A brief history of semiconductors: The United States cuts costs and gradually loses its lead

Today, the United States is still the dominant player in the development of the semiconductor industry. At the beginning of the development of the semiconductor industry, how did the United States develop and obtain its current status? ICViews has compiled a brief history of the development of semiconductors in the United States, hoping to find some answers from the development process of semiconductors in the United States.

Early U.S. industrial policy provided roles for a variety of actors: small companies experimenting at the technological frontier, while large companies pursued process improvements to scale these innovations. U.S. government demands ensure that experiments are financially feasible, while technology transfer regulations ensure that both large and small companies share progress. Importantly, recurring purchases provide businesses with the liquidity they need to continue iterating without having to rely on large-scale, one-off products. This type of industrial policy encourages innovation by ensuring that small companies have access to domestic mass production of innovative designs, while allowing large companies to reap the benefits of mass production of these innovative designs.

As the industry matures and the competitive environment changes, the U.S. policy framework has also changed.

Since the 1970s, industrial policy has been gradually replaced by capital-light "scientific policy" strategies, and huge leading companies and asset-light innovators have replaced a system of large and small production enterprises. Strong ecosystem. While this strategy was initially successful, it has created a fragile system. Today, the semiconductor industry is constrained by fragile supply chains that are tailored to only a few companies with large capital chains on the one hand, and many asset-light design companies on the other.

Although the U.S. semiconductor industry regained its dominance in the 1990s, due to this policy approach, the technological and commercial advantages of the U.S. semiconductor industry are now more fragile than before. As TSMC rises above Intel, the United States has lost cutting-edge technology and American companies face critical supply bottlenecks. The supply chain issues exposed by the epidemic show that semiconductors, as a universal technology, play a role in almost all major supply chains, and that semiconductor production is a vital economic and national security issue. Although policy can play an obvious role, it has its limitations on the process of technological progress, supporting the development of new ideas rather than diverting new technologies to capital. Innovation in process technology is a practical process that requires the continuous establishment and operation of new production lines. But in the low-capital environment of the United States, it is difficult for the semiconductor industry to learn by doing.

Every part of the semiconductor supply chain has technological innovation and benefits from diverse players and a dynamic labor market. Labor is not only a cost center at the technological frontier but also a key input to the innovation process. In addressing current shortages, policymakers should recognize the lessons of semiconductor industry policy and create a robust competitive ecosystem to spur innovation.

In the early days of the semiconductor industry, the U.S. government used industrial policy and scientific policy to help cultivate a diverse ecosystem of semiconductor companies. Fiscal spending provides necessary liquidity to this highly speculative industry. To maintain an innovative and vibrant competitive ecosystem, strategy also requires ongoing intervention.

The U.S. Department of Defense (DoD) uses procurement agreements and quasi-regulatory measures to ensure the broad dissemination of company ecosystems and technological advancements. U.S. government contracts create a ready market for early-stage companies, with the U.S. Department of Defense eager to play the role of first customer. Convinced that there will be demand for large-scale semiconductor production, investment in capacity is financially feasible for many early-stage small companies.

As a core customer for many companies, the U.S. Department of Defense has a clear view of the latest technological developments in the industry and uses this view to directly promote dialogue and knowledge sharing between companies and researchers. Meanwhile, "second source" contracts require that any chips purchased by the Department of Defense must be produced by at least two companies, linking procurement to technology transfer. The U.S. Department of Defense even requires Bell Labs and other large R&D departments to publish technical details and broadly license their technology to ensure that all companies that may contract with the U.S. Department of Defense have access to the cornerstone of innovation.

This system accelerated the pace of innovation in the industry and spread rapidly. Government procurement agreements ensured investors' willingness to spend and also increased spending on capital goods for recurring production, helping the process to improve significantly.

At the same time, workers move freely throughout the system, and knowledge gained at one company can be applied to improve production processes in other companies.

In this competitive environment, combined with the antitrust practices of that era, large companies were encouraged to develop large research laboratories and small companies were encouraged to conduct crazy experiments. Successful experiments help create new large companies or are scaled up by already existing large companies. Industry guidance from the U.S. Department of Defense helps drive technology in new directions while keeping industry capabilities consistent and targeted. Crucially, this strategy implicitly prioritizes the development of new technologies across the sector, rather than maximizing revenue or minimizing costs for any one company. There are also sources of financing if the company needs to invest and hold capital goods. The government protects the industry from so-called "market constraints" so that the industry focuses on innovation and production rather than narrow economic success.

By the late 1960s, however, the industry had grown so rapidly that government procurement and the government's ability to implement quasi-regulation through second-source contracting, for example, had become relatively unimportant. In the late 1940s, the semiconductor industry's existence was based on military procurement, but by the late 1960s, military procurement accounted for less than a quarter of the market.

The 1970s: Booming Commercial Markets

This period, despite the relative unimportance of U.S. government procurement and guidance, was dominated by a boom in commercial applications and a lack of serious international competition. , domestic semiconductor companies in the United States ushered in a golden age.

Although industrial policy promoted early innovation and capacity building, its relative absence during the 1970s was barely noticed. To be sure, government procurement still played a role in the 1970s, but as private companies began to incorporate electronics into their supply chains, they became more important purchasers. The beginning of mass production of computers was also closely related to the development of semiconductors, as the demand for chips drove advances in packaging and integration.

In fact, there is a divergence between the priorities of the US Department of Defense and the priorities of commercial customers. The U.S. Department of Defense seeks appropriate solutions to specific military problems, particularly the development of non-silicon-based or space-grade semiconductors that have small commercial applications. Both the government and semiconductor companies recognize that the industry no longer needs direct guidance. Therefore, the needs of both parties began to diverge.

In the 1970s, booming non-defense markets meant that successful small and large companies could survive without government support or coordination. Improvements in technology translate into improvements in processes, which in turn drive further improvements in the former. New inventions such as MOS IC, microprocessors, and DRAM have pushed the industry to new heights and recursively proposed further innovation paths.

In an environment of widespread prosperity and innovation, semiconductors have demonstrated their importance as a general-purpose technology, finding widespread use throughout the economy. Although large U.S. research labs and manufacturing sectors held significant assets, the lack of international competition and the booming market ensured that most investments were ultimately viable, both in terms of innovation and profitability.

The 1980s: fierce international competition

However, the optimism brought about by this competitive environment was interrupted in the 1980s, when the Ministry of International Trade, Industry and Trade of Japan Under the guidance of industrial policies, the United States ceded its market and technological dominance to Japanese companies.

The U.S. government initially had to create a market for semiconductors, whereas Japan was able to shape industrial policy around a rapidly growing market that already existed. As a result, Japan has been able to adopt far more stringent infrastructure-building policies than the United States and coordinate joint ventures in computers and semiconductors because it knows there is a ready commercial market for its products. While the strategy of government support and coordinated investment was the same one used by the United States in the 1950s and 1960s, the tactics used to implement it were tailored to the competitive environment of the 1980s.

Competition from Japan has had a huge impact on American companies. In the subsequent market turmoil, many people permanently exited the DRAM market. Industry has also formed advocacy groups to coordinate production and lobby for government intervention on tariffs and implementation of trade policy.

The Semiconductor Industry Association lobbied for protective measures against Japan's "dumping" and established the Semiconductor Research Corporation to organize and fund academic research in semiconductor development that was relevant to the commercial market but not related to the U.S. Department of Defense. The Semiconductor Manufacturing Alliance is jointly funded by industry members and the U.S. Department of Defense. Its initial purpose was mainly to use earlier industrial policies to promote horizontal cooperation between companies. However, in order to minimize costs, the alliance soon turned its focus to vertical integration between suppliers and manufacturers.

Lagging semiconductors have become commodities, interchangeable and judged based on unit cost. Due to a combination of technological and economic factors, traditional vertically integrated companies began to unravel in the 1980s. Given the economic situation in the United States at the time, there was little interest in investing in capacity for low value-added activities in a much more competitive global market.

Instead, large companies absorbed the productivity that small companies still possessed, creating conglomerates. With the emergence of the MOS transistor as the industry-leading design, companies began to adopt similar design principles, making it economical to specialize in manufacturing "foundries." The subsequent vertical disintegration has led to the emergence of large, vertically integrated conglomerates, alongside smaller design-focused “fabless” companies that design but do not produce chips. In theory, these "fabless" companies pursue innovative design strategies while minimizing costs and retaining flexibility. In the 1990s, as American companies pioneered new product categories and Japanese companies faced competition from South Korea, the U.S. industry's embrace of this strategy led to a recovery in market share.

In terms of policy, the United States has never returned to domestic industrial policy. Instead, the success of foreign industrial policy programs is achieved through a combination of domestic consolidation, monopoly, trade protectionism, and funding for scientific research.

The 1990s: Science Policy, Not Industrial Policy

In the 1980s the industry faced changes in technology and the competitive environment, and the 1990s witnessed the new " Science Policy” reaches its climax. In the 1990s, whether it was the kind of policies adopted by the United States in the past or practices more influenced by Japan's Ministry of International Trade and Industry, the United States did not return to industrial policy. Instead, it regarded the introduction of "science policy" as the government's role in semiconductor manufacturing. A new paradigm for action in the field. The focus of science policy is to promote public-private cooperation with individual companies, integrate industry R&D with academic R&D more closely, ensure the breadth of research power, and form an industry structure that can support innovative companies with light-asset operations.

The policy goal shifts from creating a strong competitive ecosystem with strong supply chains to creating public-private institutions to coordinate between researchers, fabless design companies, equipment suppliers and large "champions" complex switching. In this way, no business needs to invest excessively in R&D to remain globally cost competitive, and governments can avoid large-scale investment expenditures. The chart below, from the 1994 National Semiconductor Technology Roadmap produced by the Semiconductor Industry Association, illustrates the strategy behind science policy:

The central theme of "science policy" is non-redundant efficiency, which is consistent with the early 's industrial policy focuses on redundancy and duplication, in contrast. Early industrial policy greatly accelerated the pace of innovation and ensured that the failure of a single company did not affect the robustness of the supply chain, but it did mean significant duplication of investment. While this approach helps drive the adoption of process improvements, static shareholder value maximization suggests that such duplication is economically wasteful.

Industrial policies over the past few decades have promoted large-scale employment, which is a core driver of innovation. The “science policy” of the 1990s avoided this approach for the sake of minimal efficiency. Employees change companies frequently, and learning by doing is the core approach to innovation. Indeed, the “non-transactional interdependence” literature in Economic Geography goes some way to explaining how the convergence of groups of workers in the semiconductor industry is important to the industry’s fast-paced innovation. While keeping a large number of workers in one place is key to many advancements, in this new competitive environment this is seen as wasteful. Labor accounts for a sizable proportion of unit costs, and companies believe that if they can strategically downsize, global competitiveness will be restored.

In the early days of the semiconductor industry, relatively price-insensitive government contracts accounted for a large portion of total sales, and this inefficiency was seen as a cost of innovation.

With the entry of foreign competitors and the cost-sensitive commercial market becoming the main buyer of semiconductors, the duplication of this capability seems like a pure cost center with little benefit to many companies. Concerns about profitability mean ensuring there is as little duplication of effort as possible in order to control costs in a price-sensitive, competitive environment. This creates a collective action problem, where cutting spending is in every business's interest, but doing so further worsens the ability of American businesses to innovate.

In the 1990s, instead of returning to industrial policy, the U.S. government opted for much lower-cost science policy programs. Ideally, "science policy" would allow governments to reconcile businesses' conflicting desire to save without falling further behind technologically. However, in order to conform to the spirit of the times, the US government is also trying to save and will not provide the large-scale financial support needed for industrial policy to succeed in the new competitive environment.

Instead, the government will spend less money and try to create a division of labor that allows all players to cut costs in pursuit of profits without sacrificing the technological frontier. It does this by funding R&D in academic research laboratories on the one hand, and industrial groups to translate research into commercial capabilities on the other. To some extent, this further reduces individual firms' R&D investments, as advances create only minimal competitive advantage. Rather than creating an ecosystem with overlapping supply chains, this structure creates a division of labor in which each business and institution is responsible for a distinct and separate part. At the same time, loose trade policies and close trade networks allow companies to enter the factory-free model more economically and develop asset-light strategies. The aim is to regain the technological frontier in the most economical way for the public and private sectors by solving a collective action problem and reducing redundancy across the system.

In the short term, this strategy worked! By the late 1990s, investment in U.S. semiconductors and other technology fields generally flourished, and the United States successfully restored its technological advantage. The industry was able to innovate while remaining internationally competitive without requiring large-scale financial support from domestic industrial policy. Most individual companies focus their R&D efforts on the next node or two in production process development, while longer-term research is organized by government-funded academic researchers. Industrial groups stepped in to translate this academic research into commercial activities and largely eliminate the duplication of labor costs in R&D and production. Large, centralized research laboratories were hollowed out and supply chains became narrower, targeting only the research needs of a few core companies.

The 21st Century: The Dot-com Bubble and Diminishing Returns

However, the short-term success of this strategy comes at a huge long-term cost. Redundancy in labor and capital helps ensure that companies can quickly improve internalized processes while also training the next generation of engineers and technicians. While this duplication may be redundant from the perspective of static maximization of shareholder returns in a single period, it is critical to ensuring a long-term innovation trajectory. "Eliminating redundancy" and "increasing vulnerability" are two sides of the same coin.

In the long term, underinvestment in labor and capital will show up in some way, whether on the balance sheet, in innovation capabilities, or both. As things stand, the United States is in danger of losing its edge in cutting-edge design, and has largely lost its dominance in cutting-edge manufacturing to TSMC. Allocating part of the investment process to each company might make each company's balance sheet look stronger, but the industry as a whole has become more fragile due to continued underinvestment. Decades of labor cost minimization have reduced the number of skilled technicians and engineers, while decades of underinvestment in capacity have hampered the ability of domestic companies to cope with the current labor shortage.

The industry's current problems are a long-term natural outcome of the science policy strategy that seemed so successful in the late 1990s and early 2000s. The drive for consolidation and vertical integration has focused on long-term research in academic labs, massive “champions” and asset-light “fabless” innovators, creating a shaky competitive ecosystem.

Because these champion companies account for a very large proportion of the competitive landscape, their R&D priorities and intermediate investment requirements set conditions for the entire industry. Big buyers like Intel can use their relative monopoly power, explicitly or covertly, to structure supply chains around their needs.

These fragile supply chains are prone to problems when broader economic demand shifts, such as since the beginning of the pandemic. This vulnerability is the result of supply chain optimization, but this optimization is geared toward short-term profitability and the elimination of redundancies rather than the needs of the economy as a whole.

Whether intentionally or unintentionally, these large companies will also formulate technological development paths around their own financial needs and plans. Thus, a policy mix of R&D in academic laboratories combined with tax optimization and private sector unit cost minimization creates significant technological path dependence. At the same time, these businesses are, in a technical sense, “too big to fail”: if they miss out on process improvements, the absence of similarly sized domestic competitors means that the entire industry misses out on that progress. In this sense, technology policy as a whole is entrusted to private actors.

Inconsistent feedback also occurs in the process from R&D to production. The key to science policy is to separate innovation in intellectual property from innovation in production processes; that is, science policy prioritizes research, design and creativity over implementation, production and investment. As a result, fabless companies have emerged that focus on design and outsource manufacturing to foundries overseas.

However, prioritizing R&D will slow down innovation. Simply subsidizing R&D is no different from incentivizing offshoring: the policy rewards the development of intellectual property rather than the ownership of tangible assets. The problem is that process improvements come from the implementation of new technologies embedded in new physical assets. “Learning by doing” is a key part of technological innovation. Good engineers want to innovate every step of the production process at every link in the supply chain. Offshoring and outsourcing production of cutting-edge designs introduces a black box around the process, where similar problems that prevent revenue maximization cannot be corrected. Focusing solely on R&D would offshore the development of these process improvements, starving domestic producers and preventing the workforce from developing new skills.

Academic research has deviated from the path of commercialization and cannot drive industrial innovation. Given that academic research often revolves around problems with low relevance to current production, it sometimes fails to provide insights into alternative applications of existing technologies or alternative process-driven innovation paths. As science policy puts this group in charge of long-term innovation strategies for the entire industry, this blind spot cannot be ignored. In fact, the failure of Moore's Law and the unique shift to heterogeneous chip design in many applications is a good example of the innovation that often implies technological development at any time.

Decades of failed investments in industrial capacity and jobs have created a situation in which American businesses are highly dependent on outside manufacturing plants. TSMC’s current plans to invest in a local manufacturing facility in Taiwan illustrate the company’s attempt to solve this problem through acquisitions without reducing our reliance on a single supplier for leading designs. Instead, we should review the history of industrial policy in the early days of semiconductor production, recapture the technological frontier, and promote innovation at every node of the supply chain.

Today, the United States faces semiconductor shortages and waning innovation capabilities, and policymakers are considering serious intervention. While it may be too late to address the current shortage, the next one can be prevented. Broad bipartisan support for infrastructure spending in the United States, the need to build back better after the pandemic, and national security concerns about semiconductor procurement should encourage policymakers to believe that now is the time for ambitious reforms. As noted above, the history of semiconductor industry policy provides many lessons on how best to create high employment, technological innovation, and strong domestic supply chains.

History shows that science policy is a necessary complement to industrial policy, but it is not sufficient by itself. Coordinated research and development is an essential part of any solution, but it is not the whole solution. To capture process improvements and ensure the workforce is sufficiently skilled to operate at the cutting edge of technology, the industry needs to see continued capacity expansion. However, as we have shown previously, in a low-demand environment there is a clear reluctance by private firms to make uncertain investments. Industrial policy, through a combination of government procurement and financing guarantees, direct financing, etc., is the only way to provide sufficient liquidity to the industry to ensure that capacity expansion is fast enough and that the industry remains at the technological forefront. At the same time, the government has the financial ability to allow domestic companies to produce backward semiconductor products to ensure national security and supply chain flexibility. In the long run, an industrial outsourcing policy aimed at maximizing shareholders has not yet been formed.

It is also important to recognize that strong economic demand and consequent tight labor markets, particularly for semiconductor production, are critical to the success of these policies. Strong government-led investment will create good jobs for people of all experience and skill levels. This will create a highly skilled workforce and ample opportunities for learning-by-doing that drives meaningful process improvements. In high-skill, high-capital-intensive industries, labor is almost like another form of capital good, paying clear dividends on investment. However, in the absence of adequate employment opportunities, these specialized skills disappear as workers move to other industries. This is not to say that upskilling the workforce is enough: legislation that creates training programs without simultaneously creating the necessary jobs and investments can quickly become self-defeating.

Some may balk at the scale of investment needed for industrial policy in semiconductors and other key industries. It's a huge market with a huge price tag, with modern manufacturing plants costing billions of dollars. However, semiconductors are a critical, versatile technology that enters nearly every supply chain. Large-scale industrial policy can prevent bottleneck periods from dragging down economic growth while creating a strong domestic supply chain for national security needs. Relative to the initial investment in semiconductor technology, the cost of returning to industrial policy is much higher, but the return will be higher. Revitalizing lagging and leading industries and restoring a strong competitive ecosystem as part of a $4 trillion infrastructure or bipartisan supply chain bill is an investment not to be missed.

The policy goal is simple: develop an expanded industrial policy toolkit that encourages innovation, tight domestic labor markets and preserves critical supply chain infrastructure. Semiconductors, as an industry, are an ideal starting point for developing these policy tools due to the scale of investment and the jobs required. Rebuilding a strong innovation environment will also help the United States return to the technological frontier in a lasting way and create jobs and investments that will pay dividends for years to come. Semiconductors play a vital role in the modern industrial economy, and their technological routes are too important to be guided by short-term profitability. The government has an opportunity and a responsibility to use industrial policy to stop the next shortage before it occurs while ensuring that the United States maintains its position at the technological forefront.