This article is a state-of-the-industry survey on the dwindling silicon foundries that manufacture integrated circuits. The authors rightly question the trifecta of factors that currently constrict and afflict business: i) the viability of the ever-changing integrated circuit improvements (shrinking sizes, currently seven nanometers); ii) the corresponding demand from companies for newer designs; and iii) the current availability of one pure play foundry, Taiwan Semiconductor Manufacturing Company (TSMC), that continues to pursue the latest silicon generations to support these evolving needs, along with two companies (Samsung and Intel) that also make their own chips. The author wonders “how the industry can adapt as physical limits eventually make further shrinkage impossible (or [prohibitively] expensive).”
Over the past few decades, in correspondence with Moore’s law, the exponentially increasing transistor counts were made possible through corresponding intensive capital investment increases “to build fabrication plants and develop more aggressive processes and novel device structures.” The possibility of keeping up with Moore’s law was largely due to continued exponential shrinkage in the 2000s, through “multiple revolutionary innovations, such as high-dielectric-constant (high-k) gate dielectrics, metal gates, strained silicon, and the nonplanar transistors known as [fin field-effect transistors, FinFETs].” These allowed manufacturers to work around “fundamental physical limits ... such as gate oxides just a few atom layers thick, as well as large leakage currents and other non-idealities in the tiny devices.”
To make innovations beyond the seven nanometer limit, and for it to be economically viable with respect to throughput and yield, requires large-scale investments. This article seems to be largely triggered by the decision of GlobalFoundries, in August 2018, to halt development of its seven nanometer process. Since GlobalFoundries “was created in 2009 from the manufacturing operations of Intel’s arch competitor Advanced Micro Devices (AMD),” it has also acquired both Chartered Semiconductor (a Singapore-based foundry) and “the manufacturing operations of IBM”; such consolidation has shrunk the competitive landscape.
With the reduction of suppliers to one pure play entity, and two others with their own businesses to cater to, such investments are quite unsustainable for supplying needs at a global scale. The article highlights this risk and its possible impact on US semiconductor industries.