A flurry of announcements is coming out of China about advancements in semiconductor precision equipment and manufacturing. As China pushes for self-sufficiency in this critical sector, non-China media reactions often swing between surprise at their progress at the start, followed by the “but” arguments that often appear lifted from the previous articles at the time of previous such announcements. They effectively express skepticism about their actual capabilities, citing issues like potential low yields or unverified information. While most doubts are valid, particularly given the lack of transparency and China’s need for peacocking, the core issue is that hardware manufacturing is an incredibly challenging journey with multiple routes to higher levels of precision, and China is progressing on its path.
This note is on a heavy topic, so let’s start with the worst possible analogy! To illustrate, think of the characters Sheldon Cooper and Howard Wolowitz from The Big Bang Theory. Sheldon, the theoretical physicist, represents software—cognitive, individualistic, and ever-evolving in new directions. Conveniently homonymed Howard embodies hardware—not because of anything else, but because he is the looked down upon practical engineer amongst his innovating, scientist friends. This crude analogy is to highlight that hardware development is a cumulative process requiring meticulous teamwork and discipline. But let's not stretch the comparison too far; after all, not every software developer is a Sheldon, Hardware making is about hundreds of Howards working together, not one.
Software development is not child’s play, but anyone can get started. Thanks to user-friendly tools and vast libraries, even young kids can start programming in Python or Java without a deep understanding of the underlying codes. Ace software developers often become celebrities, gracing conference stages and sharing innovations. Software development has many accumulative, complex aspects and requires tremendous collaboration in any meaningful project, but the field permits individuality and visibility.
The starting point in hardware making is a vast amount of capital, which, at the cutting edge, is now similar in scale to the highest-profile infrastructure projects. It's getting more expensive and challenging by the day. Building advanced semiconductors isn't just about pouring in billions of dollars or buying the latest equipment. It demands the whitest white-collar professionals to work with the discipline of assembly line forepersons, adhering to processes where even tiny contamination can render a silicon wafer useless (something this author learned, to his chagrin, in his engineering project years ago). This level of disciplined teamwork has been challenging to perfect by all but a few. It's rare to see hardware innovators from fabs giving keynote speeches; their achievements are embedded in the silent efficiency of a well-coordinated team.
Plus, there are no unique paths of progress. A better fab, for instance, is an accumulative result of tens of thousands of process innovations, including those in equipment manufacturing. Even two companies with the same ASML equipment may have completely different processes afterwards to produce the same nanometer chip eventually. The evidence of this has been around for a while: companies like TSMC and Samsung Electronics have reached new heights through different technological paths, each with their own set of processes and innovations.
In other words, shortcuts are virtually nonexistent in hardware manufacturing. You can't simply start building a 5nm fabrication plant by purchasing equipment from leading manufacturers. The equipment itself is the culmination of countless process innovations that need to work in perfect harmony. Even beyond the equipment, the entire manufacturing process requires mastering numerous steps involved in the processes at lower levels.
In its quest for semiconductor advancement, China is charting its unique manufacturing path for a while. Unable to access cutting-edge EUV lithography equipment due to export controls, companies like Huawei and SMIC have turned to existing DUV equipment and techniques like double patterning. This method involves multiple laser passes to etch designs at sub-20 nm resolutions, similar to techniques previously explored by Intel. They are currently trying self-aligned quadruple patterning (SAQP), which uses even more passes to increase transistor density and reduce power consumption.
China’s determined journey is yielding results beyond manufacturing, particularly in equipment manufacturing as well as in EDA toolmaking. This note is not about its geopolitical implications but about what it may mean for technology investments.
As we have repeatedly said with our following catchphrases, in the GenAI era, competitive corporations and nations face the same Prisoners’ Dilemma-like choice set: if they don’t do it, the entities they hate the most will. This is most manifested in all the hardware segments. With India and GCC nations, apart from a few others, beginning their journey in the Semi manufacturing, the
The semi race will spur innovations of different types also because of the different end objectives. The GenAI use cases in China are most pursued in automation, viz., mobility and robotics. Outside, the priorities are in consumer applications and corporate efficiencies. For investors, the compounding of progress is assured, as no one will rest easy.