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

by Chris Miller

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.

Huang gave away CUDA for free, but the software only works with Nvidia’s chips. By making the chips useful beyond the graphics industry, Nvidia discovered a vast new market for parallel processing, from computational chemistry to weather forecasting. At the time, Huang could only dimly perceive the potential growth in what would become the biggest use case for parallel processing: artificial intelligence.

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. The company’s patents are so fundamental it’s impossible to make a cell phone without them.

Qualcomm has made hundreds of billions of dollars selling chips and licensing intellectual property. But it hasn’t fabricated any chips: they’re all designed in-house but fabricated by companies like Samsung or TSMC.

The investment arm of the Abu Dhabi government, Mubadala, became the primary investor in the new foundry, an unexpected position for a country known more for hydrocarbons than for high-tech. CFIUS, the U.S. government body that reviews foreign purchases of strategic assets, waved the sale through, judging that it had no national security implications. But the fate of AMD’s production capabilities would end up shaping the chip industry—and guaranteeing that the most advanced chipmaking would take place offshore. GlobalFoundries, as this new company that inherited AMD’s fabs was known, entered an industry that was as competitive and unforgiving as ever.

Jobs gave a lecture that asked, “What is software?” “The only thing I can think of,” he answered, “is software is something that is changing too rapidly, or you don’t exactly know what you want yet, or you didn’t have time to get it into hardware.”

Ultimately, Zeiss created mirrors that were the smoothest objects ever made, with impurities that were almost imperceptibly small. If the mirrors in an EUV system were scaled to the size of Germany, the company said, their biggest irregularities would be a tenth of a millimeter.

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.

ASML’s EUV lithography tool is the most expensive mass-produced machine tool in history, so complex it’s impossible to use without extensive training from ASML personnel, who remain on-site for the tool’s entire life span.

Because manufacturing tools account for much of the cost of an advanced fab, keeping the equipment operating is crucial for profitability. In the U.S., Chiang said, if something broke at 1 a.m., the engineer would fix it the next morning. At TSMC, they’d fix it by 2 a.m. “They do not complain,” he explained, and “their spouse does not complain” either.

Softbank had purchased Arm in 2016 for $40 billion, but it sold a 51 percent stake in the China division—which according to Softbank accounted for a fifth of Arm’s global sales—for only $775 million.

First, assiduously cultivate political relationships to garner favorable regulation and cheap capital. Second, identify products pioneered in the West and Japan and learn to build them at equivalent quality and lower cost. Third, globalize relentlessly, not only to seek new customers but also to learn by competing with the world’s best companies.

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.

“We will never try to match our opponents or our competitors tank for tank, plane for plane, person for person,” declared Bob Work, the former deputy defense secretary who is the intellectual godfather of this new offset, in a clear echo of the logic of the late 1970s. The U.S. military will only succeed, in other words, if it has a decisive technological advantage.

The U.S. military is already fielding the first generation of new autonomous vehicles, like Saildrone, an unmanned windsurfer that can spend months roving the oceans while tracking submarines or intercepting adversaries’ communications.

The challenge isn’t simply to embed computing power in a single device, like a guided missile, but to network thousands of devices across a battlefield, letting them share data and putting machines in a position to make more decisions.

The military worries that chips fabricated or assembled abroad are more susceptible to tampering, with back doors added or errors written in. However, even chips designed and produced domestically can have unintended vulnerabilities.

Taiwan isn’t simply the source of the advanced chips that both countries’ militaries are betting on. It’s also the most likely future battleground.

Newspapers and academics alike reported that globalization was in fact “global,” that technological diffusion was unstoppable, that other countries’ advancing technological capabilities were in the U.S. interest, and that even if they weren’t, nothing could halt technological progress.

“Policy can, in principle, slow the diffusion of technology, but it cannot stop the spread.” Neither of these claims was backed by evidence; they were simply assumed to be true. However, “globalization” of chip fabrication hadn’t occurred; “Taiwanization” had. Technology hadn’t diffused. It was monopolized by a handful of irreplaceable companies.

America’s technological lead in fabrication, lithography, and other fields had dissipated because Washington convinced itself that companies should compete but that governments should simply provide a level playing field.

So Washington kept telling itself that the U.S. was running faster, blindly ignoring the deterioration in the U.S. position, the rise in China’s capabilities, and the staggering reliance on Taiwan and South Korea, which grew more conspicuous every year.

Many officials worried that China’s leverage over the world’s critical technology systems was growing.

Inaction wasn’t a viable option, they believed. Nor was “running faster”—which they saw as code for inaction.

One U.S. semiconductor executive wryly summed things up to a White House official: “Our fundamental problem is that our number one customer is our number one competitor.”

When Micron sued UMC and Jinhua for violating its patents, they countersued in China’s Fujian Province. A Fujian court ruled that Micron was responsible for violating UMC and Jinhua’s patents—patents that had been filed using material stolen from Micron. To “remedy” the situation, Fuzhou Intermediate People’s Court banned Micron from selling twenty-six products in China, the company’s biggest market.

After detailed consultations with officials at Japan’s powerful Ministry of Economics, Trade, and Industry, the Trump administration was confident Tokyo supported a tough move against Jinhua and would ensure Japanese companies didn’t undercut American restrictions on the firm. This gave the U.S. a powerful new tool to put out of business any chipmaker, anywhere in the world.

So after Jinhua paid invoices to the U.S. firms that supplied its crucial chipmaking tools, the U.S. banned their export. Within months, production at Jinhua ground to a halt. China’s most advanced DRAM firm was destroyed.

Nearly every chip in the world uses software from at least one of three U.S.-based companies, Cadence, Synopsys, and Mentor (the latter of which is owned by Germany’s Siemens but based in Oregon).

Excluding the chips Intel builds in-house, all the most advanced logic chips are fabricated by just two companies, Samsung and TSMC, both located in countries that rely on the U.S. military for their security. Moreover, making advanced processors requires EUV lithography machines produced by just one company, the Netherlands’ ASML, which in turn relies on its San Diego subsidiary, Cymer (which it purchased in 2013), to supply the irreplaceable light sources in its EUV lithography tools.

These academics worried that the U.S. government’s use of trade and capital flows as political weapons threatened globalization and risked dangerous unintended consequences. The Trump administration, by contrast, concluded it had unique power to weaponize semiconductor supply chains.

Mark Liu, promised not only to abide by the letter of the law but also its spirit. “This is something that can be solved not solely through the interpretation of the rules, but also has to do with the intentions of the U.S. government,”

Huawei’s been forced to divest part of its smartphone business and its server business, since it can’t get the necessary chips.

China’s rollout of its own 5G telecoms network, which was once a high-profile government priority, has been...

Car companies at first cut chip orders, expecting car sales to slump. When demand quickly recovered, they found that chipmakers had already reallocated capacity to other customers.

The Biden administration and most of the media interpreted the chip shortage as a supply chain problem.

However, the semiconductor shortage wasn’t primarily caused by issues in the chip supply chain. There were some supply disruptions, like COVID lockdowns in Malaysia, which impacted semiconductor packaging operations there. But the world produced more chips in 2021 than ever before—over 1.1 trillion semiconductor devices, according to research firm IC Insights.

The semiconductor shortage is mostly a story of demand growth rather than supply issues.

Such problems that emerged, notably the shortage of auto chips, are mostly the fault of carmakers’ frantic and ill-advised cancelation of chip orders in the early days of the pandemic coupled with their just-in-time manufacturing practices that provide little margin of error.

China’s ruling party has no higher goal than asserting control over Taiwan. Its leaders constantly promise to do so.

Taiwan produces 11 percent of the world’s memory chips. More important, it fabricates 37 percent of the world’s logic chips. Computers, phones, data centers, and most other electronic devices simply can’t work without them, so if Taiwan’s fabs were knocked offline, we’d produce 37 percent less computing power during the following year.

It would take at least half a decade to rebuild the lost chipmaking capacity. These days, when we look five years out we hope to be building 5G networks and metaverses, but if Taiwan were taken offline we might find ourselves struggling to acquire dishwashers.

A facility to fabricate the most advanced logic chips costs twice as much as an aircraft carrier but will only be cutting-edge for a couple of years.

we’re seeing a “decline of computers as a general purpose technology.”

future of computing will be divided between “ ‘fast lane’ applications that get powerful customized chips and ‘slow lane’ applications that get stuck using general-purpose chips whose progress fades.”