Background

When supply chains froze up during the COVID-19 pandemic, it was a rude awakening for many living in North America and Europe how dependent their economies were on critical products being produced in Asia ranging from PPE to microchips. COVID also accelerated the political and economic tensions that had been brewing between China and United States since 2013 when Xi Jinping became President. Since 1945, the US-led economic order had promoted a (relatively) free trade regime which had ushered in unprecedented economic development in formerly poor East Asian countries like Korea, Singapore, Hong Kong, and China. These countries developed world-class export-oriented companies through a mix of market and state-led industrial strategies.

In 2020 as the COVID virus ripped across the globe, Western nations found themselves unable to manufacture critical medicines and equipment they needed to keep their healthcare workers safe. Furthermore, the sectors which had been partially exempted from the lockdown policies, like manufacturing, were made idle due a lack of computer chips which had become ubiquitous in mainly production lines. It was now clear to all that a globalized economy and just-in-time supply chains had produced profound fragilities. The entire premise of the benefits that a globalized economy would bring, a hallmark of the post-war world economic thought, was being called into question.

It was not of course the first time globalization had been challenged, especially by political constituents who faced job losses in sectors like textiles, steel, appliances, automotives, and many other heavy industries. Occasionally, protectionist policies would be enacted like the “voluntary” export quotas on Japanese automobiles in 1981, or Bush Jr’s steel tariff of 2002. And some sectors had been coddled and insulated for decades like the Canadian dairy cartel or European agriculture. But overall, these were exceptions to the rule of increased market access and global integration.

However, as China increasingly came to be seen as a national security threat,[¹] the danger of economic dislocation was surpassed by a concern of military and political hegemony. Sure, Chinese economic growth might raise GDP per capita on the margin, but since China’s population was twice the size of the US and the EU combined, its ability to project military and diplomatic power would eventually overtake that of the US-led world. President Trump’s election in 2016 signaled an end to many things, but most relevant to this discussion was an end to treating China as just another economy in Asia. A series of trade wars was launched in 2018, with mixed results, but a clear emphasis that business would no longer be conducted as usual. Now, the US would be explicitly pursuing a policy of decoupling.

At the start of the US-China trade wars, Trump sought to cause as much damage to Huawei and ZTE as possible, two prominent Chinese telecoms companies that were fighting for global market share of the 5G market. Trump banned US companies from purchasing or selling technology to these firms. President Biden continued Trump’s general stance of economic hostility towards China, albeit with less bombastic rhetoric. In 2022, the Biden Administration upped the ante and put in export controls on the sale of advanced computing and semiconductor manufacturing items to PRC.

The emphasis on semiconductors (aka “chips”) as key elements of US national security highlighted the profound importance of computing in both the military and economic spheres. Presidents Trump, Biden, and the American political class have embraced the idea of the need for an American industrial policy. While state-led development of specific sectors is hardly new in US history (most famously Alexander Hamilton’s Report of Manufactures), it is arguably the most important shift in US economic policy since the 1980s. American companies have fully internalized this change constantly speak of reshoring, nearshoring, and friendshoring.

Arguably, the most important industry coming back to North America is semiconductor device fabrication. The signature legislation in this area so far has been the CHIPS Act which was signed by Joe Biden into law in 2022. The US pioneered the semiconductor industry; inventing the integrated circuit and microprocessor. Silicon Valley takes its name semiconductor companies like Fairchild Semiconductor, General Microelectronics, and Intel. From around 1980s to today, the US semiconductor industry has slowly lost global market share and has retreated to R&D intensive subsectors like chip design, whereas the East Asian economies have focused on the capital intensive processes (i.e. chip fabrication or “fabs”).

Needless to say, semiconductors are the workhorse of the 21st century economy and power everything from airplanes to computers (obviously) to cars to missiles to medical devices. The story of the microprocessor’s development and its current place in the geopolitical struggle between US and Chinese hegemony is therefore required reading for anyone interested in understanding the modern world.

Review

Enter Chip War by Chris Miller, a highly readable but sophisticated overview of both the history and science underpinning microchips and their place in the modern world. Miller is a highly credentialed professor at Tufts, with appointments at the usual DC Think tanks. He is also a director at a macroeconomic and geopolitical consultancy company. The book is told from a US perspective, which is fair given his background and audience (it was one of Obama’s favorite books of 2023). Miller’s academic background is in Cold War history, and he does a good job at showing that the first chip war was actually between the US and the USSR. The Soviets were incapable of developing their semiconductor industry able to keep pace with the US. Instead, they followed a policy of “steal and copy”, which them the Soviets were always a decade behind the US in chip technology. This deficit was reflected in the sorts of weapons the Soviets used and their emphasis on lower-compute platforms like mass tank formations and (more concerning) nuclear weapons.

The first Gulf War highlighted the staggering technological lead the US army had compared with Saddam’s Soviet-equipped army. Systems like precision guided munitions, stealth bombers, satellite surveillance, and missile defense systems were able to pulverize Iraqi troops and neutralize counter attacks. The US and coalition forces inflicted casualties at a 20:1 ratio against Saddam’s army, a ratio akin the level of military imbalance European colonial armies had near the end of the 19th century against their indigenous opponents. Soviet military planners realized that the future of warfare would be dominated by technology and information warfare.

Since semiconductors have continued to follow Moore’s law, the computing power of chips has gone up by a factor of 1-million since from 1990 to 2020. This means the number of transistors on a dense integrated circuit in 2020 is 1-million times larger than it was in 1990. Since computing power, usually measured as FLOPs, will be a multiple even larger than that (due to improvements processor architecture and parallel computing techniques), the abilities of computers and their integrated components in military systems is now profound. Such power is what enables the algorithms inside your phone to be able to use 5G cellular technology or AI algorithms to power object detection and automation like driverless cars. An army, let alone an economy, without access to these chips will be unable to maintain a leading position.

The book does a good job at describing the mind-boggling complexity of what it takes to keep Moore’s law going (see quotes below). Miller demonstrates why semiconductors have become incredibly capital intensive, and have necessarily been perfected by only a handful in firms.[²] For chip fabrication, the main players are Taiwan Semiconductor Manufacturing Company (TSMC), Samsung, and Intel. Currently, the former two are able to produce 3-nm (nanometer) chips, while Intel is at 5nm, but all are racing to get down to 2nm. A fab factory might cost up to $15 billion USD, and will need to be retooled for every new generation of chip. These high upfront costs are coupled with the high costs of actually running the factories including massive amounts of electricity (cleaning and cooling), skilled labor, and materials like silicon wafers.

The most important chip manufacturer is arguably TSMC, since it operates purely as foundry for its clients. This means that it doesn’t design chips (by default anyways) but rather produces chips meeting client specifications. This form of outsourcing means that a company like NVIDIA can focus solely on product and chip design, and leave the actual “printing” to TSMC. While Intel and Samsung offer foundry services, they also act as a vertically integrated business and create the chips needed for their own products like CPUs and GPUs.

As Chip War shows, having the world’s most important foundry headquartered and producing chips mainly in Taiwan, gives the Western world’s policy makers atrial fibrillation every time the PRC makes not-so-subtle references to a planned takeover of the island in the next five years. TSMC is considered Taiwan’s golden goose, accounting for 25% of the country’s GDP, and is a point of national pride. As Miller points out, it also provides Taiwan with an asset of military importance that acts as a deterrent, since help from the US, Japan, and Korea is seen as more likely to be forthcoming given the company’s critical position in the global economy. Famously, the Republican Presidential candidate Vivek Ramaswamy said that once the US was able to produce advanced chips, he would let China invade Taiwan. While this view might not be universally held, it’s certainly indicative of what many world leaders might be thinking behind closed doors.

But TSMC is not the only player that matters of course. State-of-the-art (SOTA) chip manufacturing requires cutting edge lithography (creating transistors by etching silicon with light), which is currently produced by one Dutch company: ASML (one of their machines costs $380 million). These EUV machines developed by ASML are also dependent on components like optical systems developed by Zeiss, a German company. To determine the most efficient layout of transistors, British-based companies like Arm Holdings sell access to their own proprietary software (e.g. the ARM64 processor). This string of complicated dependencies between firms found throughout the US and Europe gives the Western world a certain amount of leverage over China, but also gives the PRC the ability to play sides off against each other.

Since Chip War was published at the end of 2022, the battle for semiconductor supremacy has continued with Chinese companies increasingly decoupling from the US, Japanese, and European industrial ecosystem. The US has courted TSMC to expand fab plants in Phoenix, although production delays and cultural frictions are showing the challenges that the US will have in becoming and industrial superpower again.

Overall I’d give Miller’s book a solid 8/10, and recommend anyone interested in science, technology, or geopolitics to read the book as soon as possible to have an improved understanding the world’s current chip wars.

Quotes

Lathrop had used simple visible light and off-the-shelf photoresists produced by Kodak. Using more complex lenses and chemicals, it eventually became possible to print shapes as small as a couple hundred nanometers on silicon wafers. The wavelength of visible light is itself several hundred nanometers, depending on the color, so it eventually faced limits as transistors were made ever smaller. The industry later moved to different types of ultraviolet light with wavelengths of 248 and 193 nanometers. These wavelengths could carve shapes more precise than visible light, but they, too, had limits, so the industry placed its hope on extreme ultraviolet light with a wavelength of 13.5 nanometers.

Using EUV light introduced new difficulties that proved almost impossible to resolve. Where Lathrop used a microscope, visible light, and photoresists produced by Kodak, all the key EUV components had to be specially created. You can’t simply buy an EUV lightbulb. Producing enough EUV light requires pulverizing a small ball of tin with a laser. Cymer, a company founded by two laser experts from the University of California, San Diego, had been a major player in lithographic light sources since the 1980s. 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.

Trumpf had a reputation and a track record for delivering the precision and reliability Cymer needed. Could it deliver the power? Lasers for EUV needed to be substantially more powerful than the lasers Trumpf already produced. Moreover, the precision Cymer demanded was more exacting than anything Trumpf had previously dealt with. The company proposed a laser with four components: two “seed” lasers that are low power but accurately time each pulse so that the laser can hit 50 million tin drops a second; four resonators that increase the beam’s power; an ultra- accurate “beam transport system” that directs the beam over thirty meters toward the tin droplet chamber; and a final focusing device to ensure the laser scores a direct hit, millions of times a second.

Zeiss’ primary challenge was that EUV is difficult to reflect. The 13.5nm wavelength of EUV is closer to X-rays than to visible light, and as is the case with X-rays, many materials absorb EUV rather than reflect it. Zeiss began developing mirrors made of one hundred alternating layers of molybdenum and silicon, each layer a couple nanometers thick. Researchers in Lawrence Livermore National Lab had identified this as an optimal EUV mirror in a paper published in 1998, but building such a mirror with nanoscale precision proved almost impossible. 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.

Footnotes