Recent scarcity in supply and the growing role of silicon chips in turning dumb products into smart ones have sparked renewed interest in the ever-growing semiconductor industry. On the one hand, incumbent players are aggressive in holding their positions, even through massive subsidies and creating war-like situations. On the other, aspiring entrants are desperate to find an entry route. The underlying cause of the apparent rigidity of the industry in synchronising with demand and creating a high entry barrier is a monopoly spread across the value chain. Despite the distribution of the semiconductor value chain over more than 25 countries, every significant industry segment has a monopoly-like competition scenario. Unlike conventional reasons like collusion, treaty, natural resource stock, or vertical foreclosure, the race to increase the quality and reduce the cost through improving precision has led to this situation. As a result, the industry is neither flexible nor amenable to new entries.
At the bottom of the value chain is silicon wafer production, constituting four per cent of industry value added. Although silicon wafers are produced from nature's widely available sand, Japan has a monopolistic position with a 60 per cent market share.
The underlying cause has been the race of increasing precision or purity-reaching 99.9999999 per cent purity. As it contributes to both yield and performance, producers have an endless desire to improve the purity of silicon wafers. As a result, R&D conducted over the decades has resulted in high capital intensiveness and patent barriers.
After having a highly pure silicon wafer, the next step is to design chips and project their images on the wafer. The growing complexity, scale, and precision have led to the development of highly specialised software tools. These tools are at the core of articulating production process-related knowledge and design cores as a reusable library. Growing specialisation, zero cost of copying software-centric assets, and exponentially increasing R&D costs have ended up in a monopoly in design automation software. It happens to be American firms like Cadence, Mentor Graphics, and Synopsys are at the helm.
The projection of design needs a photolithography machine. Although it started as the makeshift arrangement of optical devices, over the last 60 years, these machines have become highly sophisticated. Instead of being a supporting machine, photolithography is now at the centre of silicon chip production competence-creating geopolitical tension. In addition to demanding high-precision engineering design, this segment has been requiring ever growing risk-taking R&D investment. For example, Dutch company ASML had to make a highly risky -- as high as $10 billion -- investment over more than 20 years before shipping its first extreme ultraviolet (EUV) machine. Consequentially, ASML has attained the stage of sole supplier of EUV machine-the only option for producing chips at 5nm node. But ASML cannot make its machine on its own. In addition to developing core technologies like the EUV light source, ASML needs to rely on sourcing components from 4700+ suppliers and precisely integrating them with high-precision engineering.
The industry has been demanding fine chemicals and gases to process the ever-shrinking features of silicon chips. The same journey of reaching precision has led to monopolisation in this segment by Japanese and German companies. Furthermore, ever-growing complexity and accuracy have led to more than 400 discrete steps, running over more than a week to turn silicon chip design into 3D prints on silicon wafers. The industry has been demanding increasingly specialised machines in each of these steps. As a result, a long race has resulted in monopolistic dominance of American and Japanese firms. For many machines, there has been only one sourcing option.
At the core of semiconductor production is a fab-factory of processing silicon wafers for turning designs as digital content into the 3D nanostructure. Unlike many other manufacturing plants, fab is fiendishly complex and capital intensive. The transition to a high-end node demands more than a year of R&D to optimise the yield. It's so complex that even Intel has been struggling for over a year to migrate from 10nm to 7nm node. And more than 40 per cent of the personnel of TSMC work on process R&D for integrating various machines and fine-tune their integrated operation for maximising yield. Process R&D, growing capital need (reaching over $15 billion for 5nm node), and scale have also been weakening competition in this segment, resulting in only two high-end fab operators-TSMC and Samsung.
In the end, processed silicon wafers proceed to slicing, testing, bonding, and packaging. Once this step used to be labour intensive. But growing precision has been demanding knowledge and skill, which often human endeavours cannot provide. Hence, there has been increasing automation, making it highly capital intensive-creating economies of scale and scope advantage. Thus, despite the growth in semiconductor volume, this segment has not had a proportionate increase in labor demand. As a result, even this segment did not keep spreading and migrating worldwide to source low-cost labor. Instead, it has been consolidating as part of a cluster, growing from a humble beginning in Malaysia and Taiwan.
In addition to the growing monopolistic situation, geopolitics in the polarising semiconductor value chain has now been capitalising the edge-creating China's red and American blue channels. America is posing a barrier to the supply of high-end software, machines, and chemicals from the US firms and firms having supply chain linkage with American firms to China. Consequently, China has been compelled to develop substitutes at each value chain layer. But due to the extreme level of specialisation developed over decades, it has become challenging to tackle complex engineering and manufacturing problems. Generating a financial return exceeding the cost of capital is another dimension of this challenge. Hence, even large domestic market is not sufficient to pursue import substitution strategy.
So far, any kind of physical law or natural phenomenon has not been the underlying cause. Instead, research and innovation carried out by private firms in response to market forces have been the driver of this journey. As a result, this precision race creating monopoly has also been lowering labour requirements; but there has been a growing demand for high-end human capital. Hence, countries like Taiwan, Japan, and the USA have found the semiconductor industry as a source of high-paying jobs. On the other hand, aspiring less developed countries are increasingly finding themselves in a disadvantageous position-which has created a race to build infrastructure and give incentives for alluring foreign direct investment. But as the risk of turning investment into profitable return keeps growing and the scope of adding value keeps eroding, it's time to rethink how to leverage this lucrative industry.
Rokonuzzaman, Ph.D is academic and researcher on technology, innovation and policy. [email protected]