This is not a tech talk club, it's about redividing interests. But no matter how it is divided, it implies control over terminal capacity and the commercial deployment of capacity.
Author 丨 Meng Hua
Edited 丨tian grass
Produced 丨 Automan Media
On March 29, the South Korean government and companies confirmed a message that the U.S. government has invited South Korea's Samsung and SK Hynix to participate in a semiconductor alliance called "Chip4" (Chip Four-Party Alliance).
In addition to Korean companies, Japan's Toshiba, Renesas, Tokyo Electronics, Taiwan's MediaTek, TSMC, Riyueguang, and the United States Intel, Qualcomm, Micron, Broadcom and other chip companies are invited.
The list includes chip design companies, raw material consumables companies, as well as foundry capacity and packaging and testing companies, in fact, including the entire chip industry chain: from semiconductor materials and manufacturing equipment, from gluing/developing equipment to consumables, from physical vapor deposition equipment to chemical deposition, from ion implantation to etching and polishing, equivalent to a global chip production capacity mobilization.
1
The objectives of the Alliance
This time, the focus is clearly on chip production. It was the first global "producer alliance" of chips officially led by the United States. This alliance that excludes China points to a fairly clear message that is to achieve the integration of the whole chain of advanced process chips and curb the competition that China may participate in.
From the passage of the Chip Act by the US Senate last year, the Passage of the Innovation and Competition Act of 2022 by the House of Representatives, and the US Department of Commerce's requirement to designate foundries to hand over production and customer data, it can be easily judged that the United States is sorting out and pooling its chips, using technological and industrial advantages more completely to contain China, the largest geopolitical competitor.
It should be pointed out that the United States does not have all the knowledge and production capacity related to chips today, in fact, the production capacity of chips in the United States has dropped to 12%. The United States is better at the development environment (EDA software), accounting for 85% of the world; core IP accounts for 52% of the world; semiconductor manufacturing equipment comprehensive output (the largest OEM players are the Netherlands Asma, Japan's Nikon and Canon) account for 50% of the world.
What the United States can do now is to forcibly aggregate these industrial chains with financial and military hegemony and obey its own orders. The proportion of intellectual property rights and manufacturing capacity is not important.
But contrary to this, the value of the chip industry is increasingly flowing downstream. R&D expenses in the manufacturing sector account for 25% of the entire chain, while value output accounts for 45% of the entire chain. In other words, the United States occupies the upstream of design, and Japan occupies the upstream of raw materials, and the profits earned are not as much as those of foundries.
In the context of the global "lack of core", upstream manufacturers have seen that profits have been taken away by foundries. Because it is closer to the customer and the relationship between the distribution product, the distribution of benefits is more favorable to the foundry. This is the internal contradiction of the chip industry chain.
The largest market for chips is in China. According to 2021 data, semiconductor sales in the Chinese market were $192.5 billion, an increase of 27.1%, accounting for 34.6% of the global market, continuing to rank firmly in the world's largest semiconductor market.
China's own chip production capacity caught up with the United States in 2010, and the process is far from it. If the current inertia is followed, by 2030, the US chip market share will fall to 10%, while China will rise to 25%-30%. Taking into account the factors of TSMC and Samsung setting up factories in the United States, the process gap between China and the United States will still be at least a generation apart.
The 2nm chip process has reached the physical limit of silicon-based semiconductors, how to go next, the academic community is not ready at the basic theoretical level, and the engineering ideas in the industry are impossible to talk about. From this point of view, it is conducive to the catch-up of the laggards.
Therefore, the current chips of the United States are at risk of falling, and must be fully used while they are still in hand, at least to delay the pace of China's research and development.
2
Not good for car companies
What is missing in the automotive industry is mainly mature process (40nm-200nm) chips, which rely on global production to be resolved in the process of moving towards the boom cycle, that is, after 2024.
What the original engine factory could do was to strengthen procurement management, strengthen stockpiling, and transfer the inventory of chips from suppliers to their own hands. This process has to bear the corresponding inventory costs, contrary to the previous principle of "lean production", but it is better than cutting off supply and production.
Is the U.S.-led "Chip4" alliance supposed to take shape and operate effectively, good for producers, or good for users?
This question seems difficult to answer. But if you look at it in another way, that is, chip suppliers achieve full-chain coordination under a certain mechanism, that is, to achieve the integration of resources and production capacity, will there be more products, or will they become less?
This is an easy question to answer. All industrial policies aimed at distorting the laws of the market will eventually distort the market itself. Any trust behavior is detrimental to pan-users.
From automakers to Tier 1 suppliers, there is no appetite for similar upstream alliances, whether it is government-led or dominated by a leading enterprise. This is not a technical talk club, but there is a possibility of redividing interests. Regardless of the division, it implies control over the commercial deployment of terminal capacity and capacity.
As customers, it is difficult for auto companies to achieve alliances to cope with production integration. Moreover, this integration is still entangled by the US government and has obvious political purposes. This makes both non-China car companies in China worry that their orders or relationships with chip suppliers will be affected.
3
The abacus of the members
Requesting data is a one-time administrative act that can be obtained by threatening sanctions. But what mechanism does a multinational industrial alliance led by a government rely on for long-term operation? In other words, benefits must be given to drive supplier cooperation.
The Chip Act, passed by the U.S. Senate, supports a $52 billion stimulus package, but for now it only sends the bill to the House of Representatives, and ultimately depends on the actual appropriations of the House of Representatives.
If you want to see the real intentions of suppliers, you can look at the "inter-enterprise alliance" they established last May, that is, the American semiconductor alliance SIAC established by 64 semiconductor manufacturers. But the purpose of the alliance is not to hit Chinese customers, but to urge the U.S. Congress to compensate semiconductors for their distance from East Asia's largest customer base through a $52 billion grant without compromise.
TSMC and Samsung invested $12 billion and $17 billion, respectively, to build advanced process facilities (5nm and 3nm) in Arizona and Texas. Both complained, however, of favouritism.
Samsung urged the U.S. government to ensure that all eligible companies can compete for federal funding on a level playing field. They refer to Intel, which is believed to have received subsidies that exceed its share of production capacity. The current distribution plan is not very friendly to Korean companies.
Samsung and SK Hynix have both established large-scale memory chip factories in China. Samsung's vice president in 2021 said that investment in China has reached $46 billion, and the production capacity of Samsung Electronics' Xi'an factory accounts for 42% of Samsung's total flash memory and 15% of global production capacity.
In addition to investing in the DRAM chip factory in Wuxi, SK hynix is introducing foundry manufacturing and sales businesses such as CIS, DDS, and PMIC to Wuxi. The joint venture between SK Hynix and Taiji Industry, "Haitai", ranks among the top ten in the country in terms of sealing and testing scale.
Last year, SK Hynix intended to introduce Asmay's EUV lithography machines for retrofitting its fab in Wuxi to upgrade its production line capacity, but it was stranded due to opposition from the United States. The United States has yellowed the upgrade plan by ordering Asma.
No more than 30 percent of asma's upstream components, particularly optical systems, depend on U.S. supplies. This is not a problem of the upstream and downstream of the supply chain, but the ability of the United States to comprehensively use hegemony is reflected in the chip industry chain.
Korean companies invest in China and in the United States on a large scale, but investing in China cannot substantially help China improve its process capabilities, which is a hidden red line.
TSMC's investment in the mainland has the same problem. TSMC itself does not intend to put the first decimal process overseas, whether it is the United States or the mainland, but the mainland's process and wafer manufacturing level are two generations behind it.
How do Japanese companies in the "Chip4" alliance view the alliance? At several meetings at the end of 2021 and January 2022, the Ministry of Economy, Trade and Industry summarized several factors that have been declining in the global proportion of semiconductors in Japan, two of which are more eye-catching.
First, it has not adapted to new trends. In the process of chip design and manufacturing separation, the vertical division of labor gradually shifted to the horizontal division of labor, which directly contributed to the rise of professional foundries. The Japanese company's requirement of "their own people" to control manufacturing is the key crux.
The second is to fall into the trap of "self-research". Using only Japan's existing technology and human resources for joint research and development, trying to exclude the participation of foreign companies, in the case of declining willingness to invest, no obvious results have been achieved.
From last year to this year, TSMC set up a wafer manufacturing subsidiary, JASM, in Japan, and announced that it will introduce 22-28nm process technology to Japan.
However, Gan Liming, president of the "Semiconductor Strategy Promotion Parliamentarians Alliance" in the currently ruling Liberal Democratic Party, proposed that if this process is only met, it will lose the significance of promoting Japanese semiconductors, and the subsidies invested by the government will be wasted. Denso and Sony announced that they would join JASM to develop a 12-16nm process, but it still could not satisfy the government. TSMC's investment in Japan is therefore variable.
Japan hopes to gather official and private investment of 7 trillion to 10 trillion yen (the amount is not determined, that is, this matter has not yet become a proposal), seize the new outlet of photovoltaic chips, and use 10 years to promote the development of 2nm processes.
The US $52 billion stimulus policy is also aimed at advanced processes, but it does not mention 2nm very clearly.
In this way, within the "Chip4" alliance, everyone actually has their own ideas, but they carry out technical containment of China, which is an area of overlapping interests.
Therefore, the establishment of the "Chip4" alliance is no problem, it depends on the next step of operation. After all, China's multiple identities as a production base + customer + competitor make the United States, Japan, South Korea and Taiwan very entangled, and if they want to put forward a consistent action plan, they need to be carefully designed.
China doesn't really have a very good response to this. China's large-scale investment in semiconductors is a strategic action, not a temporary event. But for quite some time, it will be subject to the supply and containment of the members of this alliance.
The consulting agency commissioned by the United States believes that Chinese mainland "lack of industry knowledge, hindering the development of self-sufficient supply chains", so independent semiconductor manufacturing capabilities will not be developed in 5-10 years.
The prediction of the future itself is not important, but it reflects the technological complexity of the chip industry and involves business and geopolitics. Chips not only need lithography machines, but also a full set of technology. Instead of focusing on what the alliance is going to do, focus on what it's going to do.
【Copyright Notice】
This article is the original manuscript of Automan Media
Unauthorized reproduction is not permitted
▼