- Leszek Bukowski
And China is losing it.
According to Intel's current CEO Pat Gelsinger, semiconductors will be a significant factor in shaping global international politics in the coming decades. He is joined by Eric Schmidt, former Google CEO and former chairman of the National Commission on Security and Artificial Intelligence, who believes that the development of artificial intelligence is already a dominant force shaping international relations.
The astonishing development of artificial intelligence is an unprecedented event in the history of the world. No one needs to be convinced that it will have a decisive impact on the lives of us all. It will inevitably have an equally fundamental impact on the geopolitical situation. According to Schmidt, the state that achieves supremacy in AI will also dominate its opponents technologically. Dominance in AI, however, will only be achieved with the hardware on which it is built. This is a particular problem for the People's Republic of China.
Silicon Empires
A Polish science fiction writer, Stanislaw Lem, once said: "Technology has been linked to science in a positive feedback loop for three hundred years.”
So science, through its discoveries and theories, drives technological progress, which in turn creates tools to aid scientific understanding of the world. In one such feedback loop, just after the Second World War, a group of scientists at Bell Labs began researching materials such as silicon and germanium. This research will later lead to the creation of transistors - electronic semiconductor devices that are a switch that regulates the flow of electricity. They are used to create physical logic gates - the elementary building blocks of modern computers. This is also where Chris Miller begins his book - Chip War, on which this episode is based. At first, few people understood the significance of the discoveries made by William Shockley, John Bardeen and Walter Brattain in the laboratories of the AT&T telecommunications company, which included Bell Labs.
In 1948, no one could have predicted that transistors on a silicon wafer would become a major factor in the global politics of the world's major powers. Nor could anyone have foreseen that miniature integrated circuits would make it possible to build personal computers that fit in our pockets. Before the first calculating machines were developed, it was people who did the calculations themselves for various purposes. Computing has always been, and still is, an inalienable element necessary for the existence and development of human civilisations. Cuneiform tablets from the Sumerian civilisation contain the calculations the Sumerians needed to make to run their state. But never before has the issue of computational autonomy been the subject of dispute between the superpowers - and it is today, as the current conflict between the United States and China is waged over the ability to use silicon in the advanced computing needed to build artificial intelligence models.
But let us go back to the origins of integrated circuits. Silicon Valley takes its name from the first companies that focused on creating transistors that could be assembled into a circuit on a single silicon wafer. Companies such as Fairchild Semiconductor, Texas Instruments and, later, Intel laid the foundations for a system that in later years saw the creation of other companies involved in disruptive, advanced technologies: Apple, Microsoft, Google, and OpenAI. Silicon Valley is still the place where the most powerful companies are headquartered. Nowhere else in the world is there such an accumulation of capital as in the northern part of the Santa Clara Valley.
On the other hand, chip manufacturing has been one of the driving forces behind globalisation. As early as the beginning of the 1960s, Fairchild Semiconductor began manufacturing its chips in Hong Kong, then administered by the United Kingdom. It was a process that contributed significantly to the rise of Asian countries such as Japan, South Korea and Taiwan - all of which became important centres for producing semiconductors, which require the most complex manufacturing process ever designed by humankind. Mainland China, on the other hand, was a late entrant to the semiconductor race. Only since the 2000s has it sought to become a significant, if not a major, player in the field of advanced integrated circuits.
But the current battle is not the first silicon war the US has fought. It was opposed as far back as the 1980s by Japan. The industrial might, ingenuity and diligence of the Japanese-made electronics, including chips made in Japan, were synonymous with modernity. Chris Miller, for example, describes how FBI agents caught the Japanese company Hitachi in industrial espionage.
The start of the 1990s marked the era of Intel. The proliferation of PC-like devices meant that Intel's processors monopolised the logic chip market. Intel's then-CEO Andy Grove struck a deal with Bill Gates - CEO and founder of the then-emerging Microsoft. To this day, Microsoft's software and Intel's x86 architecture dominate the personal computer industry.
Today, if you pick up any personal computer, you can be sure that its central processing unit (CPU) has been designed by one of the American companies: Intel, AMD (Advanced Micro Devices) or Apple. Intel still makes most of its chips in the US, while AMD and Apple operate a so-called 'fabless' system - they do the R&D and design of their chips but outsource production to Taiwanese giant TSMC, which in turn makes chips exclusively for third parties.
Nowadays, graphics processing units (GPUs), popularly known as 'graphics cards', have also become very important. Initially, these types of chips were only used for graphics-related calculations in computer games or graphics programs. However, around 2012, their specific architecture attracted researchers' attention in artificial intelligence and, in particular, neural networks. A team of researchers led by Geoffrey Hinton began using GPUs to train neural networks, which could become increasingly sophisticated and multi-layered by running their computations on GPUs. It was also the time when the concept of deep learning was born. Today, the GPU market is dominated by two US companies: Nvidia and AMD - with a clear advantage for Nvidia, which capitalised on the formative moment of the new field of deep learning to dominate the market for the chips needed to train artificial intelligence. Miller writes that it is estimated that up to 95% of the GPUs in Chinese servers doing AI computing are made by Nvidia.
And it is precisely for this reason that China has embarked on a battle for computing independence from the West, particularly from US chip-making technologies. As Xi Jinping himself has said, "Without cyber security, there is no national security, and without computerisation, there is no modernisation". In 2015, China announced the 'Made in China 2025' programme, the core of which is to make the Chinese semiconductor market independent of US technologies. In Chinese power circles, these technologies are known as critical.
The start of silicon war between China and the US began in 2016 when the US sanctioned the Chinese DRAM chip company Jinhua. This followed an investigation into a Chinese employee's theft of intellectual property from a US company Micron. As a result, production at Jinhua was halted. The next pivotal moment came in April 2018, when the US imposed devastating sanctions - later to be tightened in 2020 - on leading Chinese telecoms companies ZTE and Huawei. According to many experts, it was also a pivotal moment in shaping China's strategy towards the semiconductor challenge. It became clear to policymakers in the Middle Kingdom that the Chinese economy was highly vulnerable to sanctions in the high-tech sector. Since then, many companies have begun buying and stockpiling US chips in case sanctions are tightened. Moreover, it quickly became clear that the might of Chinese internet giants such as Baidu and Tencent was insufficient to ensure strategic and economic independence from the United States.
A worrying dependence
On the one hand, Beijing has created a two-decade-old fusion of artificial intelligence and authoritarianism, maximising the use of technology to monitor its citizens while remaining dependent on Silicon Valley companies for critical technologies. As Miller points out, "In powering China's economic growth, efficient chips have been as important as petroleum [...] China's problem is not just chip manufacturing. At almost every stage of the manufacturing process, China is surprisingly dependent on foreign technology. Nearly all of it is controlled by its geopolitical rivals Taiwan, Japan, South Korea and the United States. The Georgetown University study found that in the entire integrated circuit supply chain, including design, intellectual property, equipment, manufacturing and all the other links, Chinese companies account for just 6 per cent, compared with 39 per cent for American companies, 16 per cent for South Korean companies and 12 per cent for Taiwanese companies.
One reason for this is that state-of-the-art integrated circuits can only be produced on EUV machines manufactured by the Dutch company ASML. As a result, ASML has an absolute monopoly on the world market in developing and selling this type of equipment. Chips produced on EUV machines by Taiwan's TSMC end up in flagship phones from companies such as Apple and Samsung, as well as Chinese companies such as Xiaomi, Oppo and, until the sanctions, Huawei. The chips are mainly made for US companies Qualcomm and Apple. Qualcomm later sells its SnapDragon processors to mobile device manufacturers.
Nvidia's advanced chips for neural network training are also produced by Taiwan's TSMC using ASML machines. In addition to the Taiwanese giant, EUV machines are also used by Intel and Samsung. However, the Dutch company ASML also relies on technology from the US: in 2013, it bought San Diego-based company Cymer, which makes advanced light emitters for EUV machines. This bottleneck in the chain of dependency in chip production has also been exploited by Washington, which has banned the sale of EUV machines to China and, more recently, after negotiations with the Dutch, machines using older DUV technology. As early as 2018, one of the lawyers working for ASML said during a case brought by the company against a Chinese employee accused of industrial espionage that the employee was involved in a conspiracy planned by the Chinese government. This view was echoed in the Dutch intelligence service's 2023 annual report, which described China as "the greatest threat to the economic security of the Netherlands".
However, the sanctions imposed by the Biden administration in the autumn of 2022 on China's semiconductor industry and artificial intelligence companies are nothing like previous hits of this kind. We talked more about this in our October 2022 episode. On the one hand, buying chips for ordinary consumer applications in China is still possible. Still, for example, Nvidia's A100 and H100 chips, which are used to train models such as ChatGPT, are currently inaccessible to Chinese big tech companies. US citizens are also barred from working for Chinese chip companies. From the perspective of a few months, it can be said that the current strategy of the US government has two goals: first, to hamper the ability of Chinese companies to produce advanced semiconductors, and second, to limit China's capabilities in the field of artificial intelligence.
According to current Intel CEO Pat Gelsinger, semiconductors will be a significant factor in shaping global international politics in the coming decades. Meanwhile, Eric Schmidt, CEO of Google from 2011 to 2015 and former chairman of the National Commission on Security and Artificial Intelligence, says that the development of AI is already a dominant force shaping international relations. According to Schmmidt, AI is an autocatalytic technology - meaning its growth makes it develop even faster on its own and stimulates the development of other fields. A state that achieves supremacy in AI will dominate its opponents technologically, says Schmmidt.
Beijing's main strategic problem is that the West is dependent on China for consumer goods, none of which are critical to the survival and development of civilisation. The West can buy cheap white goods, telephones and other products from Chinese companies. Still, at the same time, it can reconstitute supply chains should the network of economic ties with China be radically disrupted.
The situation is different on the Chinese side. Critical goods such as advanced integrated circuits depend on the West and its Asian allies such as Japan, South Korea and Taiwan. According to a recent CSIS (Center for Strategic and International Studies) report, Xi's speeches repeatedly raise the issue of China's economic-strategic self-sufficiency, the path to which lies in mastering the technology for manufacturing advanced integrated circuits in mainland China. In the same report, we can read that the key elements of China's strategy to counter US sanctions are
- Avoidance of new sanctions and continued access to US technology;
- Trying to break the unity between the United States and its allies;
- Acquisition of foreign technology through industrial espionage and recruitment of foreign talent;
- Pressure on Chinese companies to buy Chinese products and eliminate US semiconductor equipment suppliers;
- Conducting a symmetric response to US and Allied sanctions
And indeed, we can expect Chinese sanctions on US memory manufacturer Micron.
A race to decide the fate of the world
One cannot also forget that any military build-up around Taiwan is a significant threat and unknown. This does not have to be a direct invasion, but could be a naval blockade of the island.In either case, the whole world would face a technological ice age lasting several years. There would be no more annual launches of new mobile phone models, and research into artificial intelligence would probably lose its momentum.
Apple, the world's most valuable company, manufactures all the processors used in its devices here at Taiwan's TSMC. While the production of Apple devices takes place at assembly plants owned by Foxconn, which, although a Taiwanese company, is also the largest employer in the People's Republic of China. The disruption of supply chains in the US-China-Taiwan network would therefore result in a drastic loss of value for a company that is also the biggest driver of the US economy. Nvidia also makes artificial intelligence chips in TSMC factories and would face a similar fate to Apple. Behind them, the value of all the other Silicon Valley giants would plummet. For the West, such a situation would mean years of profound economic crisis. On the other hand, every link in this chain can be reproduced in the West - but at a skyrocketing cost.
From China's point of view, the situation is different: in the event of an armed confrontation in the Taiwan Strait, the supply of even consumer chips for computers and phones would probably be blocked. Access to Microsoft, Google and Apple software would also be cut off. China could only use its own internal resources to produce chips several generations older than those used in the West. This would not be an economic crisis but a civilisational collapse. For these reasons, many observers believe that an invasion or blockade of Taiwan will never happen.
All this makes it inevitable that China will invest huge sums in its domestic semiconductor industry in the coming years. At the same time, the US will seek to diversify its chip supply chains and isolate China. The current bottlenecks in semiconductor manufacturing - such as TSMC - allow the US to block China's access to advanced technologies while at the same time putting the US itself and the global economy at risk if one of these bottlenecks is blocked. Washington's strategy is thus designed to densify the network of nodes essential to semiconductor production while at the same time cutting China off from this network.
One thing is sure - the battle for computing supremacy will shape the international situation for the next decade and probably beyond. And whoever gains the upper hand in computing power will be the favourite in the race for artificial intelligence, whose full potential is impossible to estimate.
China experienced incredible growth in recent years. Yet decades of building technological supremacy have put the West, and the United States in particular, at the forefront of perhaps the most crucial race in the modern world.
