Pages

15 January 2023

The Microchips Are Down

EDOARDO CAMPANELLA

CAMBRIDGE – “Potato chips, computer chips, what’s the difference?” a top economic adviser to US President George H.W. Bush supposedly asked in the early 1990s. “A hundred dollars of one or a hundred dollars of the other is still a hundred dollars.” At the time, Japanese firms were pushing their American competitors out of the market for memory chips, but free-market elites in Washington, DC, staunchly opposed any form of industrial policy to protect the domestic semiconductor industry. If foreign companies could produce chips at a lower price, they argued, American consumers would pocket the cost savings and direct their spending to other sectors.

Chipmakers in Silicon Valley were understandably dismayed. But they were hardly the only industry suffering from Japanese competition, and the federal government could not afford to protect them all (nor could it risk the political blowback of saving some but not others). In the years that followed, the structure and composition of the US economy evolved accordingly. The American share of global semiconductor manufacturing dropped from 37% in 1990 to 12% today. Asia now accounts for over 70% of semiconductor production, with the most advanced chips being produced exclusively in Taiwan and South Korea.

Yet not until the COVID-19 pandemic did the United States (and the rest of the world) seem to realize that computer chips really are quite different from potato chips. Because semiconductors are used in a vast range of goods – from computers, smartphones, and coffee makers to toys, automobiles, and weapons systems – the global chip shortages of 2021 cost the US economy $240 billion (around 1% of GDP). The global auto industry alone produced 7.7 million fewer cars than it would have, and the health-care, defense, space, and energy industries all suffered significant losses.

By triggering a cascade of production stoppages and backlogs, the chip shortages helped unleash the inflationary pressures that central banks are now struggling to suppress. As part of a broader policy response, both Europe and the US are implementing measures to strengthen their domestic production capabilities. The European Chips Act, announced last February, aims to mobilize billions of euros worth of public and private investment to preempt future supply-chain disruptions.

Similarly, the US CHIPS and Science Act, which was signed into law last August, promises to foster a domestic manufacturing renaissance. In parallel, the US has weaponized the semiconductor industry in its strategic rivalry with China, imposing sweeping export controls to block Chinese companies from accessing the most advanced chips and chip-making equipment.

As the semiconductor industry has evolved, it also has become one of the most important geopolitical and economic issues of our time.

SILICON EL DORADO

Chips are the building blocks of artificial intelligence, 5G communications, the Internet of Things, quantum computers, and other technologies of the future. In line with Moore’s Law (which is actually just an astute observation), the number of transistors on a single microchip has roughly doubled every two years since the 1960s, yielding a one trillion-fold increase in computing power. A modern smartphone has around 100,000 times more processing power than the computer used for the Apollo moon missions.

Although part of this story is global in nature, such extraordinary progress would not have been possible without the visionary entrepreneurs who made Silicon Valley what it is today. Nonetheless, in The Code, historian Margaret O’Mara of the University of Washington dismantles the Valley’s own founding mythology of lone geniuses changing the world from their garages. “Silicon Valley is neither a big-government story nor a free-market one,” she writes. “It’s both.” It would not exist without its leaders, but nor would it have achieved so much without the publicly funded research and innovation that sent humanity to the Moon and gave rise to the internet.

To be sure, this argument is hardly new. But O’Mara is a talented writer, and she brings meticulous research to bear on the story. While Silicon Valley’s denizens may be singularly focused on disrupting the status quo and inventing the future, those who want to understand their world need to start with its deeper history.

That history starts in 1939, when two Stanford University graduates, William Hewlett and David Packard, created a series of improved vacuum tubes for radio and other electronics applications. They founded a company together in a garage at 367 Addison Avenue in Palo Alto – now considered the birthplace of Silicon Valley – and their timing was impeccable. Thanks to hefty orders from the US Department of Defense during World War II, the Hewlett-Packard Company quickly rose to global prominence, facilitating the subsequent commercialization of the new electronics industry.

Meanwhile, Stanford’s provost, Frederick Terman, remade his institution by elevating the science and engineering disciplines with federal defense dollars, and by leasing university land to high-tech firms to create the Stanford Industrial Park. Among the firms to set up shop there were HP and a company owned by William Shockley, the co-inventor of the transistor. The semiconductor industry was beginning to take shape.

The turning point came in 1957, when a handful of Shockley’s most talented employees – the “Traitorous Eight” – left his company (because of his detestable management style) to found Fairchild Semiconductor. Fairchild owed its early growth to its role as a government contractor for the Apollo program in the 1960s. But over time, its founders led and invested in hundreds of start-ups that came to define the Valley – including Intel, which introduced the first microprocessor, and the venture-capital firm Kleiner Perkins.

For its part, the US federal government was not simply buying up chips and circuits from these companies. It was actively shaping incentives with the goal of advancing scientific research. Entrepreneurs were given ample freedom to push technological boundaries. “The US government,” O’Mara explains, “got into the electronics business and became, in a sense, the Valley’s first, and perhaps its greatest, venture capitalist.”

Of course, more recent denizens of the Valley have been in denial about the government’s role as a catalyst. In 1996, for example, future Apple CEO Steve Jobs boasted that “Silicon Valley doesn’t traditionally look for handouts.” And yet, his crowning achievement, the iPhone, would not exist without a suite of key technologies that had been funded by the federal government over the years.

CHIP WARS

Toward the end of her book, O’Mara notes that Silicon Valley is “no longer merely a place in Northern California,” but instead the hub of a “vast supply chain” stretching across three continents. In Chip War, Chris Miller, a historian at Tufts University, offers a detailed account of the industry’s evolution, including through interviews with some of the key players. Starting in the 1970s, semiconductor firms (like many others) found it advantageous to move production to East Asia, with its abundant cheap labor and looser regulation of the toxic chemicals involved in chipmaking. Once this trend started, fierce market competition took care of the rest.

When the US semiconductor sector was still at its infancy, in the 1960s, it faced almost no competition. Though the Soviets soon adopted a “copy it” strategy, this proved inadequate for a technology that was evolving at the speed of Moore’s Law. By the time the Soviets were mastering a new-model chip, it was already obsolete. But by the 1980s, Japan had emerged as a serious challenger. Intel (then under the leadership of Andy Grove) found itself being pushed out of the memory-chip market and had to reinvent itself as a microprocessor company.

That choice paid off and set the stage for the rise of the US computer industry. But the semiconductor industry, more narrowly, has evolved in a peculiar way. On the one hand, it is among the most internationally integrated sectors in the world, with the different phases of research, design, production, and assembly spread across North America, Europe, and East Asia. US semiconductor firms work with around 16,000 suppliers, on average.

On the other hand, it is also one of the most concentrated industries in the world. Just a few firms play a disproportionate role at each stage of the global value chain, owing to the complexity of the technologies and the size of the investments they require. Problems at any one of these companies can disrupt the entire industry.

While a handful of US companies control the upstream of the supply chain – particularly software design – there are two other key chokepoints in chip production. The Dutch company ASML has a monopoly over the cutting-edge lithography systems that are used to print the minuscule designs that determine what chips will do.

Similarly, Taiwan Semiconductor Manufacturing Company manufactures over 90% of the leading-edge chips that serve as the “brains” in electronic devices. (These are logic chips that process information to complete a task, as opposed to the less technologically sophisticated memory chips that store information.) At its most advanced facility, Fab 18, TSMC can make cutting-edge chips with features as small as three nanometers – that is about one-fortieth the size of the coronavirus. And, underscoring the fragility of the supply chain, ASML and TSMC’s fates are intertwined, because the latter would not be able to operate without the lithographic technologies sold by the former.

TAIWAN’S SILICON SHEEN

TSMC was founded in 1987 by Morris Chang, who was born in China but spent a large part of his career working at Texas Instruments in the US, where he played a key role in optimizing the production of chips. After being passed over for the role of CEO at TI, he decamped to Taiwan to establish his chip empire.

As with Silicon Valley, public money was fundamental. Government support was integral to the success of semiconductor firms that took root not just in Taiwan, but also in Japan and South Korea. If Chang had been offered that C-suite job in Texas, Taiwan would probably be far less relevant – geopolitically and economically – than it is today. The current Sino-American tensions over the island stem partly from China’s domestic politics, but also from US and Chinese fears of losing access to the most advanced chips. Taiwan’s hope is that its “silicon shield” will deter outright Chinese aggression and induce the West to protect it. But as China proved with its military exercises following then-US Speaker of the House Nancy Pelosi’s visit to Taiwan in August 2022, it can effectively prevent goods from leaving the island.

Were China to invade Taiwan, even with the express intent of seizing TSMC, the firm would hardly remain viable, and the global economic impact would be profound. As TSMC Chairman Mark Liu pointed out last August, “nobody can control TSMC by force.” An invasion would render the company’s factories in Taiwan “non-operable,” because chip production depends “on the real-time connection with the outside world: with Europe, with Japan, with the US.” Moreover, in the event of an invasion, the Taiwanese might simply destroy the fabs.

Companies and governments are increasingly aware of these risks – and increasingly concerned about Taiwan’s centrality in the industry. Private firms are trying to diversify their supply chains away from the island by striking agreements with other Asian countries, and governments are luring firms as they try to rebuild their domestic capabilities. US President Joe Biden’s administration recently persuaded TSMC to build new fabs in Arizona, deploying a form of government support that echoes the industrial policies adopted by Asian governments in the 1980s.

THE SANCTIONS TEMPTATION

The US has accelerated this re-shoring process by weaponizing its own position in the global semiconductor industry. According to Biden’s national security adviser, Jake Sullivan, the US can no longer assume that it is sufficient merely to stay a couple of generations ahead of its competitors in key technologies. “Given the foundational nature of certain technologies, such as advanced logic and memory chips,” he said last September, “we must maintain as large a lead as possible.”

Thus, with its October export ban, the Biden administration targeted sales of high-performance chips to China. US companies are no longer allowed to supply advanced computing chips, chip-making equipment, or related products to Chinese buyers except under a special license. The ban also applies to foreign companies that use US semiconductor technology.

The goal is to establish a stranglehold on advanced computing and semiconductor technologies. Since China’s chip producers specialize in manufacturing less-advanced microprocessors (for use in appliances such as refrigerators and washing machines), they do not have the ability to produce the leading-edge chips used in AI applications. In depriving China of high-end chips and the personnel and equipment needed to build them on its own, the US is effectively blocking China’s technological and economic development path.

In Backfire, Agathe Demarais, the forecasting director of the Economist Intelligence Unit, highlights the risks of such measures. Though her manuscript was completed before the chip export ban, her argument is prescient: sanctions have a poor success rate, high economic costs, and massive unintended consequences. To be effective, they should be targeted, short-term, and backed by allies.

None of these conditions are met by the chip export ban. Its goal – suppressing China’s economic rise – is both sweeping and long-term, and US allies were not consulted about it ahead of time. Worse, such sanctions will invite retaliation: further escalation of tensions over Taiwan, a cutoff of rare earth minerals (81% of which come from China), or a push by China to develop its own domestic capabilities.

True, without US know-how, Chinese firms cannot produce the smallest and most advanced chips. But, according to one credible analysis, China could still rely on non-American suppliers for around 70% of its microchip needs. Policymakers are also channeling investment to indigenous companies focused on chip and software design and semiconductor manufacturing equipment. And beyond trying to create national champions in these fields, the Chinese leadership is creating an ecosystem of smaller, highly specialized industries that are key for chip production – such as vapor deposition, wafer cleaning, rapid thermal processing, and metrology.

Under President Xi Jinping’s Made in China 2025 plan, the country aims to be semiconductor self-sufficient by mid-decade, with domestic suppliers meeting 70% of its needs. To support this effort, the government is leveraging state-backed investment funds to provide capital for homegrown semiconductor development and manufacturing. Around $120 billion has already been committed to achieving technological parity with the US by 2035.

ONE TECH, TWO SYSTEMS

The competition over semiconductors is part of a broader geopolitical battle. Chips provide the computing power that is necessary to analyze data, and the more data you produce, the more computing power you need. China produces a great deal of data, which is an advantage in AI. But without access to the most sophisticated chips, it cannot remain at the forefront of AI developments (which in turn are used to bolster its data-hungry surveillance state).

In The Wires of War, Jacob Helberg, the former global lead for news policy at Google, discusses how “techno-totalitarian” regimes are trying to use the hardware and software of the internet to divide the world into twentieth-century-style spheres of influence. Helberg describes how malign foreign powers use the internet to conduct asymmetric warfare. The “software war” consists of social-media misinformation, whereas the “hardware war” is more about illegally accessing tech devices and private information. Russia stands out in the former, and China in the latter. America’s own reshoring of semiconductor production is a clear example of such strategic positioning. Unlike in the 1980s, when economic competition from Japan reshaped the semiconductor industry, the main driver of change nowadays is geopolitical rivalry. Countries are competing over access to the most advanced chips, and governments are becoming active planners of the industry, rather than just its customers. Even in Washington, industrial policy is no longer anathema.

As a result, supply chains are likely to become shorter and more regional, implying higher production costs and substantial efficiency losses. Taiwan’s central role will probably be considerably diminished. Foreseeable technical constraints will transform the industry once again, because the miniaturization of chips will hit its physical limits (Moore’s Law, eventually, must end).

For Silicon Valley to preserve its supremacy, staying ahead of the curve on innovation will not be enough. It also needs to strengthen its partnership with the federal government. It needs to refocus on the production of physical goods, rather than just digital services. And it needs to identify the potential sources of future crises so that it can prepare for the kind of abrupt adaptations that the new geopolitical environment might require. As Grove put it when recounting his radical overhaul of Intel under the pressure of Japanese competition, “Only the paranoid survive.” The same is true for today’s tech entrepreneurs. Disruption is a two-way street.

No comments:

Post a Comment