Communia

Chips are in short supply. Souring the hopes for improvement heralded a few months ago, the industrial world is still in turmoil due to a lack of components. Imperialist squabbles are aggravating the situation and driving the militarization of chip production.

The effects of the chip production bottleneck are filtering through the entire industrial network

Chips and microcontrollers for electronics and home appliances are in short supply. Canon finds itself unable to incorporate chips and teaches its customers how to bypass its own electronic checks.

Some of the most serious effects occur in the automotive production chain. The big companies keep announcing shutdowns of entire lines and plants, as well as slowdowns in the pace of production. So far this month alone, Ford stopped production in 8 plants, Volkswagen slowed it down and Jaguar Land Rover incurred millionaire losses after lowering its production by 34% compared to 2019 due to the lack of components. The list of companies does not stop there: Daimler, Toyota, Nissan etc. are in a similar situation.

And all this not only affects companies that depend directly on electronic components, steel and metallurgical companies that depend on orders from automakers also experience a multitude of stoppages, such as ArcelorMittal, and wind turbine manufacturers are facing both the shortage of components and the simultaneous rise in raw material prices.

But the industrial world is much more than just production or profit figures. No plant operates without workers, and the chaos has turned into a further worsening of working conditions.

In the automotive and metallurgical sector, the plans for closures and layoffs already scheduled in order to make the much-vaunted transition to electric cars profitable have been amplified by stoppages and production downsizing, with additional redundancies and temporary layoffs.

In the electronics sector itself, the situation is different, since the shortage of components is not really due to a drop in production. The demand for computers and devices - caused by the rise of teleworking and distance learning - far exceeds the production capacity of the foundries (fabs).

These foundries are plants full of expensive material employing a small number of workers and are located in the territory of large national capitals, but the plants that carry out the next assembly steps are much more labor-intensive and are generally located in semi-colonial countries. Here the pace of work has become infernal. The infamous conditions in assembly plants such as Foxconn (China) were already notorious before the pandemic, and have not improved at the height of demand.

In Vietnam, the assembly plants of Korea's Samsung continued to operate at full throttle at the height of the pandemic, cramming workers into makeshift dormitories which turned into pandemic hotspots. In India, there were several strikes and even a revolt in an assembly plant in the south of the country.

These are just a few examples, but they clearly illustrate that, whether they end up being on the side of stopped or of accelerated production, workers end up being squashed by the productive chaos.

A rigid and unadaptable production system

TSMC's fab in Taiwan

The fact that the first link in the chain, the foundries, have been unable to adapt to the new conditions during the almost two years of the pandemic seem surprising. This is due to the combination of both the deformity and gigantism imposed by the type of society in which we live and also due to the increase in large-scale imperialist squabbles.

The manufacturing process of electronic microcircuits on a large scale is an almost parodic example of a process molded and favored by the accumulation of capital, enhanced by the ability to allocate enormous masses of money invested in machinery and facilities which yield dividends multiplied in a predictable and almost linear way.

But this same optimization for squeezing dividends has not only multiplied the physical productivity of the process, it has primarily multiplied its fragility in the face of unexpected conditions and its enormous concentration in a few places and under the control of a few hands.

Read also: The war of the chips and the contradictions of capitalism, 6/14/2021

The drive to increase scale and produce increasingly more microcircuits from the same amount of silicon has led to longer, more expensive and more complicated silicon crystal growth and lithography processes... Each cycle takes at least six months and fabs are running at 80-90% by default. And the remaining 20% capacity leaves extremely limited responsiveness.

The process is a juggernaut virtually insensitive to feedback, no matter what happens in the rest of the production chain. Concentration, the need to maintain conditions of extreme air purity (which does not allow the movement of a large part of the processes from plant to plant) and the need to put as much machinery together as possible only push even more towards the concentration of all production in a very few plants around the world.

And building new plants not only takes time, in fact it is not really feasible because there is not enough machine production to equip them. The very expensive machines for the fabs are made to order and in very small numbers. There are bottlenecks and problems that cannot be solved just by throwing heaps of money at them.

When the degree of concentration of capital becomes an obstacle to the development of productive capabilities

World chip production by company and country of origin of the company

The degree of concentration of capital is truly monstrous: there are companies which practically monopolize not only the market, but also global production capacities.

Taiwan's TSMC, the leader in the mobile microprocessor sector, has 65% of the world chip foundry market, 40% of the world's logic circuit capacity and 92% of the production of chips with tracks of less than 10 nm (especially less than 2 nm).

Korea's Samsung and SK Hynix monopolize world RAM production (they have been involved in price-fixing scandals in the past), while Intel monopolizes the production of microprocessors for PCs and other devices.

Faced with this degree of monopoly, Chinese attempts to gain production autonomy are anecdotal in their global impact. And the rest of the world either manufactures microcontrollers and other minor chips or watches from behind the fence.

Capitalist chaos does not disappear even with global monopolies.

FPGA circuit

However, one shall not be fooled by this picture of established giant capitals, investments in fabs and processes still show a rather marked casino gambling character.

Due to the type of processes and the time it takes to pay investments back, there is often a significant mismatch between what customers want and what foundry owners want, bankrupting entire processes.

For example, the customer comes with a "blueprint" that can be lithographed at different track widths, but the owner needs to render profitable his processes producing a certain track width that may already be obsolete a few years or even months after construction. That's what's happening right now. There is a lot of production capacity at a certain track width (28 nm) that is free and available, while the rest is all clogged.

Producers - including TSMC - have invested in new plants and lines at 28 nm and are urging customers to use them as a solution to the bottleneck but nobody is using them because customers requiring for example microcontrollers prefer to use smaller nodes (22 nm) which in theory are cheaper for them as they can squeeze more chips per silicon disk, but in practice nothing moves because there is not enough capacity for that node.

On the other hand, customers at different levels have been looking for alternative solutions on their own, from Tesla, which has simply eliminated specific chips in its cars, to companies in the sector such as Qualcomm, which are opting to use the production capacities available in different kinds of chips.

The chips we have been discussing so far were mostly chips that are pre-designed for a function (ASIC) and whose blueprints are "printed" on the silicon by foundries, but there are also chips that harbor a large number of logic gates whose connections are not preset but are programmable after printing, these are the FPGA chips.

Of course, no one in their right mind would try to program a microprocessor - the core of a computer - on an FPGA, but they are theoretically useful for microcontrollers and other components badly needed by the pther industrial sectors. The point is that it is a cost/benefit decision, and producing FPGAs on a certain scale is much less cost-effective than standard static circuits:

While it is true that new design tools allow lower production costs for specialized chips (ASICs) above 50,000 units, since the COVID crisis there is no production capacity left available. FPGA chips are available, and we can make them do whatever we want. ‎...‎ We will continue to look at the ASIC and SoC side when the need arises.

Qualcomm engineer

Which brings us back to the original problem... It is no news that capitalism - and especially its main force, accumulation - multiply both the system's fragility and productive chaos the more they develop. The reality is light years away from the ideal delusion of a frictionless productive equilibrium, but much worse things than a "simple" industrial chaos are on the horizon. The big national capitals want to get in on the act, and not always in a good way.

Contradictions and imperialist moves

Samsung's second foundry line in Pyeongtaek

The effects of this struggle of capitals can be felt at ground level, out of which huge new fabs are growing in the countries with the most concentrated capitals. Anyone who has traveled by KTX high-speed train to or from Seoul in the last year will have seen the huge box of Samsung's new foundry in Pyeongtaek rising out of the ground, a symbol of the ambitions of the controversial heir to the Lee dynasty, owner of the RAM emporium, rising out of the ground.

The first production line in Pyeongtaek was founded after Lee received a pardon and was released from prison where he was serving time for bribing the previous government. In the second line that has just emerged from the ground, the company has opted to shift its focus to logic circuits - processors and controllers - with an investment of some $100 billion. The Korean state, as we already commented a few months ago, plans to invest hundreds of billions to stay ahead in its field of accumulation. But it won't be that easy.

SK Hynix, the other major RAM manufacturer, tried to expand its fab in China late last year but ran into a roadblock from its supposed U.S. ally. The machines needed for lithography on the scale SK wanted are made in the Netherlands, and the US blocked their move into China, upsetting Korean plans. To the enormous annoyance of Korean capital, the vast majority of Korean investments are in China, and China accounts for 40% of Korean semiconductor sales.

The result is that the U.S. goal - to stifle the Chinese technology industry "from below" - is directly confronting Asian imperialism and its interests. It is explicit for example in the statements of Intel's CEO:

Do we want to own the intellectual property, the R&D and the fiscal flow associated with that or do we want it to go back to Asia?

Does it make sense to help "the enemy."?

Taiwan is not a stable place. Beijing sent 27 fighter jets into Taiwan's air defense identification zone this week - does that make us more comfortable or less?

How do you compete with a 30-40% subsidy? Because that means we're not competing with TSMC or Samsung, we're competing with Taiwan and Korea. The subsidies in China are even more significant.

Pat Gelsinger, CEO of Intel

Indeed, the US "invites" Samsung and TSMC to invest in new fabs on US soil (whose subsidies infuriate Gelsinger). But make no mistake, it was immediately made clear that they need to accept a subordinate position vis-à-vis the US and its companies: the Biden government forced them to disclose their confidential data if they want to invest on US soil. Consequently, a part of the Korean bourgeoisie showed its discontent and stated the obvious: the US wants the data to undermine their bargaining power and plan how to get ahead of them economically.

The European Union, the technological trailing end, then began to announce its strategic plan for semiconductors and negotiated with TSMC and the Koreans to open facilities. Both the construction of the first fab on European soil by TSMC and the idea that the a href="https://www.euractiv.com/section/digital/news/eus-breton-on-asia-tour-to-foster-digital-cooperation-tackle-chips-shortages/">EU and Korea are jointly engaged in a "war-like effort" were touted by the European press.

The plan announces an official investment of 45 billion over 8 years, a pittance compared to the 100 billion absorbed by a single Korean or American fab, while announcing that it plans to move from producing 10% of the world's semiconductors to 20% by 2030. Of course, with limitations and partial export blockades.

If the Americans let them do so, of course. Intel announced a few days ago that it intends to monopolize the entire supply of fab machinery available on the market in order to eliminate competition and thus "lead the industry". And the attack is occurring on several simultaneous fronts, not only in the machinery, but also in the architecture that governs the blueprints themselves.

A few months ago, U.S.-based Nvidia tried to take over the ARM architecture to the horror of many countries and companies around the world which depend on it. The move was blocked, but a couple of days ago Intel announced that it will effectively run and control the RISC architecture standard in its V version, something which is interpreted by many analysts as a victory in its attempt to increase its weight in the worldwide design and production of microprocessors for mobiles and PCs (the R in ARM stands for RISC).

Towards the militarization of chip production

In case the bellicose tone of Intel's CEO and his allusions to war tensions were not clear enough, the company plans to build a huge new fab in Ohio with an investment of more than $100 billion.

The U.S. military is its first major known customer.

After years behind TSMC and now after a long journey and not without problems Intel is again (for a short time) ahead. We must certainly look to the future, where the fight will be more dizzying than ever, entering a scenario where we will no longer speak of processors every year, but of complete lithographic processes every 12 or 18 months.

In the midst of a trade war, military tensions can only grow and worsen. But what chips are also showing us is that a parallel process is underway: the transformation of the productive apparatus into a war economy, a machine of death and destruction on a planetary scale.

The current form of society twists the entire productive apparatus in its own image to the point of confronting humanity's most basic needs.

Technological development, which, oriented to the satisfaction of human needs, opens the possibility of an abundant and free world, thus becomes, as the expression of the growth and accumulation of the capital it effectively is, antagonistic to human development, to the point of threatening the pure and simple preservation of the lives of millions of human beings.

There is nothing we can do about this within the categories of the system: neither as "individuals" nor as "consumers" nor as "citizens". Only by fighting as a class, asserting the primacy of universal human needs in every place and circumstance where it clashes with the logic of profit, can we put this drift in check and open the doors to a new world.