After more than a year of global shortages, the chip industry has developed an ironic problem: It doesn’t have enough chips to build the machines and tools needed to make the chips the world so desperately needs.
Some of the largest makers of the complex, expensive tools needed to fabricate microprocessors recently began discussing contingency plans if they are unable to get enough chips and other components. Several have had to reduce their financial forecasts or limit their growth estimates for next year, disclosing that shortages of chips and components are the culprits.
Specifically, the chip equipment makers need programmable processors called field-programmable gate arrays, made by the likes of Intel’s Altera unit and Xilinx, among other components. FPGAs tend to be less powerful compared to chips designed for a specific function, but are useful for emerging technology such as 5G infrastructure, certain components inside big data centers and chip-manufacturing machines — where custom chips don’t yet exist.
“Our FPGA chips go into a lot of semiconductor equipment,” Intel's co-general manager of Logic Technology Development Sanjay Natarajan told Protocol. “We will have those discussions with [tool makers], where we say, 'Hey, we want this equipment,' and they go, 'Hey, we want some FPGAs.'”
The fact that there aren’t enough components to allow the chip tools makers to ramp up production is prolonging the shortage, and is just one example of the complexities around both solving it and predicting when it will end. The length of the supply chain, range of materials needed and even unpredictable events — such as a March fire at a Renesas Electronics-owned auto-chip factory in Japan or the extreme cold weather in Texas that triggered factory shutdown — mean all bets are off.
Part of the issue is data. To determine when the tool companies might be able to regain an ample supply of components and chips, industry analysts need more data, according to Syed Alam, global semiconductor lead at Accenture.
“I'm not sure if I have the data point to connect the dots to say X percent of the chip shortage is because of the tools not being available,” Alam told Protocol. “Because a shortage could be for a lot of reasons. You don't have this raw material, you don't have the substrate, you don't have the equipment ready, you don't have the people, you don't have the [factory] build-out done.”
The troubles the equipment makers are facing heading into 2022 are just a piece of a much larger concern. After a big shock to chip manufacturing because of the pandemic, there is also considerably more demand for microchips.
According to Wells Fargo analyst Aaron Rakers and his team’s projections, the semiconductor market could double within five years to $1 trillion, from roughly half that this year.
The pace of transformation in the industry means even high-ranking U.S. officials are reluctant to take a stab at determining when the shortage might end. To help promote a $52 billion government subsidy package to help alleviate the shortage, Deputy Commerce Secretary Don Graves made an appearance at a December chip conference in San Francisco called Semicon West.
But in a telephone interview after the speech, Graves said even he couldn’t forecast the end: “I don’t know that I can give you an accurate prediction on that."
The 2008 financial crisis may provide some guidance on how long the current shortage will last, according to Deloitte supply chain expert Chris Richard. Back then, tech and consumer electronics companies cut back on orders and burned through their inventory in much the same way automakers and others did in the early months of the pandemic. And when the economy began to recover, orders accelerated as companies tried to rebuild inventory.
“It was well into 2010 before some of our clients and semiconductor companies were back in a normal situation and customers weren’t screaming at them because their [production] line is down,” Richard said.
Discussing the shortage is made more complicated by the fact that there are dozens of types of chips made for markets that range from home appliances — which don’t need very advanced processors — to high-performance computing, which requires the latest silicon. As a result, different industries will operate on different timelines because it might be easier to get some chips more quickly than others.
For example, buyers of data-center chips don’t follow the same patterns as consumers. There may not be enough semiconductors or equipment to satiate demand from large data-center operators such as AWS, Facebook, Google or Microsoft, but those companies typically have tight relationships with the necessary vendors and make capacity plans that account for whatever chips they can get their hands on.
Wall Street analysts and investors often ask chip companies to make predictions because they are attempting to determine when the historically cyclical industry will begin to see supply in line with demand once again, or if they’ll go too far and find themselves with a glut of inventory. If there is an easy consensus among the chip giants, it’s that the shortage will continue through 2022; beyond that is anyone’s guess.
To ease the shortage, Intel has promised to spend $20 billion on two new factories in Arizona, and $3.5 billion to expand capacity at its packaging facility in New Mexico. The company’s executives have long pointed to 2023 as the minimum timeframe for a shortage to ease, in part because of the amount of time it takes to construct new factories and ensure they are up and running.
“But maybe a year and a half total, as you think about how long does it really take to unwind most of these issues and add the necessary capacity in the system,” Intel CFO George Davis said at a recent investor conference. “So I don’t think it’s over quickly. If you're looking at it from a cyclical standpoint, the question is, will we overshoot? Or will we have just what's needed to normalize in 2023?”
Intel rival TSMC is the largest contract chip maker on the planet, and derives a big chunk of its revenue from selling advanced processors found in computers and smartphones fabricated with the latest technology. Executives there are tight-lipped about its current situation, but it has promised $100 billion of capital spending to expand capacity over three years, and on an October call with investors, CEO C.C. Wei gave his version of a prediction.
“Let me say that while we do not rule out the possibility of an inventory correction, we expect TSMC's capacity to remain very tight in 2021 and throughout 2022,” he said. “This is because of our technology leadership position.”
According to Accenture’s Alam, companies started to build more capacity over the past year, which he expects will begin to come online in the next nine to 12 months. In addition, what he described as inflated demand will begin to fall off at around that time; companies have been ordering more chips than they actually need in the hopes they will get the actual quantity necessary to build products.
“I think this double-counting and double-ordering will begin to ease up,” he said. “There will be a correction in terms of the inflated demand.”
Unfortunately, the industry hasn’t learned much from the prior shortages and cyclical fluctuations, according to Deloitte’s Richard. He said that the industry can do better at dealing with future shortages across the chip supply chain by building in more excess capacity and, if a company is reliant on chips, holding a greater amount of inventory. They're both costly options.
Solving the over-ordering problem could help too. Because real-time exchange of information between various parts of the supply chain doesn’t exist at the moment, signals from different tiers are often amplified through several intermediaries as they travel from a chip buyer such as Apple to a chip maker such as TSMC. Finding ways to get more accurate signals about true demand could potentially dampen the instinct to order in bulk.
What’s clear is that there has been a significant, if hard to calculate, increase in demand for chips of all shapes and sizes, which the pandemic has pushed forward.
“The importance of semiconductors as a foundational pillar for the accelerated digitization (from enterprise to consumer) will continue in the post-COVID world,” the team at Wells Fargo wrote. “From the way we consume goods and services to the way we work (virtually; hybrid), and increasingly the vehicles we drive, we think the COVID pandemic and supply chain constrained recovery will continue.”
According to Graves, part of the rationale behind the Commerce Department’s interest is that the industry is made up of hundreds of companies focused on turning a profit, and it doesn’t spend much time thinking about a national perspective that includes economic and national security. Beyond cash, the federal government can use other mechanisms to ensure there is sufficient talent to run the new chip factories in the coming years and decades.
“We need to make sure that our workforce development system, our community colleges or universities, are putting out the type of talent that’s needed for the upcoming years and decades,” Graves said. “It’s certainly the Ph.D.s at the cutting edge, but it’s also the engineers or the trained technicians up and down the supply chain.”