Your PlayStation (NYSE: SONY) might be on backorder, or that new router might be weeks away, or the dealer might say they've only got seven models of that particular car left on the lot.
These aren't just minor inconveniences. The 2021 global chip shortage is wreaking havoc across swathes of industries, threatening to derail the fragile global economic recovery.
While politicos from Beijing to Washington debate how to shore up domestic supplies, chip foundries are swamped under a tidal wave of backorders, and the end seems nowhere in sight.
At the beating heart of this issue is one straightforward question: can't we build more factories to build more chips?
Likely for the same reason that we don't often build space new stations. Semiconductor fabrication plants are powered by the bleeding edge of modern industrial technology, and therefore constructing one is a capital-intensive investment, to put it mildly.
The fabrication plants themselves are honeycombed with chambers so clean that they could make the Centers for Disease Control and Prevention blush. Robots hum as they snake their way across the ceiling while pulses of extreme ultraviolet light wink within the airtight confines of multimillion-dollar machines.
But unlike space stations, fabrication plants have a shelf life of only five years or so. Moore's law dictates that the number of transistors--the tiny switches that ferry electrical impulses across a circuit board-- must double every two years.
The facts bear this out. In 2017 Intel's (NASDAQ: INTC) flagship processor, the Core i7 Broadwell-E, contained about 4.3 billion transistors spaced at about 14 nanometers. For reference, the Covid-19 virus is anywhere between 50-140 nanometers in size.
True to Moore's law, roughly four years later, in 2020, Apple (NASDAQ: AAPL) produced the M1 CPU with four times as many transistors, 16 billion, all spaced at a width of 5 nanometers.
For semiconductor companies, these timed leaps in technology every two years amount to an almost automated obsolescence. Each time the industry adopts a new standard, be it in the number of transistors or in the size of circuits, semiconductor companies must upgrade each facility along the semiconductor supply chain to keep up. At the same time, these companies have to develop whole new processes capable of cramming more and more transistors into smaller and smaller spaces.
Last year Intel discovered the high cost of failing to keep pace with the rest of the industry after it delayed the release of its next-gen 7nm processors. Since then, Intel's relevance has waned, and the company has been forced to cede some of its market share to rivals like AMD (NASDAQ: AMD).
There's also an extreme degree of care needed to produce these electronic brains. Chip foundries have to contend with components smaller than the smallest virus known to man. A single wayward dust particle could ruin an entire production run, resulting in a loss of millions. Therefore human hands never touch silicon wafers while they're in production. Instead, they flit upon conveyor belts, safely locked in airtight cartridges, being delivered by unseen hands from one stage of the production process to the next.
That process requires over 59 pieces of equipment, each likely costing several million dollars apiece.
For with darkroom experience, think of making chips as being similar to making prints. An insulated silicon wafer, coated with photoresist, acts like photo paper. The pattern of the finished circuit board, like a film negative, is projected onto the wafer as it's exposed to extreme ultraviolet light. The wafer is then "developed." After exposure, regions without photoresist react chemically, and then they are effectively washed away. This process creates the indentations that will go on to form the basis of the finished circuit board. These remaining indentations are then intentionally coated with impurities and fitted with metal connectors to control the flow of electric current across the board. Usually, this process is repeated hundreds of times until the desired configuration is achieved. Then, three months later, sections of the finished wafer are cut out, becoming chips.
Knowing how semiconductors are produced gives insight into the brutal economics inherent to the industry.
Sure, companies could build run-of-the-mill chip foundries for $15 billion apiece to help deal with the chip shortage. But then there's the cost of upgrading every few years to consider, the untold millions that could be lost due to failed production runs and the slow turnover due to a long and technically demanding production process.
The economic picture is so brutal that only three players really dominate the industry: Intel, Samsung, and Taiwan Semiconductor Manufacturing Co (NYSE: TSM), so everyone from Apple to Ford (NYSE: F) literally has nowhere else to source their silicon. Like the only restaurant in town on a Saturday night, plants owned by these three players are overwhelmed and struggling to keep pace with orders.
There may be an end in sight.
President Joe Biden has pledged to shore up America's chip supply chain through domestic manufacturing. Arguably, such pledges ignore the reality that a chip true assembly line can stretch across continents, with separate components made by a milieu of highly specialized companies. TSMC has pledged to spend 20 billion on new facilities this year alone. However, this news is too little too late as far as 2021 is concerned. It will take several years for TSMC's new fabrication plants to come online.
In the meantime, hopefully, you'll get your new PlayStation sometime before 2023.