Who’s winning the chip war between China & US?

Right now, it seems China and the US start to exchange punches.
June 25, 2023
The China Academy

The so-called “chip war” between the world’s two largest economies has been going on for a few years now, and Western media is eager to declare victory and the loss of China at every turn.

But not everyone in the West shares this sentiment.

Jensen Huang, CEO and founder of NVIDIA, for example, has expressed his concerns recently, saying that “If we are deprived of the Chinese market, we don’t have a contingency for that. There is no other China, there is only one China,” then he added that there would be “enormous damage to American companies” if they could not trade with China.

Similarly, ASML’s CEO Peter Wennink says his company’s access to China is ‘essential’, before pointing out that it was “logical” that China would seek to develop its own semiconductor equipment when restricted from purchasing similar products from abroad.

So why are those executives so concerned if China is supposedly losing the chip war as described by the Western media?

Before diving into details about the semiconductor industry, let’s take a look at the numbers.

  • For Q1 2023, Intel reported a 36% decline in revenue compared to Q1’22, its lowest since 2010.
  • AMD is not much better. It registers a loss of 139 million USD in Q1, a staggering 118% drop from Q1’22. They are not alone in their bad fortune.
  • As a whole, the global semiconductor industry revenue is expected to drop by 11.2% this year, according to a new Gartner forecast.
  • There are certainly a number of factors at play here besides increasingly stringent US export restrictions to China. But the revenue declines are hardly surprising when one gets into a fight with one’s biggest customer and the world’s biggest semiconductor market. Those companies are forbidden to sell high-end semiconductors to China by the US government. Inevitably, as ASML’s CEO pointed out, China would seek to develop its own technology and eventually become a formidable competitor for the West in the semiconductor industry.

    In fact, China’s domestic semiconductor companies see the US chip embargo as a windfall and are moving quickly to seize any market share ceded by foreign competitors. Over the last 3 years, these Chinese firms have seen phenomenal growth in both revenues and profits.

    But can we declare China the “winner” in this contest yet? Not really.

    The dynamics of the semiconductor industry are highly complex.

    Not all semiconductors are the same. One way to broadly categorize the market is to divide it into lagging-edge & leading-edge processes. The terms “lagging edge” and “leading edge” in chip manufacturing refer to the size of one transistor on a chip, as determined by the width of the gate. Generally, a smaller individual transistor size means an increased total number of transistors that can fit into a single computer chip of a given size, and therefore enhance the speed thereof. Smaller transistors also consume less electricity, primarily because electrons must travel a shorter distance from one transistor to another, chips can operate on a lower voltage, and leakage current is reduced. Therefore, competitiveness as defined in the semiconductor industry comes down to calculation speed per unit of power consumed. Processes with transistor sizes of 28nm and above are typically considered “lagging-edge” or “mature”, while those below 28nm can be considered “advanced” processes, while the smallest possible semiconductor size that can be manufactured on a commercial scale is labeled as the “leading-edge”. Smaller transistors result in higher performance but are far more complex and expensive to manufacture.

    The market landscape for lagging-edge and leading-edge processes differ significantly.

  • Leading-edge processes are primarily used to build processors that end up in mobile phones, PCs, data centers, and other devices that attempt to maximize speed while minimizing power consumption. TSMC, Intel, and Samsung are among the leading manufacturers of advanced and leading-edge processes, which range from 16nm/14nm to 3nm.
  • By contrast, mature processes are mainly used in industries such as automobiles, the military, infrastructure, customized chips, and other use cases where reliability, stability, and/or lower prices are crucial. Mature processes are also used for small and medium-capacity memory chips, analog chips, MCUs, power management chips, analog-digital hybrid chips, CMOS sensors, ASICs, and sensors. The demand for lagging edge semiconductors has increased significantly due to the rapid growth of new energy vehicles, next-generation communications technology, security, and cloud computing.
  • The Chinese company SMIC has mastered manufacturing processes up to 7nm, although 7nm manufacturing likely requires the use of foreign-made equipment, such as ASML’s DUV lithography machines – on which the US is trying to impose ever more stringent export restrictions.

    At present, chips manufactured using mature processes account for approximately 70% of the global semiconductor market. Chinese manufacturers are quickly gaining ground in this market segment at the expense of their Western competitors. China has the advantage of being the biggest market in the world for lagging edge semiconductors. Recently, SMIC also officially stated that the capacity utilization rate of its 40nm and 28nm factories has returned to 100%, which shows how popular mature processes are. At present, the company is the 3rd biggest provider of mature process chips worldwide, with the gap between it and the top 2 shrinkings rapidly. Therefore, the business is lucrative enough for SMIC and other lagging-edge fabs to make a good living.

    Of course, SMIC is not content with just being a low-end provider forever, last year Bloomberg reported that SMIC had successfully delivered 7nm ASICs (custom-made chips optimized for a specific function) to its customers, although the production volume is unknown.

    On the other hand, Chinese scientists have been working around the clock to enhance the domestic chip-manufacturing ecosystem.

    A key piece of equipment for chip manufacturing is the lithography machine, which can project patterns of transistors onto a silicon wafer using light. A lithography machine uses different types of light and optical systems depending on the size and resolution of the patterns to be printed. For example, extreme ultraviolet (EUV) lithography machines use a wavelength of 13.5 nm to print the most intricate layers on a chip. EUV lithography enables semiconductor production smaller than 7nm. A different but related technology – deep ultraviolet (DUV) lithography – can be used to produce 7nm chips, but DUV would have difficulty scaling beyond the 7nm node. ASML, a Dutch company, is currently the only company in the world that can produce EUV lithography machines, and it’s restricted from selling these machines to China due to export controls imposed by the US government. It becomes the biggest obstacle for Chinese firms to manufacture advanced process chips.

    Fortunately, the Chinese Academy of Sciences has made breakthroughs in the development of several core technologies for EUV lithography, such as the EUV light source, the dual-stage system, and the particle accelerator. These technologies are essential for producing high-end chips on sub-7nm nodes using EUV lithography. The Chinese Academy of Sciences has collaborated with Tsinghua University to propose a new type of particle accelerator that can generate EUV light with greater stability and efficiency. CAS has also mastered the dual-stage system, which is a key component for achieving high precision and accuracy in EUV lithography. However, there are still some challenges and gaps for China to develop a complete EUV lithography machine, such as the optical lens system, which requires extremely high quality and purity. Putting these breakthroughs together & commercializing them into a working EUV lithography machine will also take time & investment.

    The CAS’s effort toward China’s self-reliance on chips doesn’t stop here. It has recently showcased its latest advancements in open-source chip technology at the 2023 Zhongguancun Forum. Two major breakthroughs have been introduced: the “Xiangshan” high-performance open-source RISC-V processor core and the “Aolai” RISC-V native operating system.

    The open-source chip design could one day turn out to be a game-changer, but at present, Chinese firms are facing multiple uphill battles.

    Under US pressure, the Japanese government recently announced a measure to restrict exports of 23 types of advanced semiconductor manufacturing equipment, starting from July 2023. This would certainly affect the revenues of the Japanese manufacturers involved, as well as disrupt the supply chain for most Chinese chipmakers. The Chinese government immediately protested and vowed to retaliate.

    On the other hand, sanctions and restrictions are no longer a one-sided game played by the US. On May 21, the Cyberspace Administration of China (CAC) announced a ban on the American chipmaker Micron, prohibiting it from selling its products to operators of key infrastructure projects in China, such as telecommunications, energy, transportation, and finance. Micron failed a network security review and the restrictions were imposed as a result. The CAC claimed that Micron’s products have serious security gaps that pose significant risks to China’s critical information infrastructure supply chain and would negatively impact Chinese national security.

    Micron is the biggest US memory chipmaker and derives more than 10% of its revenue from mainland China. The company warned that the ban could cost it as much as a high single-digit percentage of its annual revenue. The ban on Micron is a potential indicator that domestic memory suppliers, such as Yangtze Memory Technologies Corporation (YMTC) are starting to produce adequate substitutes to US memory products.

    None of the US firms can manufacture advanced process chips by themselves, just like their Chinese counterparts. They have to rely on companies from allied countries, such as Samsung, TSMC, ASML, NXP, Siemens EDA, etc. for products and services across large parts of the semiconductor value chain. And increasingly, these companies have found themselves caught in the tug-of-war between the two giants, as they have significant business interests in both countries.

    Right now, it seems China and the US start to exchange punches, and the jury is out on who will eventually emerge as the winner of this chip war. What’s certain is the collateral damages are shared by companies from both sides, and huge opportunities are lost. In the near term, it appears that neither will win a decisive victory against the other, as they may be pursuing different objectives in this conflict.

    The US is attempting to forestall China’s ability to produce leading-edge semiconductors that are competitive with the high-end products of companies such as Intel, AMD, & Nvidia. Meanwhile, China is attempting to domestically build a complete, end-to-end semiconductor design and manufacturing value-chain, regardless of whether they have lagging-edge or leading-edge capabilities, to no longer be susceptible to the whims of foreign restrictions. This is a pattern commonly manifested in Chinese industrial policy.

    When developing a domestic industry in any sector in which China is a relative latecomer, it does not initially seek to compete in the high-end segments of the market, nor will it seek to completely exclude foreign suppliers. China tends to prioritize the completeness of the end-to-end industry value chain, even if it’s on the low end of the spectrum.

    In the context of semiconductors, this means China is likely trying to master every aspect of the design and manufacturing process, be it refining raw materials, EDA tools, architectural design, manufacturing, or packaging. We have to say building a complete end-to-end semiconductor industry in one country is a feat in and of itself, since no country or region in the world (including the US) has accomplished such a task to date.

    In the short-term China would likely be content to focus on building on 28nm or even 45nm nodes, as long as the end-to-end processes can be fully mastered by Chinese companies. So in the short and medium term, there will still be plenty of room for the likes of Intel & Nvidia to profit from the Chinese market.

    However, China’s mastery of the semiconductor value chain, even if it remains on the low end for another decade or so, means that no country or coalition can deny the benefits of digital transformation to anyone else through sanctions and export restrictions, nor can global semiconductor giants overcharge for products built on mature nodes.


    The China Academy
    Share This Post