Understanding Critical Dependencies
What do we use semiconductor chips for? Or put another way, why do we care about integrated circuits from the vantage of national security? The answer is two-fold: (a) their role in supporting critical national security applications and (b) their role in underpinning geopolitically significant technology and critical infrastructure.
In defense and military applications, the power to compute faster than an adversary gives you a strategic advantage, whether it be outmaneuvering an adversary or having a higher success rate in general when attacking or defending. To that end, semiconductors are used for a diversity of purposes including communications, command and control systems, targeting systems, radar systems, high fidelity wargaming, and autonomous systems such as drones. For national security purposes like these, as Eric Burger, Research Professor of Computer Science at Georgetown University, pointed out, semiconductors—ultra high density central processing units (CPUs), graphics processing units (GPUs), and application-specific integrated circuits (ASICs)—must offer both high performance and reliability.
Yet, the national security importance of semiconductor chips goes well beyond the direct national security applications into the realm of national economic security, geopolitically significant technologies, and critical infrastructure and services. Take AI as just one example. Here, semiconductors are an essential driver of the evolution, proliferation, and functionality of current and future machine learning applications. Why? AI is underpinned by three mutually reinforcing pillars: data, algorithms, and computational power. While both data and algorithms have been the subject of increasing public debate, with data at times being dubbed the new oil, computational power is often overlooked. Yet, it is this computational power that makes the data and the algorithms come alive. Consequently, as AI systems become increasingly sophisticated, they also require increased computational power and, therefore, place increasing demands on the functionality of integrated circuits.
Importantly, this isn’t an either-or conversation: national security versus commercial applications. High performance computing tools are dual use technologies. That means that in supporting such technologies, we have the opportunity to create a positive reinforcement cycle leveraging both national security and commercial applications. According to Burger, “[the same technologies that] help build a better bomber can also help build a better transport plane. And digital signal processing and high frequency power amplifiers that build better radars can also power better Wi-Fi and 5G.” As a result, “more success with commercial chips means more experience designing chips, which lowers the unit costs or creates more value per chip, which means more profit, which results in more ability to pay for national security needs” such as more specialized chips for national security applications. Over time, this cycle deepens American chip design capabilities, lowers per unit costs, and increases chip functionality. Leveraging this dual use cycle simultaneously allows for a secure nation, which protects economic activity, but also increased economic activity and innovation, which supports a secure nation.
Ensuring the Availability and Security of Supply Chains and Bolstering U.S. Leadership
Given that semiconductors represent a critical dependency for the U.S., why should we be concerned about the state of the industry? This question takes two forms. First, how do we ensure the availability of semiconductors and the security of semiconductor supply chains? If the past year has demonstrated anything, it’s how critical supply chain security and resilience has become, particularly given the recent automotive chip shortages. Second, how do we maintain and bolster U.S. leadership going forward?
The semiconductor supply chain is long, complex, and global. This includes, according to Meg Hardon, Vice President Government Relations at Infineon Technologies Americas Corp, “anywhere from 400 to 800 process steps and a timeline of 6 to 18 months in multiple locations around the world.” As a result, there is limited flexibility in the semiconductor supply chain, both in terms of its ability to adjust to sudden shifts in demand but also in its ability to rapidly compensate for decreases in functionality or capacity at various stages across the supply chain.
This inflexibility drives home the necessity of making deep investment in the availability and security of supply. Doing so means implementing best practices across industry including but not limited to cybersecurity, business and operational continuity, and environmental compliance. As Hardon pointed out, this requires standardized practices with built-in accountability such as “backup production sites, long term contracts with our suppliers with a very holistic approach and asking them to mirror our own best practices for secure supply chains, manufacturing in multiple locations [around] the world, [and] having plant harmonization practices so that we always have real time visibility into our production capacity and can deliver when we promise.”
Yet, despite the challenges of pursuing the twin goals of availability and security within this complex, global ecosystem, much of our focus in the U.S. to-date has been two-fold: (a) on competition with China and (b) on a lack of domestic high end manufacturing capabilities at scale.
China, a near-peer technology competitor, is investing heavily in this space. However, the concern here is less that the ‘sky is falling’ and more that ‘the writing is on the wall’. When we look across the entire semiconductor ecosystem, or stack, a less dire geostrategic picture emerges. For example, while China has an advantage in input materials (namely, rare earth materials) and has invested significantly in tools for assembly and packaging, the U.S. is a leader in the area of intellectual property, namely R&D, licensing, and design tools. The U.S., Japan, and the Netherlands dominate the manufacturing of the equipment necessary for production of semiconductors, while Taiwan’s TSMC and South Korea’s Samsung lead in manufacturing at scale at the most advanced process nodes, followed closely by the U.S.’ Intel. In short, the U.S. and our allies today maintain significant points of leverage over many of the segments of the semiconductor supply chain. The live question is whether - and if so, how - we will continue to do so in the future. And the answer to that question is as dependent on the choices we make as it is on the choices China makes.
The second line of concern has been on high end production at the most advanced nodes (7nm and moving toward 5nm). Why? Most of the state of the art semiconductor production sites have moved overseas due to a combination of environmental concerns, foreign government policy, rising production costs, and lower cost of labor, and resultantly, the U.S. really has one state of the art foundry: Intel. A foundry manufactures chips designed by and for companies that lack their own fabrication capabilities. Notably, Intel, in addition to making chips for others, has also historically coupled in-house chip design with its own in-house production. Indeed, collectively, the U.S. faces significant gaps in our domestic capabilities around the production, if not the design and manufacturing equipment that support production, of integrated circuits at scale. Manufacturing matters. High end production matters. But it isn’t the only thing that does.
Focusing solely on this issue fails to capture the broader geostrategic stakes at play and areas of concern, notably the ability to maintain leadership over time. As Eileen Tanghal, Senior Partner at In-Q-Tel, noted “production capacity is not going to be and should not be the sole measure of leadership. It is helpful from a national security supply chain point of view” but such a measure misses a broader assessment of the entire semiconductor stack. That assessment should include, for example, design tools, raw materials, manufacturing equipment, and packaging.
This narrow approach also risks overlooking concerning and persisting gaps in our current approach, such as the role of startups in this ecosystem and the need to provide more funding for companies seeking to scale (the middle of the funnel) rather than at the seed stage (the beginning of the funnel) or at the point of acquisition when the technology is proven (the end of the funnel). Tanghal further illustrated this concern by highlighting that from 2015 to 2017, while most funders for late stage companies were American, 66% of the funders were foreign for these middle stage companies. In other words, U.S. companies raising Series-B and Series-C stage capital were forced to rely on foreign sources 66% of the time. This is just one of many concerning gaps in the U.S. policy ecosystem to-date, and yet are simply overlooked in the ‘build manufacturing at scale capacity in the U.S.’ conversation.
A Call to Action: Lessons for Policymakers
Develop a Cohesive and Comprehensive National Strategy: from anti-dumping laws to export restrictions and from the CHIPS Act to the recent Executive Order on supply chains, the U.S. has and continues to pursue many policy opportunities at the government’s disposal. Despite these efforts, we have yet to craft and commit to a sustained, comprehensive, and cohesive national strategy that addresses the industry as a whole and the corresponding national security concerns in depth. There are numerous levers at our disposal, many of which we are already pulling. Now, we must pull them strategically and in concert or we risk facing a national security landscape in the future that is far more concerning than the one we face today.
Avoid Decoupling: the semiconductor industry is far too complex and far too global for an America-only approach, a concern one of us (Griffith) has spoken on before: “[s]uch a model simply cannot support the current and future demand for integrated circuits both in the commercial and military spaces.” While there are strong national security incentives and opportunities to bolster American semiconductor leadership, no single country or region can do it alone. In the wake of China’s drive toward self-reliance, the EU’s digital sovereignty movement, and our own increasing focus on domestic production and capacity, it behooves us to remember that significant global decoupling is neither feasible nor desirable. Moreover, while there are significant national security concerns associated with relying on near—peer competitors (or geopolitical rivals) for foundational technology, the U.S. should avoid the misperception that we must solve this problem alone. We have allies—very capable ones at that—occupying positions of strength across the semiconductor stack. We can and should lean on them.
Mind the Gaps: given how complex this ecosystem is, the solutions need to span the full stack of industry segments including R&D, design, production, distribution, and finally, integration into use cases. Our focus on domestic production at scale has proven too narrow. Yes, the U.S. needs to continue increasing its domestic manufacturing and production capacity, but we must also look to innovate and invest across the ecosystem (including equipment, raw materials, and packaging) and across a diversity of players and stages of innovation (from investing in early stage research to the middle of the funnel and expanding our focus from incumbent industry players to include the wider range of players, such as startups). We should also dare to dream big, and invest in innovation related to the use cases semiconductors support, such as AI, quantum computing, and 5G. After all, it is that very demand that will continue to fuel innovation around and investment in semiconductors well into the future.
Act Now: although the sky may not be falling just yet, the window of opportunity to shape the semiconductor ecosystem is certainly closing on the U.S. As Burger echoed, “the best time to have started building a new factory was ten years ago.” In order to maintain American leadership in the semiconductor industry, we need to step up the innovation and the capital today, or risk being sidelined by other players who have already made and continue to make those same commitments domestically.