Innovation policy for semiconductor chips: role of design and manufacturing innovation

• In Event: AI, Innovation Policy, and Global Sustainability

Abstract

The passing of the CHIPS and Science Act in the United States and the German government’s failed attempt to localize an Intel factory have made clear that domestic semiconductor chip manufacturing is a key priority for governments. In the policy arena, the erosion of semiconductor manufacturing capabilities in high-income countries and their gradual transfer to countries in East Asia is primarily discussed from a security of supply perspective, especially due to post-COVID chip shortages. In this paper, we investigate a critically overlooked aspect of losing domestic chip manufacturing capabilities: its impact on subsequent generations of innovation. With increasing global concerns about economic competitiveness, understanding the relationship between manufacturing capabilities and their linkage to technological innovation directly informs innovation and industrial policy formulation.

We study the question using a custom-assembled patent dataset of over 190,000 patents, spanning semiconductor chip innovation from years 1960 to 2020. First, we study the complete patent citation network. We trace the knowledge flows contained in the citations by categorizing the patent claims as either chip manufacturing or design knowledge. We confirm a general pattern of manufacturing innovation knowledge flowing into design innovation, but not vice versa. We also observe a decade-by-decade increase in the proportion of mixed-claim (both manufacturing and design) patents. Second, we identify the patents associated with key technological breakthroughs (e.g., SiO2 surface passivation and FinFET design) in the dataset based on historical documents. In addition to analyzing the knowledge flows, we investigate clustering of the citation networks surrounding these key patents based on geography and firm. We find that breakthrough manufacturing innovations result in an uptick in design innovations. There is no evidence of geographic co-location of the manufacturing innovations and the follow-on design innovations. In many cases, manufacturing knowledge feeds into design innovation within the same firm, suggesting the need for close collaboration.

Our findings have important implications for the design of industrial policy to localize chip manufacturing. It has been established that manufacturing knowledge is easier to transfer globally, primarily through trade in advanced capital equipment. A secondary channel is the movement of skilled personnel with tacit knowledge of the manufacturing processes. However, given the competitive nature of the semiconductor chip industry, the delay in knowledge transfer can lead to economically costly lags in realizing the next generation of design innovations. This effect may be exacerbated by the increasing trade tensions, where access to capital equipment and personnel is restricted. Second, the rising co-patenting of manufacturing and design claims suggests that manufacturing capabilities are becoming integral to design innovation.

More generally, patents underrepresent manufacturing innovations, since they are easier to protect as trade secrets. Given the semiconductor chip technology's economic value and technological sophistication, we can observe manufacturing innovations through patents. Based on this case, we hypothesize that similar patterns exist for technologies undergoing decade-by-decade improvement in performance characteristics, such as Li-ion batteries. However, they might require methods beyond patent analysis to observe linkages between manufacturing and innovation.

Author

• Anurag Panda, ETH Zurich
  Presenting Author