Last year, a lot of semiconductor companies promised they’d be exiting the current supply crunch by the end of Q1 2021. In January, we started hearing companies say they thought the shortages could last until the end of Q2. Now, analysts are predicting they could actually persist for the next year.
There’s supposedly a 30 percent gap between semiconductor demand and supply, at present. The big-picture view of the market, near as I can tell, looks something like this: First, the pandemic hit and certain customers, like automakers, slashed their chip orders to the bone, on the expectation that vehicle sales would not begin to recover until 2021.
Second, the ongoing trade war between the US and China led some companies, like Huawei, to order as much product as possible. This kept TSMC’s 7nm and some of its early 5nm capacity utilization very high, especially in the first part of the year. Rising demand for AMD-designed products like the PlayStation 5, Xbox Series S|X, Ryzen 5000, and Radeon 6000 GPUs allowed AMD to become TSMC’s largest 7nm customer, replacing the demand from Huawei. When automotive chip demand surged back more quickly than anticipated, the spare capacity that could have otherwise absorbed that surge was already allocated to other customers.
Third, the semiconductor industry was already capacity-constrained in certain markets, especially when it comes to parts being built on 200mm fab lines using older process nodes. While these nodes were once expected to fade off the radar, the growth of IoT and the general proliferation of chips into every product on Earth has kept demand for this manufacturing technology high, with foundries reporting 80-100 percent capacity utilization.
Some IoT manufacturing is done on these smaller wafers, along with automotive chips, MEMS, RF silicon, and analog circuitry. Even before the coronavirus pandemic, chip manufacturers across the world were planning to install more 200mm fab capacity, with the number of 200mm fabs growing from 193 in 2017 to an estimated 217 by 2022. Mark Lapedus has written an excellent overall guide to the 200mm capacity problem and I recommend it if you want more information on the topic. Not all of the supply constraint in semiconductors is being caused by limited 200mm capacity, but at least some of it is.
Fourth, the complexity of modern device manufacturing means that any given product may incorporate many different chips, and shortages of any one of them may leave you without a product to sell, much as the shortage of ABF is already causing problems in the general industry.
J.P. Morgan analyst Harlan Sur recently released an investor note, saying “[W]e believe semi companies are shipping 10 percent to 30 percent BELOW current demand levels and it will take at least 3-4 quarters for supply to catch up with demand and then another 1-2 quarters for inventories at customers/distribution channels to be replenished back to normal levels.”
MarketWatch quotes Stifel analyst Matthew Sheerin, an analyst specializing in the technical supply chain, as saying, “We don’t see any major correction on the horizon, given ongoing supply constraints as well as continued optimism about improving demand in 2H21…We remain more concerned with continued supply disruptions, and increased materials costs, than we do an imminent multi-quarter inventory correction.”
It’s going to be very interesting to see if Intel can keep Rocket Lake on store shelves once it launches, and what demand for gaming CPUs will be if GPUs remain unavailable. I’d like to end on an upbeat note about how things will get better more quickly than a year from now — to be clear, I think it’s still possible they will — but the guidance right now continues to trend in the wrong direction.