Chips are in short supply. The production chains of the semiconductor sector, have been in crisis for several months. Industrial branches that depend on electronic components – 60% of all factories in some French regions, for example- report that they are suffering from critical shortages and automotive production lines are forced to slow down their production. Producers of microprocessor manufacturing machinery and other components are running out of capacity. And this situation will last for long.
Table of Contents
- Chips in the imperialist conflict
- From the baroque dream to the mega-concentrated factories of chips
- The flip side of the huge microprocessor industry
- Chips: How to get out of this quagmire
Chips in the imperialist conflict
The huge microprocessor and RAM production plants (which also use monolithic transistors) are in the eye of the imperialist hurricane.
Chips production has become the most capitalized sector in history. It is hard to find an industrial sector that more closely reflects the ability of the forces of accumulation to shape social production.
Today, the largest and leading chips and RAM production plants (in a sector anachronistically called the semiconductor one) are concentrated in a handful of East Asian countries and belong to a few ultra-concentrated capitals producing the vast majority of components. Just 3 companies capture the majority of the world’s revenues. For instance, TSMC (Taiwan Semiconductor Manufacturing Company) owns 28% of the world’s manufacturing capacity making 1.8 billion cell phone microprocessors a year.
These are no small investments, building a new microfabrication plant (a fab in industry parlance), involves an investment of between $15 billion and $20 billion… not taking into account all the ancillary industry. Most of that money goes to the very expensive mountains of machinery (fixed capital) needed to produce chips. A notoriously slow and expensive process to boot.
However, these chips and components are of undoubted commercial and strategic importance, and the mega-concentration of their production outside the borders and direct control of major national capitals is in direct contradiction with the escalating trade war.
It is no longer enough for the U.S. to veto Chinese purchases of machinery and material to produce microprocessors as it has been doing for months. To give an idea of the scale of the problem for the US, the percentage of US-based worldwide production capacity has dropped from 37% to 12% in 20 years, which is obviously unacceptable to the US in the midst of a trade war. Hence the patriotic technological warfare plans of the Biden administration now approved by the US Senate and described by one senator as
one of the most important things this chamber has done in a very long time, a statement of faith in America’s ability to seize the opportunities of the 21st century.
Even Intel’s plans – the only U.S. company in the top four – to invest $20 billion in order to resurrect production in the U.S. and European imperialist sphere reflect this attempt at a global imperialist reorganization of microprocessor production.
The company announced in March its plan to respond to chips shortages in U.S. automotive industries… but doesn’t plan to start producing chips for at least six months. Maybe nine. The slowness is not accidental: the plan is based on the construction of two factories in Arizona. The self-confessed goal is to displace TSMC (Taiwan) and Samsung Electronics (South Korea), which currently produce the most sophisticated chips and have a 70% market share.
We plan to expand to other locations in the US and Europe, to ensure a sustainable and secure semiconductor supply chain for the worldPat Gelsinger, CEO of Intel
Of course, a $20 billion investment won’t even get the plan off the ground… so on top of that comes the tens of billions from the U.S. government plan. Intel’s own chief announces that the component shortage will last for several more years. He is well placed to know, because one of the main bottlenecks behind the current situation is the lack of microfabrication machinery on the market. A situation that had been going on for several years, and seriously aggravated by the attempt of several national capitals to expand production all at the same time.
But how did it get here? Why is chips production so concentrated and equipment so expensive?
From the baroque dream to the mega-concentrated factories of chips
The great luminaries of the early modern age dreamed of a machine that could solve problems mechanically, asserting (against the dogma of their predecessors) that logic could be – a priori – decomposed into a series of mathematical operations and that every problem could be solved if posed in the form of those operations.
But the technical basis, the way to build such a machine, did not appear until more than a century later pushed by the rise of capitalism. Certainly, the automatic loom, a machine whose operations were not encoded in its structure but rather in an external flexible program, was the main influence of the creators of the first computers.
A computer (more precisely one of its several processors, either CPU or GPU), from its origins to today, is a modified version of the automatic loom, where the needles are replaced by a series of logic gates which carry out the operations dictated by the program, one by one but at a high speed governed by an internal clock.
It is these logic gates that will determine the rest of the story. In principle and as far as computational theorists are concerned, these gates are an abstraction that can be instantiated on any material basis. In theory, computers can be mechanical, hydraulic, electrical, or even biological.
But this is not a neutral point. Ultimately it is in the material basis of these doors where the great success of this industry lies as a way of raising capital.
Logic gates can be built by combining a number of parts which transform incoming signals into an output signal according to a logical rule. For instance, one of the simplest gates is the AND gate, which outputs a 1 if and only if all incoming signals are 1. This gate is relatively simple to construct if one has a device which only lets current flow in one direction when activated by an input signal, simply connect two of these devices in series between a current source and the output.
The design of these devices evolved over time, from large vacuum tubes to the first small doped silicon transistors of the 1960s. All of these devices allow unidirectional current flow when turned on.
When transistor production evolved to allow transistors to be printed directly onto thin sheets of silicon, a semiconductor, the large-scale chips really industry took off.
This new process was initially presented as a much simpler and cheaper way to produce computers. No longer would transistors have to be assembled on top of a surface, the silicon surface itself could be converted into thousands of transistors through a photolithography process similar to the developing of photographic film.
As expected, rather than making chips and component manufacturing easier and more affordable for other competing capitals, the new process allowed for an impressive increase in the scale of operations and concentration of investment.
By a happy coincidence, increasing the number of transistors per board did not proportionally increase either power consumption or pose heat dissipation problems, so all that seemed necessary was to invest in further and further reducing microprocessor track and transistor sizes to achieve proportionally better yields at an almost constant rate. The detailed drawings of the microprocessors the engineers were working with grew to be as much as 50 meters wide.
In the golden age of the microprocessor industry, small capitals had no chance against the large concentrations that could move much faster than them, while the capitals invested in the big ones had an almost assured profitability. From thousands of transistors in the 1970s it grew to billions today, with huge masses of money invested in ever more expensive machinery.
The flip side of the huge microprocessor industry
Naturally, this picture of capitalist success and mega-concentration of capital has a flip side. Similar to what had already happened with the steel mills at the end of the 19th century and with the chemical industry in the 20th century, plants became increasingly huge and few in number, and, especially in the case of the chips industry, they ended up concentrated in a few countries. And that now translates into supply problems and imperialist scuffles.
Concentration also has an effect on the very design of microprocessors. Squeezing logic gates all together like sardines may be very advantageous from the point of view of designing circuits with maximum cost-effectiveness, but it starts to cause major problems in memory access.
The average distance from the logic gates to the registers and memory keeps growing and growing. The time it takes for information to reach its destination now accounts for as much as 70-80% of the total in each internal clock cycle. Similarly, the happy coincidence we were talking about earlier is no longer true, and now processors are starting to suffer from serious heat dissipation problems.
The long race to bigger scales also left other after-effects, such as an extremely long process unable to respond quickly to changes in demand. The process from starting a new batch of wafers on which microprocessors are printed until they leave the fab consumes 26 weeks or more, i.e., half a year.
And that’s without taking into account that an additional six weeks are needed for final assembly. What’s more, because of the immense investments, which often have to be recouped over five-year periods, companies are very reluctant to change the slightest bit of their machinery and manufacturing methods, which further slows down the reaction speed of an industry completely focused on continuing to invest in a single type of design:
The manufacturing process itself has become more expensive as it has become more complex, says Qing Cao, now a materials scientist at the University of Illinois at Urbana-Champaign. Cao worked at IBM until last year. The performance gains for each new generation of chips are also marginal—only about 10–15%, Cao says. […] But the semiconductor industry is loath to add expense by changing anything about the equipment already used to build silicon circuits. “We thought people would be open to using new tools,” Schmergel says. That was a pipe dream.
Chips: How to get out of this quagmire
There exist a whole variety of technologies that can not only get the industry out of this quagmire (and some probably will), but would allow chips and microprocessors to be manufactured in a more distributed and needs-driven way, from modular approaches to using nanotubes that are simple to manufacture and can be applied in ink form as semiconductors.
But under capitalism every step that facilitates or lightens production leads not toward abundance or the satisfaction of human needs, but toward the concentration of production in the hands of a few and scarcity. And at current scales, as we have seen, also to massive inefficiencies, scarcity and dysfunctional technologies. the rules of the system themselves have become a fetter on the productive capacities that the system itself fostered.
It doesn’t matter that Intel is now touting the idea of creating a safe and sustainable production chain for the whole world – showing its willingness to further reduce the global monopoly in its favor – or whether Nature calls for distributing production to low-income countries. South Korea just announced a $451 billion investment to establish the world’s largest chips component supply chain as well.
Increased scale, increased concentration… expanded capital. This is always the same solution which can only raise the level of contradictions instead of resolving them. Contradictions which express in each area and in a particular way the fundamental contradiction of the system: the increasingly violent opposition between human needs and capitalism, or what is the same, between its ruling class, the bourgeoisie and the worldwide class of wage-workers, the proletariat.
The needs of capital – to find ever larger placements for ever larger volumes of capital – the trade war and the plans of the imperialist contenders literally push against all those magnanimous plans which, in reality, barely conceal the desire to establish an even more restricted global monopoly.
Microprocessors and electronics are essential for today and tomorrow, but neither need they be manufactured as they currently are, nor will universal access and their production according to the needs of humanity come from the plans or kindness of one or another imperialist contender. Only the class movement towards decommodification and the new productive system it implies can free the productive capacities of humanity in the service of humanity itself.