With its massive ecological footprint, the semiconductor industry has been the material foundation for the IT assault. Now it is also driving economic war between the US and China
From capulcu
In most debates, the hype surrounding so-called “artificial intelligence” (AI) appears to us in a purely virtual form; as a promise of the virtually prerequisite-free automation of almost all areas of life, through a profound redesign of human-machine interaction based on human language. We have already discussed the political-economic consequences of this technological leap, in particular the increase in social inequality and dependence on a technocratic oligopoly (= the few guardians of the large language models) in The Gold Diggers of Artificial Intelligence.
In the text ChatGPT we examined a politically relevant narrowing of broad discourses due to an exaltation of the mainstream, coupled with an expected shift to the right as a hegemony amplifier. The interaction of social media with language generators such as ChatGPT has been proven to reduce discursive diversity and promote social fragmentation.
This text is deliberately intended to deal with the material side of artificial intelligence in the context of multiple interconnected crises — in particular the ecological crisis in connection with the crisis of new wars over a multipolar world order. Our first text Climate – the green vehicle of the AI offensive has already touched on the topic. AI is turning out to be an accelerant in terms of climate destruction and not, as is often fantasized, a central tool in solving an optimization problem that is too complex for humans.
The massive expansion of AI data centers not only consumes an enormous amount of energy when training and operating the large language models (a single training run of the current GPT-4 language model costs $64 million in electricity). The development and production of the chips consumes enormous amounts of energy and water — rare metals such as germanium and gallium are also required. Most computer hardware has therefore already developed a large part of its climate-damaging impact before it is switched on for the first time (this also applies to standard components such as laptops and smartphones). In addition to the energy-intensive operation of the data centers (added power consumption of the processors + their active cooling), the disposal of the high-performance hardware, which has sometimes only been used for two years, also contributes to the enormous ecological footprint.
The USA and the EU are currently spending a lot of money and other resources to rebuild a “domestic” semiconductor industry — with the aim of securing the crisis-ridden technological dominance over the declared “system rival” China. In particular, artificial intelligence is identified as a key technology: a technology that would be unthinkable without the most modern microchips – designed by Nvidia in Silicon Valley, USA and manufactured by TSMC in Taiwan with globally unique exposure machines from ASML from Eindhoven, Netherlands. Meanwhile, China has experienced an unprecedentedly rapid rise economically and technologically — so far, with no end in sight, even if Chinese economic growth has slowed somewhat in recent years. In some core areas such as electromobility, high-speed trains, renewable energies and 5G, Chinese companies are now leaving their Western competition behind. If none of the sides can assert their interests through other means, the trade war that is currently escalating could turn into a hot war. This raises the practical question of whether and how war and further militarization of society can be stopped.
Chips have been an important military technology since their inception. We also interpret the planned chip factories as part of the necessary economic unbundling to prepare for war. These chip factories that are now being built in Germany (e.g. Intel in Magdeburg or Infineon in Dresden) are therefore points at which resistance can and should begin in order to formulate a critique from the left of both progressive ecological destruction and a normalization of the logic of war in the current multiple crisis. The chip industry is at the same time highly specialized and globally integrated: entire supply chains therefore depend on the products and knowledge of individual companies and locations.
Ecological consequences of the chip boom
The “chip manufacturer” Nvidia is currently the biggest beneficiary of the AI boom. When it comes to high-end graphics chips, production no longer meets demand. Since ChatGPT was released at the end of 2022, the company’s value has increased sixfold to over two trillion dollars – even though Nvidia does not produce chips itself, but rather only designs and commissions them. Nvidia designs special high-performance chips designed for machine learning of so-called artificial neural networks, which are particularly effective at processing simple computing operations in parallel operation of many interconnected “processor cores”.
According to an empirical prediction (“Moore’s Law”) made in the 1960s, the further development of lithographic methods in chip production would cause the number of circuits on the same surface to double at least every 2 years. This roughly equates to doubling the performance of the chips. This results in a quasi-cyclical replacement of computer hardware with newer, more powerful hardware in many areas of application.
This constant replacement produces massive environmental damage. Particularly in the world of ever-increasing data processing and networking envisaged by technocrats, in which everything should communicate with everything else (smartification via 5G networks / Industry 4.0), this cyclical replacement production effort for computer chips increases exponentially.
In addition, the extent of ecological destruction required to produce a single chip increases with increasing power density. It takes three to four months for a silicon wafer to go through the various stages of processing into the finished product. The wafers are elaborately processed in a growing number of process steps in which microscopic layers are applied, patterns are burned in and unnecessary parts are scraped off in fully automatic processes. Rinsing with huge amounts of ultrapure water is an important part of this procedure.
If we assume the same size piece of silicon wafer for a computer chip, the most modern 2nm process technology requires significantly more electricity (x3.5) and ultra-pure water (x2.3) for production compared to the older 28nm technology. Greenhouse gas emissions (in CO2 equivalents) increase 2.5 times (per computer chip).
The Taiwanese chip manufacturer TSMC, the largest contract manufacturer in the world, which also supplies Apple, among others, is responsible for six percent of national electricity consumption. The ecological balance is catastrophic because almost half of Taiwan’s electricity comes from dirty coal power. To clean the wafers with ultrapure water, the company uses 150,000 cubic meters of water per day. And this is despite the fact that Taiwan has been suffering from a drinking water shortage for years. A lack of rain and dry periods have caused the levels of the water reservoirs to drop significantly. In some cities in Taiwan, drinking water has already had to be rationed and water pressure reduced so that the global supply chains of important semiconductors are not disrupted. The government is drilling wells across the country and is trying to calm angry rice farmers with compensation payments.
In an October 2020 paper, researchers at Harvard University used publicly available sustainability reports from companies like TSMC, Intel, and Apple to show how as computers become more common, so does their environmental harm. Information and computing technology is expected to account for up to 20% of global energy demand by 2030, with hardware responsible for a larger share of this environmental footprint than operating a system, researchers say: “Most of the Carbon emissions come from chip manufacturing, not hardware usage and energy consumption.”
The result is that the most advanced chipmakers already have a larger carbon footprint than some traditionally polluting industries, such as the automotive industry. The company’s information shows, for example, that in 2019, Intel’s factories used more than three times as much water as the facilities of the automobile manufacturer General Motors and generated more than twice as much waste classified as hazardous.
Information technology as a technology of power
The concept of technological attack helps us to develop criticism of technology as a criticism of power and society. As the writers of Disrupt! contend, to understand why we characterize innovation and technical ‘progress’ as an attack,
We have to realize that capitalist innovation theorists and strategists in particular conceive innovation as a comprehensive offensive, as a comprehensive shock. A shock that aims at the destruction and reorganization not only of work, but of society as a whole in all its areas, from work to transport, family, and finally education and culture. They don’t see innovations simply as “inventions”. They conceive of them as the use of basic technologies that have the potential for comprehensive destruction or “disruption”, and for reorganizational subjugation and adjustment.
The information technology attack we are referring to here is not the first innovation attack:
In the so-called “industrial revolution”, new machines (steam engines, automatic looms, etc.) served not only to destroy traditional forms of work and lifestyles based on them, but also to “shake up” the entire population. […] A subsequent wave of violence was launched around the electrical and chemical industries. It was closely linked to the forms of behavioral discipline and mental training through Taylorism and Fordism. Its material core lay in the attack of the technology of the electrically powered assembly line and its utopia on society as a whole. As its central “inventor” or “innovator”, the American Frederick Taylor himself explicitly called his system a “war” against the autonomy of workers (mostly migrant farm workers) and their unregulated ways of life.
In China as well as the USA and Europe, information technologies today represent a central pillar in the stabilization and enforcement of capitalist rule — both civil and military/police. Ubiquitous computing (for the ubiquitous recording and making available of all everyday data, e.g. via smartphone) and artificially intelligent modeling of this data (e.g. for behavior prediction), especially with machine learning techniques, have only become possible due to the enormous increases in the storage and computing capacities of microchips made possible in the last two decades.
Economic productivity has long depended on the quality and availability of IT applications and their hardware, especially where labor is expensive. For example, the automotive industry in Germany was temporarily unable to produce during the corona pandemic because the necessary chips or simple microelectronics were not available from the Far East.
The repertoire of government technologies that are ultimately based on information technologies and the computing power required for them, also known as “digitalization”, is quite diverse. Nudging, for example, seems to fit well with the EU’s (post-)democratic self-image. Ultimately, this technique allows you to create desired behavior with a high degree of probability without having to use coercion. Another example is the social credit systems that the Chinese government relies on and which put a lot of pressure on individuals to behave in a socially-compliant manner. These different and complementary techniques of rule mean a profound transformation of society. They are therefore an attack on people’s lives and work. The governments of both – supposedly completely different – political systems recognize the opportunities presented by the development of digital technologies and are promoting their implementation. A central effect of the technological attack is the creative destruction (Schumpeter) of existing social structures and forms of sociality. This has two important purposes. On the one hand, as a result of the attack, ever new areas of human life are valued for capital in the sense of original accumulation. On the other hand, the technological attack makes it possible to (approximately) predict and control social behavior. It therefore aims to create a predictive control society.
A critical resource in the war industry
Shortly after coming to power in 1933, the Nazis began preparing the German economy for war. The share of armaments costs in GDP rose from 1% to 20% between 1933 and 1938. The measures taken also included the construction of steelworks, which were highly uneconomical because they could not compete with cheap (e.g. Soviet) steel on the world market. The Nazi state massively subsidized industry and justified this with the need for economic “self-sufficiency”. Today, those who want to make their national economy “independent” of foreign goods and thus ready for war no longer primarily subsidize steel, aluminum or rubber factories, but above all chip factories and energy companies. The massive increase in defense spending by many states since the beginning of the war in Ukraine, coupled with the striving for economic “independence” in key industries such as semiconductor production and energy supply, as well as the simultaneous ideological rearmament (“ruining Russia”), serve a common goal: “being ready for war”. At least that is what the German War Minister called it. Once this preliminary goal has been achieved, it will only be a small step to actually waging war. Russia has recently shown how quickly this can happen.
Although an increasingly comprehensive digital society has been built since the 1970s through the spread of PCs, laptops and ultimately mobile devices in the civilian sector, military use continues to be a driving force behind semiconductor development. Military-relevant computer technologies have been promoted in huge research and development projects for decades.
Even conventional weapons such as rockets, bombs, etc. have been equipped with chips for a long time. Allegedly, during the Ukraine war, due to the shortage of microchips, Russia had to expand them to circumvent export bans on white goods (washing machines, etc.) and incorporate them into its own weapon system. With increasing networking and autonomy of weapon systems, the performance of the IT systems and thus also the availability of microchips to provide the necessary computing power will increasingly determine military strength. More and more AI applications are finding their way into military IT systems. Chatbots á la Chat-GPT are being implemented in battle management systems (e.g. Palantir’s “AIP for Defense”) as well as in simulation systems for the development of complex decision-making processes, for example in defense enemy drone attacks (e.g. “Ghostplay” by the German military’s Center for Digitalization and Technology Research) or for propaganda purposes and targeted disinformation campaigns using AI-generated fake images and texts. Not only since the company OpenAI deleted the civil clause for the use of ChatGPT in January 2024 have the large language models become dual use (civil+military).
US-China semiconductor race
In March 2023, the Dutch government announced new restrictions on the export of lithographic machines, i.e. the exposure machines that are central to the production of increasingly powerful chips, under pressure from the US government. Since then, ASML, the largest producer of chip production systems with a market share of almost 90%, can only be export its machines to China with special permission. The aim of this measure is to make it more difficult for China to set up its own chip production. China is already the largest chip producer in the world for outdated manufacturing technologies from 80nm and therefore has its own relevant industry in this sector. However, it is not able to produce the crucial high-performance chips that are required for modern servers, laptops, smartphones and graphics cards. The market share of the Taiwanese group TSMC is over 90%, particularly for semiconductors in manufacturing technologies below 14nm, although significant breakthroughs by Chinese companies have been repeatedly reported in recent months. Overall, TSMC also produces significantly more than half of all chips worldwide. The outsourcing of semiconductor manufacturing from the capitalist centers to Taiwan on the other side of the Pacific is a product of neoliberal globalization and the associated deindustrialization in many of the centers themselves.
Overall, it can be said that the semiconductor industry has a high degree of specialization and that (global) supply chains depend in many places on individual companies or plants. For example, it is not just the Dutch company ASML on which the supply of new production facilities depends — because ASML itself is dependent on other suppliers. The high-performance laser used to expose the wafers was developed by the mechanical engineering company Trumpf, based in Ditzingen. The mirror system with which these lasers are directed to the target comes from the Zeiss company in Oberkochen. Zeiss also boasts that 80% of all microchips manufactured worldwide are produced using its optical systems. But that’s not all – hundreds of chemicals are required in semiconductor production itself. Some of these can only be manufactured by a few companies. In Germany, BASF and Merck should be mentioned. And – you guessed it – Merck also claims that the chemicals it produces are contained in almost every microchip in the world. Because of their importance for semiconductor production, the federal government also discussed export restrictions for chemicals manufactured in Germany to China in April 2023.
China, in turn, will respond in August 2023 with a restriction on the export of gallium and germanium to the EU. These raw materials are essential for the production of microchips. China is the world’s largest producer of the minerals gallium and germanium. The EU gets 71 or 45 percent of it from China. The EU and its most important Western ally, the USA, are trying hard to build up their own raw material base.
Losses don’t stop at one’s “own” companies
The initiatives and investment programs of the USA, China and now also the EU to become more independent of Taiwan in semiconductor production date back to the mid-2010s. Behind this is the realization that microchips are not just any industrial product, but rather represent a “key technology”, as is particularly evident in the current debate about the export bans on Nvidia chips for training AI. By “becoming more independent”, “de-coupling”, and more recently in the EU’s terms, “de-risking”, the governments of these countries primarily aim for other states not to restrict their ability to generate capitalist profits for “their” chip factories, or their ability to wage war. There can be no question of one-sided sanctions or protection policies on the part of Western states. China is also taking tough measures to gain ground in the “chip war”. The Chinese government banned large companies in its own borders from buying (memory) chips from the US company Micron Technology. These chips can be manufactured in China – even if the market in this segment is currently (still) dominated by others.
All the essential software tools for chip design belong to Western companies. China has less than one percent market share in global chip design. The exposure systems for high-performance chips can only be manufactured by the Dutch company ASML. Chinese companies, on the other hand, can only compete on the global market with very outdated manufacturing technologies. The USA currently still has a certain lead over China in the chip sector, but it is rapidly shrinking. The US and EU are also just as “dependent” on manufacturing in Taiwan as China is. Given their economic and military importance, the vulnerable “supply” of chips from Taiwanese production represents an enormous risk from the governments’ point of view. This is especially since none of the states can be sure that they will be able to maintain this “supply” through military threats and, ultimately, war. This initial situation gives hope that the risk of war is assessed as too high on all sides – at least for now. The US government wants to do all it can to prevent China from catching up, both in terms of the necessary know-how in chip design, for example, and the existing manufacturing capacities. The Chinese government, for its part, has set itself the goal of becoming the world’s leading manufacturing power by 2049 with its Made in China 2025 program. In China, people assume that their own economy can benefit more from the “rules-based world order” than those who have been its main beneficiaries since its implementation after the Second World War.
The selective departure from the neoliberal paradigm of free trade in the USA and the EU is confirmation that they also share this assessment. The neoliberal dogma of the past decades is increasingly being questioned in these countries against the backdrop of successful Chinese state interventionism. In order to maintain its position of global power, the US government is willing to inflict economic damage even on large tech companies. Apple, for example, is not very enthusiastic about the US’s economic protection policy towards China. After all, the company’s devices are assembled there, even if they are designed in California and the required chips can only be manufactured in sufficient quality and quantity by TSMC in Taiwan. Unlike companies like Google, which have to fear Chinese companies as competitors, especially in the area of artificial intelligence, Apple has primarily benefited from the globalized division of labor with China and would like to continue this business.
From “developing country” to “system rival”
The governments in the EU states and the USA have recognized China as the first serious “system rival” since the collapse of the Soviet Union. They see that it is on the way to becoming possibly the largest economic power in the world and is already leaving leading Western companies behind in parts of the high-tech sector. Huawei is the company that registered the most patents worldwide in 2020, and the source of a large part of the network technology installed worldwide, for example for 5G. In other core areas such as artificial intelligence, most scientific publications are now of Chinese origin (although the quality of these publications is questionable).
When viewed naively, it may seem paradoxical that Western states are fighting China as a rival or competitor. After all, within the “rules-based world order”, China has developed from one of the poorest “developing countries” to the second strongest economy in the world – measured in terms of GDP – through clever economic policy and opening up to the capitalist world market. The fact that China was able to do this is a special case that was obviously never envisaged in this form by those who promised “development” through economic and political opening, but only thought about access to raw materials and sales markets. The Chinese Communist Party (CCP) under Deng Xiaoping relied heavily on foreign investment for economic development. However, Western capital was not allowed into the country unconditionally. The foreign companies had to enter into joint ventures with Chinese companies for their investments. In addition, a technology transfer to China and local Chinese suppliers were a prerequisite for the investments and only productive investments (e.g. the construction of production facilities), not purely financial investments (e.g. purchasing shares in existing companies) were permitted. In addition, there was a very restrictive currency and credit policy. With this strategy, China has succeeded in creating its own competitive companies. The fact that the CCP was able to impose these conditions on capitalist foreign countries is due to the fact that China had both a huge reservoir of cheap labor and a corresponding sales market. The size of the country made the location so attractive for foreign capital that the governments of industrialized countries made compromises that other countries could not achieve.
China has now been declared a “system rival”. The expectation that the opening of the market would also be accompanied by a political opening has only been fulfilled in rudimentary approaches, for example the creation of a rule of law as a prerequisite for capital investments. Many of the power techniques of creating economic dependence (sanctions, debt, takeovers, etc.) and of influencing civil society (financing NGO networks, journalists and activists for one’s own benefit) that have been successfully used elsewhere have hardly been effectively applicable from the outside in China. This is not a coincidence, but rather due to the fact that the CCP was very aware of the risks of external influence in opening up the country. If we understand the economic opening as a change in strategy by the Chinese elites after Mao’s death, it becomes clear that the continuity with socialist China lies in the nationalist goals and the inward-looking paternalistic policies of the CCP. Making China a country of global importance was already the goal under Mao (even before socialism) and in the real world this can be pursued more effectively with capitalist means than socialist ones. The fact that the development of China into a state on an equal footing (both economically and politically) is not easily tolerated by the current top dogs is shown by the increasingly loud declarations of hostility in western media reporting. And this, despite the fact that such economic development could be a “model” for many other countries according to the standards of the European and US governments. Unlike the governments of parliamentary democracies, the CCP can not only plan the course of a legislative period, but also pursue long-term strategies, such as the New Silk Road.
[…]
Conclusion
There are good reasons to resist the construction of new chip factories. The semiconductors produced there are the material basis of a technological attack that is capitalistically adding value to more and more areas of our lives and is aimed at a patriarchal optimization and control society. Emotions are captured and directed in order to optimize our lives in favor of the interests of its technocratic drivers. We are left alone and isolated. Our desire for social community cannot be fulfilled through digital interaction on the screen in “social networks”, but can only be suppressed. “Divide and conquer” is certainly not a new technique of rule, but it finds a new quality in the social atomization of digitalized society. Care, community, empathy and physicality lose meaning. The patriarchy, embodied by the German Minister of War and the “feminist” Foreign Minister, wants to restore the state of “war fitness”.
Preventing this requires unmistakable signals that we are not part of a “closed home front” and that the confrontational policy towards China in the fight for technological supremacy in the field of semiconductor and information technologies cannot be implemented without resistance. The history of social movements shows that there are many opportunities to resist (new) technologies as techniques of domination, especially on the material side.
New semiconductor factories and the AI hype that they materialize do not contribute to solving the climate crisis. On the contrary, they consume enormous amounts of resources. It makes particularly little sense to produce semiconductors in Europe, when the processing plants in which our smartphones are assembled continue to be located in East Asia because of labor costs. The climate crisis cannot be solved by AI or other technological developments, but requires profound social changes. These changes – we might call them social revolution – are prevented rather than promoted by an intellectual mobilization for the next wars for technological supremacy.
