Chip War: The Fight for the World's Most Critical Technology

by Chris Miller

Huawei discovered that, like all other Chinese companies, it was fatally dependent on foreigners to make the chips upon which all modern electronics depend. The United States still has a stranglehold on the silicon chips that gave Silicon Valley its name, though its position has weakened dangerously. China now spends more money each year importing chips than it spends on oil.

Most of the world’s GDP is produced with devices that rely on semiconductors. For a product that didn’t exist seventy-five years ago, this is an extraordinary ascent.

Chips from Taiwan provide 37 percent of the world’s new computing power each year. Two Korean companies produce 44 percent of the world’s memory chips. The Dutch company ASML builds 100 percent of the world’s extreme ultraviolet lithography machines, without which cutting-edge chips are simply impossible to make. OPEC’s 40 percent share of world oil production looks unimpressive by comparison.

Fairchild was the first semiconductor firm to offshore assembly in Asia, but Texas Instruments, Motorola, and others quickly followed. Within a decade, almost all U.S. chipmakers had foreign assembly facilities. Sporck began looking beyond Hong Kong. The city’s 25-cent hourly wages were only a tenth of American wages but were among the highest in Asia. In the mid-1960s, Taiwanese workers made 19 cents an hour, Malaysians 15 cents, Singaporeans 11 cents, and South Koreans only a dime.

The semiconductor industry was globalizing decades before anyone had heard of the word, laying the grounds for the Asia-centric supply chains we know today.

“The catalyst is the microelectronics technology and its ability to put more and more components into less and less space.” Industry outsiders only dimly perceived how the world was changing, but Intel’s leaders knew that if they succeeded in drastically expanding the availability of computing power, radical changes would follow. “We are really the revolutionaries in the world today,” Gordon Moore declared in 1973, “not the kids with the long hair and beards who were wrecking the schools a few years ago.”

In the early 1960s, it had been possible to claim the Pentagon had created Silicon Valley. In the decade since, the tables had turned. The U.S. military lost the war in Vietnam, but the chip industry won the peace that followed, binding the rest of Asia, from Singapore to Taiwan to Japan, more closely to the U.S. via rapidly expanding investment links and supply chains. The entire world was more tightly connected to America’s innovation infrastructure, and even adversaries like the USSR spent their time copying U.S. chips and chipmaking tools. Meanwhile, the chip industry had catalyzed an array of new weapons systems that were remaking how the U.S. military would fight future wars. American power was being recast. Now the entire nation depended on Silicon Valley’s success.

The aim of turning Japan into a country of democratic capitalists had worked. Now some Americans were asking whether it had worked too well. The strategy of empowering Japanese businesses seemed to be undermining America’s economic and technological edge.

A next-generation chip emerged roughly once every two years, requiring new facilities and new machinery. In the 1980s, U.S. interest rates reached 21.5 percent as the Federal Reserve sought to fight inflation. By contrast, Japanese DRAM firms got access to far cheaper capital. Chipmakers like Hitachi and Mitsubishi were part of vast conglomerates with close links to banks that provided large, long-term loans. Even when Japanese companies were unprofitable, their banks kept them afloat by extending credit long after American lenders would have driven them to bankruptcy.

The U.S. military was more dependent on electronics—and thus on chips—than ever before. By the 1980s, the report found, around 17 percent of military spending went toward electronics, compared to 6 percent at the end of World War II.

“The United States has been busy creating lawyers,” Morita lectured, while Japan has “been busier creating engineers.” Moreover, American executives were too focused on “this year’s profit,” in contrast to Japanese management, which was “long range.”

Grove realized Intel’s business model of selling DRAM chips was finished. DRAM prices might recover from the price slump, but Intel would never win back market share. It had been “disrupted” by Japanese producers. Now it would either disrupt itself or fail. Exiting the DRAM market felt impossible. Intel had pioneered memory chips, and admitting defeat would be humiliating. It was like Ford deciding to get out of cars, one employee said.

Grove’s restructuring of Intel was a textbook case of Silicon Valley capitalism. He recognized that the company’s business model was broken and decided to “disrupt” Intel himself by abandoning the DRAM chips it had been founded to build. The firm established a stranglehold on the market for PC chips, issuing a new generation of chip every year or two, offering smaller transistors and more processing power. Only the paranoid survive, Andy Grove believed. More than innovation or expertise, it was his paranoia that saved Intel.

Government efforts were effective not when they tried to resuscitate failing firms, but when they capitalized on pre-existing American strengths, providing funding to let researchers turn smart ideas into prototype products.

From day one, TSMC wasn’t really a private business: it was a project of the Taiwanese state. A crucial ingredient in TSMC’s early success was deep ties with the U.S. chip industry. Most of its customers were U.S. chip designers, and many top employees had worked in Silicon Valley.

The economics of chip manufacturing required relentless consolidation. Whichever company produced the most chips had a built-in advantage, improving its yield and spreading capital investment costs over more customers. TSMC’s business boomed during the 1990s and its manufacturing processes improved relentlessly. Morris Chang wanted to become the Gutenberg of the digital era. He ended up vastly more powerful. Hardly anyone realized it at the time, but Chang, TSMC, and Taiwan were on a path toward dominating the production of the world’s most advanced chips.

Intel’s computer processor oligopoly was too profitable to justify thinking about niche markets. Intel didn’t realize until too late that it ought to compete in another seemingly niche market for a portable computing device: the mobile phone.

Shortly after the deal to put Intel’s chips in Mac computers, Jobs came back to Otellini with a new pitch. Would Intel build a chip for Apple’s newest product, a computerized phone? All cell phones used chips to run their operating systems and manage communication with cell phone networks, but Apple wanted its phone to function like a computer. It would need a powerful computer-style processor as a result. “They wanted to pay a certain price,” Otellini told journalist Alexis Madrigal after the fact, “and not a nickel more…. I couldn’t see it. It wasn’t one of these things you can make up on volume. And in hindsight, the forecasted cost was wrong and the volume was 100× what anyone thought.” Intel turned down the iPhone contract.

So even though an American company is one of the world’s three biggest DRAM producers, most DRAM manufacturing is in East Asia.

America’s second-rate status in memory chip output, however, is nothing new. It dates to the late 1980s, when Japan first overtook the U.S. in DRAM output. The big shift in recent years is the collapse in the share of logic chips produced in the United States. Today, building an advanced logic fab costs $20 billion, an enormous capital investment that few firms can afford. As with memory chips, there’s a correlation between the number of chips a firm produces and its yield—the number of chips that actually work. Given the benefits of scale, the number of firms fabricating advanced logic chips has shrunk relentlessly.

Nvidia’s GPUs can render images quickly because, unlike Intel’s microprocessors or other general-purpose CPUs, they’re structured to conduct lots of simple calculations—like shading pixels—simultaneously. In 2006, realizing that high-speed parallel computations could be used for purposes besides computer graphics, Nvidia released CUDA, software that lets GPUs be programmed in a standard programming language, without any reference to graphics at all.

Soon the entire industry realized Qualcomm’s system would make it possible to fit far more cell phone calls into existing spectrum space by relying on Moore’s Law to run the algorithms that make sense of all the radio waves bouncing around. For each generation of cell phone technology after 2G, Qualcomm contributed key ideas about how to transmit more data via the radio spectrum and sold specialized chips with the computing power capable of deciphering this cacophony of signals.

It’s easy to lament the offshoring of semiconductor manufacturing. But companies like Qualcomm might not have survived if they’d had to invest billions of dollars each year building fabs.

Five years after Sanders retired from AMD, the company announced it was dividing its chip design and fabrication businesses. Wall Street cheered, reckoning the new AMD would be more profitable without the capital-intensive fabs.

Chang realized that TSMC could pull ahead of rivals technologically because it was a neutral player around which other companies would design their products. He called this TSMC’s “Grand Alliance,” a partnership of dozens of companies that design chips, sell intellectual property, produce materials, or manufacture machinery. Many of these companies compete with each other, but since none fabricate wafers, none compete with TSMC. TSMC could therefore coordinate between them, setting standards that most other companies in the chip industry would agree to use.

Like Qualcomm and the other chip firms that powered the mobile revolution, even though Apple designs ever more silicon, it doesn’t build any of these chips.

Application processors, the electronic brain inside each smartphone, are mostly produced in Taiwan and South Korea before being sent to China for final assembly inside a phone’s plastic case and glass screen. Apple’s iPhone processors are fabricated exclusively in Taiwan. Today, no company besides TSMC has the skill or the production capacity to build the chips Apple needs. So the text etched onto the back of each iPhone—“Designed by Apple in California. Assembled in China”—is highly misleading. The iPhone’s most irreplaceable components are indeed designed in California and assembled in China. But they can only be made in Taiwan.

The company’s engineers realized the best approach was to shoot a tiny ball of tin measuring thirty-millionths of a meter wide moving through a vacuum at a speed of around two hundred miles per hour. The tin is then struck twice with a laser, the first pulse to warm it up, the second to blast it into a plasma with a temperature around half a million degrees, many times hotter than the surface of the sun. This process of blasting tin is then repeated fifty thousand times per second to produce EUV light in the quantities necessary to fabricate chips. Jay Lathrop’s lithography process had relied on a simple bulb for a light source. The increase in complexity since then was mind-boggling.

The result was a machine with hundreds of thousands of components that took tens of billions of dollars and several decades to develop. The miracle isn’t simply that EUV lithography works, but that it does so reliably enough to produce chips cost-effectively. Extreme reliability was crucial for any component that would be put in the EUV system. ASML

The final product—chips—work so reliably because they only have a single component: a block of silicon topped with other metals. There are no moving parts in a chip, unless you count the electrons zipping around inside. Producing advanced semiconductors, however, has relied on some of the most complex machinery ever made. ASML’s EUV lithography tool is the most expensive mass-produced machine tool in history,

That a Dutch company, ASML, had commercialized a technology pioneered in America’s National Labs and largely funded by Intel would undoubtedly have rankled America’s economic nationalists, had any been aware of the history of lithography or of EUV technology. Yet ASML’s EUV tools weren’t really Dutch, though they were largely assembled in the Netherlands. Crucial components came from Cymer in California and Zeiss and Trumpf in Germany. And even these German firms relied on critical pieces of U.S.-produced equipment. The point is that, rather than a single country being able to claim pride of ownership regarding these miraculous tools, they are the product of many countries. A tool with hundreds of thousands of parts has many fathers. “Will it work?” Andy Grove had asked John Carruthers, before investing his first $200 million in EUV. After three decades of investment, billions of dollars, a series of technological innovations, and the establishment of one of the world’s most complex supply chains, by the mid-2010s, ASML’s EUV tools were finally ready to be deployed in the world’s most advanced chip fabs.

TSMC, Intel, and Samsung had financial positions that were strong enough to roll the dice and hope they could make EUV work. GlobalFoundries decided that as a medium-sized foundry, it could never make a 7nm process financially viable. It announced it would stop building ever-smaller transistors, slashed R&D spending by a third, and quickly turned a profit after several years of losses. Building cutting-edge processors was too expensive for everyone except the world’s biggest chipmakers. Even the deep pockets of the Persian Gulf royals who owned GlobalFoundries weren’t deep enough. The number of companies capable of fabricating leading-edge logic chips fell from four to three.

The PC market was stagnant, because it seemed nearly everyone already had a PC, but it remained remarkably profitable for Intel, providing billions of dollars a year that could be reinvested into R&D. The company spent over $10 billion a year on R&D throughout the 2010s, four times as much as TSMC and three times more than the entire budget of DARPA. Only a couple of companies in the world spent more. As the chip industry entered the EUV era, Intel looked poised to dominate. The company had been crucial to EUV’s emergence, thanks to Andy Grove’s initial $200 million bet on the technology in the early 1990s. Now, after billions of dollars of investment—a substantial portion of which had come from Intel—ASML had finally made the technology a reality. Yet rather than capitalizing on this new era of shrinking transistors, Intel squandered its lead, missing major shifts in semiconductor architecture needed for artificial intelligence, then bungling its manufacturing processes and failing to keep up with Moore’s Law.

Intel remains enormously profitable today. It’s still America’s biggest and most advanced chipmaker. However, its future is more in doubt than at any point since Grove’s decision in the 1980s to abandon memory and bet everything on microprocessors. It still has a shot at regaining its leadership position over the next half decade, but it could just as easily end up defunct. What’s at stake isn’t simply one company, but the future of America’s chip fabrication industry. Without Intel, there won’t be a single U.S. company—or a single facility outside of Taiwan or South Korea—capable of manufacturing cutting-edge processors.

As investors bet that data centers will require ever more GPUs, Nvidia has become America’s most valuable semiconductor company. Its ascent isn’t assured, however, because in addition to buying Nvidia chips the big cloud companies—Google, Amazon, Microsoft, Facebook, Tencent, Alibaba, and others—have also begun designing their own chips, specialized to their processing needs, with a focus on artificial intelligence and machine learning.

Most people in the industry think many of the company’s problems stem from Intel’s delayed adoption of EUV tools. By 2020, half of all EUV lithography tools, funded and nurtured by Intel, were installed at TSMC. By contrast, Intel had only barely begun to use EUV in its manufacturing process.

Far more than any other country, China has made the internet subservient to its leaders’ wishes. Foreign internet and software companies either signed on to whatever censorship rules the Communist Party desired or lost access to a vast market.

When it comes to the core technologies that undergird computing, China is staggeringly reliant on foreign products, many of which are designed in Silicon Valley and almost all of which are produced by firms based in the U.S. or one of its allies.

During most years of the 2000s and 2010s, China spent more money importing semiconductors than oil. High-powered chips were as important as hydrocarbons in fueling China’s economic growth. Unlike oil, though, the supply of chips is monopolized by China’s geopolitical rivals.

The dollar values at stake in China’s vision of reworking semiconductor supply chains were staggering. China’s import of chips—$260 billion in 2017, the year of Xi’s Davos debut—was far larger than Saudi Arabia’s export of oil or Germany’s export of cars. China spends more money buying chips each year than the entire global trade in aircraft. No product is more central to international trade than semiconductors. It wasn’t only Silicon Valley’s profits that were at risk. If China’s drive for self-sufficiency in semiconductors succeeded, its neighbors, most of whom had export-dependent economies, would suffer even more. Integrated circuits made up 15 percent of South Korea’s exports in 2017; 17 percent of Singapore’s; 19 percent of Malaysia’s; 21 percent of the Philippines’; and 36 percent of Taiwan’s. Made in China 2025 called all this into question.

Viewed on their own terms, the deals that IBM, AMD, and Arm struck in China were driven by reasonable business logic. Collectively, they risk technology leakage. U.S. and UK chip architectures and designs as well as Taiwanese foundries have played a central role in the development of China’s supercomputer programs. Compared to a decade ago, though its capabilities still meaningfully lag the cutting edge, China is substantially less reliant on foreigners to design and produce chips needed in data centers.

The problem wasn’t simply that Chinese government-linked funds were buying up foreign chip firms. They were doing so in ways that violated laws about market manipulation and insider trading.

For all the country’s export prowess, China’s internet firms make almost all their money inside of China’s domestic market, where they’re protected by regulation and censorship. Tencent, Alibaba, Pinduoduo, and Meituan would be minnows were it not for their home market dominance. When Chinese tech firms have gone abroad, they’ve often struggled to compete.

The same cell towers that transmit calls also send other types of data. So Huawei’s equipment now plays an important—and in many countries, crucial—role in transmitting the world’s data. Today it is one of the world’s three biggest providers of equipment on cell towers, alongside Finland’s Nokia and Sweden’s Ericsson.

Cell networks will identify a phone’s location and send radio waves directly toward a phone, using a technique called beamforming. A typical radio wave, like one that sends music to your car radio, sends signals out in every direction because it doesn’t know where your car is. This wastes power and creates more waves and more interference. With beamforming, a cell tower identifies a device’s location and sends the signal it needs only in that direction. Result: less interference and stronger signals for everyone.

With Huawei’s design arm proving itself world-class, it wasn’t hard to imagine a future in which Chinese chip design firms were as important customers of TSMC as Silicon Valley giants. If the trends of the late 2010s were projected forward, by 2030 China’s chip industry might rival Silicon Valley for influence. This wouldn’t simply disrupt tech firms and trade flows. It would also reset the balance of military power.

From swarms of autonomous drones to invisible battles in cyberspace and across the electromagnetic spectrum, the future of war will be defined by computing power. The U.S. military is no longer the unchallenged leader. Long gone are the days when the U.S. had unrivaled access to the world’s seas and airspace, guaranteed by precision missiles and all-seeing sensors.

Of course, computing power has been central to warfare for the past half century, though the quantity of 1s and 0s that can be harnessed to support military systems is millions of times larger than decades earlier. What’s new today is that America now has a credible challenger. The Soviet Union could match the U.S. missile-for-missile but not byte-for-byte. China thinks it can do both. The fate of China’s semiconductor industry isn’t simply a question of commerce. Whichever country can produce more 1s and 0s will have a serious military advantage, too.

For many Chinese military systems, however, acquiring U.S.-designed, Taiwan-fabricated chips hasn’t been difficult. A recent review of 343 publicly available AI-related People’s Liberation Army procurement contracts, by researchers at Georgetown University, found that less than 20 percent of the contracts involved companies that are subject to U.S. export controls. In other words, the Chinese military has had little difficulty simply buying cutting-edge U.S. chips off-the-shelf and plugging them into military systems.

Just as the Cold War was decided by electrons zipping around the guidance computers of American missiles, the fights of the future may be decided in the electromagnetic spectrum. The more the world’s militaries rely on electronic sensors and communication, the more they’ll have to battle for access to the spectrum space needed to send messages or to detect and track adversaries.