How do you feel about this article? Help us to provide better content for you.
Thank you! Your feedback has been received.
There was a problem submitting your feedback, please try again later.
What do you think of this article?
An Extreme Ultraviolet Lithography machine: the equipment needed by chip fab plants to make the latest 5nm devices. If you need one, it’ll set you back over $100 million. Don't bother shopping around - only one firm makes them.
Image credit: ASML
Over the last few years, an unprecedented situation has been developing finally coming to a head with the COVID-19 pandemic: a serious shortage of the semiconductor devices required to satisfy our craving for high-tech gadgets. For many, the seriousness of the shortfall in supply became apparent in 2020 when carmakers started laying off production line workers; the reason given was that even a basic modern car cannot function without many electronic components – ‘chips’ – and supplies were drying up. Explanations for this crisis have been many and varied. The people working for the chip manufacturers who were charged with forecasting demand were quick to blame the unexpected COVID-19 pandemic. They were right, to some extent, but the fact is this perfect storm has been brewing for some time.
In 1965 Gordon Moore, co-founder of Fairchild Semiconductor and Intel, forecast that the number of components on an integrated circuit would double every year, reaching 65,000 by 1975. In 1975 this prediction proved to be accurate, and he revised what has become known as Moore’s Law to a doubling of transistors on a chip every two years. Since then, the world semiconductor industry has invested huge amounts of time, effort and money to fulfil this promise, leading to the formulation of Moore's Second Law; that the capital cost of a semiconductor fabrication plant (fab or foundry) also increases exponentially over time. That too has proved to be valid, and at least partly explains the mess the industry has got itself into.
Fab and Fabless
In the early years of the semiconductor industry, a chip manufacturer would both design and make the device. But with the relentless demands of Moore’s Law, higher and higher levels of integration meant more and smaller components or nodes being crammed onto tiny squares of silicon. It became obvious only automated machines could construct these circuits with the required precision. Very expensive machines – see the heading photo above. And they have to operate in a sterile environment with only one-thousandth the quantity of polluting particles in the air allowable in a surgical operating theatre! With every new generation of say, a microprocessor, requiring a whole new set of machines, the economics forced big chip companies to close their foundries and out-source production. They became fabless manufacturers. So now there are only three major fab companies left in the world: TSMC in Taiwan, Samsung in South Korea and Intel in the USA. This dangerously close to a monopolistic situation has worked well enough up until a few years ago. Then international politics, a virus and even Climate Change intervened.
The seeming obsession of chip manufacturers with maintaining the validity of Moore’s Law has led to a decline in the support for earlier devices based on ‘obsolete’ technology. At first, the need for limitless processor power for server farms was the driver; now that’s been joined by training data sets for Deep Learning (AI) applications and data mining for crypto-currencies such as Bitcoin. Examples of new devices include the Apple M1 (AI for its tablets and PCs) and the Tesla D1 (AI training for its car auto-driving car ambitions). And yet the vast majority of embedded processors needed for nodes on the Internet of Things and, come to that Industry 4.0, could be made with decade-old semiconductor technology. Nobody wants to make these old, boring chips anymore; so legacy foundries were shut down and demolished. The remaining fabs can’t keep up with demand for their ludicrously over-powered products.
Apple M1 Max: 57 billion transistors and up to 64GB of fast unified memory.
Image credit: Apple Inc
In 2018, the United States enforced trade sanctions on China. Huawei, one of the biggest smartphone makers in China, then placed huge orders for chips before these sanctions were implemented. Apple and others soon followed suit, not wanting to be left behind. The US government had finally realised that unrestricted ‘out-sourcing’ led to ‘off-shoring’ and total dependence on a potential enemy for the country’s high-tech needs. Intel was struggling with its sub-10nm technology while TSMC was leaping ahead. Somewhat late in the day, the US government began implementing a policy of ‘re-shoring’.
The COVID-19 Pandemic
By late 2019 the SARS‑CoV‑2 virus was out of control and spreading all over the world. The hospitality industry and airlines began to suffer as lockdowns came into force during 2020. Car makers’ profits were high at the beginning of the year, only to crash as customers couldn’t travel to or even enter the showrooms. Manufacturers were forced to lay-off staff and suspend production in plants all over the world. Twenty-five years ago, apart from the Engine Control Unit (ECU) needed to comply with exhaust emission legal limits, the most sophisticated piece of electronics in the average car was the CD player. Now, even the turn flashers are microprocessor controlled. Car firms had no choice but to cancel their orders with the chip makers: an ever-expanding market for sophisticated electronic components suddenly dried up. This could have been a disaster for the semiconductor industry as car makers normally ordered chips in huge volumes, hundreds of thousands at a time. Over the years, vast profits had been made. To be fair though, without that demand there wouldn’t have been the economies of scale that made the consumer electronics industry possible. No smartphones, no laptop computers.
Fortunately for the chip makers, they experienced a classic example of the proverb “it's an ill wind that blows nobody any good”. The pandemic lockdowns may have killed demand for new cars, but with so many office-workers now ‘working from home’, sales of those smartphones and laptops soared. The consumer electronics industry boomed and hoovered up those surplus chips. As the demand for new cars picked up, the chip makers were selling their products elsewhere and the misery for car plants continued with customers returning, but dealers having no cars to sell them.
Yes, even climate change issues have had a part to play. Apparently, the chip manufacturing process requires very large quantities of water. Early in 2021, Taiwan experienced a serious drought which badly affected production at TSMC which, it is estimated, has a daily requirement of over 150,000 tonnes. Of course, this may have just been a freak weather condition; on the other hand, there is an increasing probability that it won’t turn out to be a rare occurrence.
As I write this in January 2022, the world is still engulfed by a third, albeit less dangerous wave of the COVID pandemic. Some sort of ‘normality’ is returning, mangled supply chains are straightening themselves out and shortages, at least in the prosperous western countries, seem to be lessening. But potential obstacles to the relentless march of ‘high-technology’ still remain:
- Over-reliance on a small number of silicon foundries producing all the basic semiconductor hardware for the whole world. Returning to the days of many, smaller fab plants, may kill-off the economies of scale, reduce consumer demand and ultimately slow the rate of technological progress.
- Despite the recent setbacks, China will seek to exploit the current chip supply shortage in order to increase its influence over world affairs. There are a lot of foundries in China able to produce ‘old technology’ chips, fulfilling orders that the other fabs in the world cannot. There is also the impact on TSMC when and if China makes a move to take over the running of Taiwan.
- Technology may help to mitigate the effects of global warming, but there is a danger it could make things a whole lot worse. Those fantastically powerful processor chips consume a lot of power, and as the tendency is to form them into arrays containing perhaps thousands of them for tasks such as AI training and running server farms, their ‘carbon footprint’ is becoming very significant. We may have to curtail some of our ambitions for a future lifestyle involving intelligent robots running intelligent factories while humanity relaxes thinking deep thoughts in the Metaverse.
If you think the chip shortage only affects manufacturers, try and find a distributor with stocks of Raspberry Pi 4 boards. You might strike lucky.
If you're stuck for something to do, follow my posts on Twitter. I link to interesting articles on new electronics and related technologies, retweeting posts I spot about robots, space exploration and other issues. To see my back catalogue of recent DesignSpark blog posts type “billsblog” into the Search box above.