the patent transfer law.
For Bell Labs, in the case of the United States and the Soviet Union competing for hegemony at the front and the relatively tight finances, the second source of goods provided them with financial support beyond the first level. But more importantly, as more companies produce semiconductors, the abundance of data points the way for scientists to iterate on technology.
For the Ministry of Defense, the second source of goods not only ensures the stability of the first chain, but also avoids financial waste and monopoly breeding.
For the company, the second source of technology sharing on the left hand and strengthening the anti-monopoly on the right hand has enabled the new technology to popularize the entire industry in a short period of time, and many well-known chip companies in later generations were born in this period of brutal expansion, such as the analog device giant Texas Instruments.
For the industry, the stable procurement of the second source of goods by the Ministry of Defense has made the industry no longer need to estimate the impact of the laws of the market economy, and almost all companies have focused on innovation, thus stimulating the prosperity of the chip semiconductor industry again.
The win-win situation has been laid, however, even if the welfare of the second source of goods is good, the military's stomach is always limited, and when the chips that cannot be eaten are piled up more and more, the market for the private sector has become a decisive winner and loser.
The bane of the collapse of the American chip Zhu Lou dream stems from this.
II. II. II
Tokyo, Haneda International Airport.
A 35-year-old middle-aged man was waving goodbye on the gangway, and if anyone at the scene was a photographer, he would have recognized him as Kazuo Iwama, the famous former president of Sony, known as CCD Kid, but at this time he was only the head of the Tokyo Telecommunications Industry Tape Recorder Manufacturing Division.
In 1951, Kazuo Iwama received a difficult assignment at the board of directors to study semiconductor transistor technology at Bell Labs in the United States.
As a top laboratory serving the U.S. side, Bell has a very high degree of confidentiality, and is not allowed to take notes, take pictures, or ask questions other than the subject matter on site.
Kazuo Iwama can only play a good memory that is not as good as a bad pen.,Learn things thoroughly in the laboratory during the day.,Go back at night and write it down little by little in the notebook.。 In this way, Kazuo Iwama spent four years writing a 256-page manuscript, not realizing that his notebook, which was almost pulped, had written a broad path for the Japanese electronics industry.
In August 1955, the Ministry of Communications in Tokyo developed the world's first TR55 tape recorder with transistors as components according to Kazuo Iwama's manuscript.
After the explosion of the tape recorder, it not only revitalized the semiconductor market in the private sector, but also opened a new era in which the United States is responsible for innovative technology and Japan is responsible for manufacturing products.
Since then, there have been many examples of similar Kazuo Iwama in the United States and Japan.
In the sixties, when low-power chip CMOS was booming in New Jersey's semiconductor companies, Japanese engineer Tadashi Sasaki envisioned applying the CMOS process to integrated circuits to produce pocket calculators for his company, Sharp.
But only one U.S. company took the order, and the others declined because of the low profit margin of CMOS orders in the civilian market and the fact that it would squeeze military chip production capacity.
Later, the number of applications for CMOS chips increased, and Japan thoroughly mastered the technology through industrial policy in the 70s. When the potential of the CMOS process in terms of energy saving and miniaturization is fully realized, Japan completely crushes the United States in terms of technology and production capacity. In 1990, three of the world's top four CMOS companies were Japanese, and only one was in the United States.
Coincidentally, in the 50s, RCA, the world's largest TV manufacturer, pioneered LCD technology, but because it was difficult to apply it to TVs, it sold all the technology patents to Sharp for $3 million.
Adhering to the principle of easy first and then difficult, Sharp first used LCD on a calculator that does not require a high display and only needs a few character changes to maintain production capacity and revenue, and then combined with technological progress to develop large-size LCD displays.
By the beginning of this century, Japan has become the world's largest panel manufacturer, creating a surplus of up to 100 billion US dollars every year.
Success is also Xiao He, defeat is also Xiao He.
Military procurement has not only paved a long stage for chip innovation, but also because of the lack of market competition, American manufacturers have gradually lost their production capacity advantages and technical sense.
Because of this, in the first world war of memory chips, American memory chip manufacturers were almost one-sided by their Japanese counterparts. By the early eighties, when the war ended, at least eighty percent of memory chip companies in the United States had permanently withdrawn from the market.
At the same time, the CPU logic chip, which was expected by the United States, inadvertently caused a large divergence in the chip production model in the process of changing lanes and overtaking, which made the American chip manufacturing industry, which was already bleeding, directly collapse in the middle of the road.
Three
Taipei, Taiwan, China.
Braving the icy winter rain in December, Li Kunxi, the head of Samsung, took advantage of the time to have breakfast and had an in-depth exchange with TSMC founder Zhang Zhongmou, and the purpose of his trip was also very clear: to poach the chip giant who once led Texas Instruments to rise to become a world manufacturer.
Although TSMC has only been established for less than two years, and Samsung has already made a statement in the semiconductor industry with the support of the Blue House and the United States, Zhang Zhongmou firmly rejected Lee's invitation.
Because he believes that the more refined the matter, the more it must be handed over to professional people to do, and only in this way can the products be more competitive and the profits of the enterprise will be maximized.
The era of chip division of labor is coming, and it is very promising to be a professional chip foundry.
In the decade that Zhang Zhongmou carefully planned, with the best-selling of home computers, the production capacity of logic chips such as CPUs increased day by day. As a result, U.S. semiconductor manufacturers who have retreated from memory chips have bet heavily on logic chips, hoping that it will complete the overtaking of Japan.
But what the United States did not expect was that when the logic chip encountered the wave of globalization, the IDM model that had lasted for more than 40 years was gradually declining, and the only remaining high-rise building in the American chip manufacturing industry collapsed in the concept of light assets and almost undefended industrial transfer.
The so-called IDM refers to the three core links of semiconductor production: design, manufacturing, packaging and testing, and the whole industry chain is responsible for the model.
The advantage of this model is that the production capacity is strong and can implement its own strategy in an all-round way, while the disadvantage is that the production front of the enterprise is long, the investment cost is large, and if the end product loses in the market competition, it will drag down the entire industrial chain.
The reason for the collapse of Japan's manufacturing industry is essentially that Japanese companies that adopt a similar IDM model cannot make money after the terminal products lose to China and South Korea.
In the era when semiconductors or memory chips were king, wafer design was relatively simple, and the real difficulty was in the production process, which needed to rely on the technical advantages of the wafer factory, so the early memory chip manufacturers had their own factories, whether it was design, manufacturing and packaging.
After the rise of logic chips, the difficulty has shifted to the design side, different designs and arrangements will lead to an absolute gap in performance, and the production capacity of the fab will only affect the power consumption and frequency performance, which is harmless at the big level.
For example, the performance of the surging series is far inferior to that of the Kirin series, which is enough to show that the decisive value affecting the performance of the chip is not the production process, but the design arrangement. As a result, most of the brands that make processors don't need their own factories at all.
This large divergence in the production model, combined with the inertia of international industrial transfer at that time, gave the United States sufficient motivation to eliminate redundancy from the first end to the enterprise end.
Four
In the White House conference hall, as the most enthusiastic about science and technology affairs since the war, Clinton was listening to the report of the Science and Technology Advisory Committee with relish.
Three months ago, the Secretary of Energy reported to him that Intel intended to break the 193nm wavelength limit once and for all.
In the early nineties, due to the limitations of basic theories and materials, the wavelength of the light source of the lithography machine was stuck at 193nm for more than 20 years.
Intel plans to emit a wavelength of 13 with a high-power CO2 laser5nm extreme ultraviolet light to completely solve this problem.
This wavelength is only one-fourteenth that of an argon fluoride laser and is capable of engraving chips below the 7nm process.
However, it involved a large number of basic scientific research and precision parts matching, and Intel couldn't do it on its own, so it pulled in Nikon, Motorola, ADM and other large manufacturers, as well as TSMC and ASML, which were unknown at the time, to study together.
The project is doable, but the risks and costs are too great for them to bear combined. Intel had no choice but to approach the Department of Energy, hoping that the United States would take the lead in doing this, and this meeting was the last trembling.
After listening to the report, Clinton made a decision on the spot, mobilized three national laboratories of Lawrence Livermore, Lawrence Berkeley and Sandia to go to the front line for research, and allocated $200 million from the Department of Science and Technology and the Department of Energy to fund the research and development of EUV lithography machines.
But he has one condition: he doesn't want Nikon, after all, the United States is betting on EUV lithography technology with all its might, in order to regain control of the global semiconductor industry chain from Japan.
In front of the national fortune, Nikon was out.
The EUV LCC Alliance, which is regarded by the industry as a city on the top of the semiconductor mountain, was born.
This conference had several particularly far-reaching influences on later generations, such as ASML and TSMC rising as world leaders on their own tracks; Another example is Japan, which has missed out on EUV technology, and the semiconductor industry has slipped into collapse.
Of course, for the United States, the most important thing is that the demanufacturing of the chip industry has been greatly accelerated.
In 1981, Reagan kicked off the New Deal reforms, advocating market fundamentalist neoliberalism instead of Keynesianism, which advocated state intervention, restrictions and regulations on foreign investment in advanced technologies and key products were lifted, and a number of high-end manufacturing enterprises began to transfer manufacturing links to other labor depressions under the guidance of the concept of light assets.
Especially in the chip industry, even the focus of the industry policy introduced by the company is to remove duplication and redundancy in the industry, maximize the value of investment, save enterprise costs, and enhance the competitiveness of semiconductor companies.
Therefore, it is natural for companies to believe that concentrating a large number of workers in one link is wasteful and inefficient, and if they can outsource the manufacturing process to reduce the size of the company, their competitiveness will be greatly enhanced.
TSMC, which is famous all over the world with the EUV LCC alliance, provides a perfect solution for this demand of the United States, so the curtain of the transfer of chip manufacturing to East Asia has begun.
Although some people have raised concerns about the voluntary abandonment of chip manufacturing, Clinton is not impressed.
The premise of chip manufacturing is equipment, and the patented technology of the three major national laboratories in the United States supports the operation logic of EUV, and tens of thousands of parts from 40 countries in the United States and Europe support the finished product composition of EUV.
On the other hand, U.S. technology decision-making is fraught with geopolitical ideas. In Clinton's view, although semiconductor manufacturing is not in China, it is still in the sphere of influence of the United States, so it is not a big problem.
But when China, a country outside of the U.S. sphere of influence, began to demonstrate its strong semiconductor manufacturing capabilities, everything took a subtle turn.
Five
On January 31, 2024, Intel suddenly announced that the wafer fab originally scheduled to be built in Ohio this year will be postponed to 2026 due to the lack of financial subsidies for a long time. Prior to this, TSMC's two 3nm factories in Arizona were also forced to be postponed due to subsidies and talent problems.
Undoubtedly, this is a further blow to Biden's plan to promote homegrown chip manufacturing.
Rebuilding the local chip industry is Biden's No. 1 national policy. The widespread application of digitalization and artificial intelligence has rewritten the underlying structure of industry.
In the age of machinery, whoever controls oil controls all countries, and in the age of AI, the subject of this phrase becomes a chip. The United States will categorically not give up the chip manufacturing industry, which can affect the global industrial trend, to China.
To this end, Biden not only used administrative means to force Samsung and TSMC to build factories in the United States, but also introduced the Chips and Science Act in 2022, specifically allocating $52 billion to support traditional American manufacturers such as Intel to rebuild wafer production plants, but the result backfired.
The postponement of the TSMC and Intel project has put the two insurmountable barriers for the United States to rebuild its local chip industry from behind the scenes, and Intel is a matter of money.
The TCO (total cost of investment) of an advanced fab consists of two parts, the upfront capital expenditure consisting of land, buildings, equipment, and power generation facilities, and the operating costs consisting of labor, utilities, materials, and taxes.
The American Semiconductor Association has calculated that because of the high cost of labor, materials, and supporting facilities, the TCO of the American wafer fab is about 25 to 30 higher than that of the same wafer fab in Taiwan, China or Singapore, and 50% higher than that of China, which has the best incentive policy, and this expense is simply unbearable for enterprises.
Intel, for example, will invest more than $100 billion in its Ohio fab project by 2030, while Intel's profit last year was just over $10 billion.
Although the new bill shows that Biden is willing to provide incentives for chip companies, it is one thing to release policies, and it is another thing whether they can be implemented. Both Intel and TSMC emphasized that they have repeatedly applied for subsidies to **, but have never received any response.
As for why there is no money, the answer is very simple, the debt is about to empty the federal finances, a turnip and a pit, and Biden can only move it from other places if he first finishes the account of mandatory spending.
But the question is, after calculating the account of mandatory expenses, how much money can be left for chip subsidies?
TSMC is a human problem, the factory cannot find skilled technicians to install key precision equipment, and there is a shortage of talents for the later operation and support of EUV lithography machines.
Even Zhang Zhongmou has warned that it is expensive, wasteful, and futile for the United States to increase domestic chip production, and that the cost of manufacturing chips in the United States is 50% more expensive than in Taiwan, because the American chip manufacturing industry does not have the talent pool needed to expand and succeed.
It is difficult to cross the mountain, who is sad and lost the way.
And what is more sad than the person who lost the road is that he knows that the road is in **, but he can never go through.
When the United States used power to open the last hurdle of the reshoring of the chip industry, it found that the real Guanshan was not outside the Guanshan, but inside the Guanshan.
The problem of capital and talent is like two insurmountable moats, blocking the ambition of the United States to reshape the local chip industry at its own gate. It takes courage to attack a city, and it takes the determination to die and survive to attack one's own city.
Scrape the bones to heal the wounds, do not break or stand!
In the life-and-death race between China and the United States in the long-term chip manufacturing industry, it is not yet known who will die.