More than two years ago, semiconductor industry pundits warned of an impending chip inventory correction. Turns out, they were right.
That “correction” has been cited repeatedly by chip executives in earnings calls over the past two or more quarters. They sometimes use different terms, maybe to avoid the negative connotation of the word “correction,” but what matters is that a lot of inventory of silicon is sitting on their shelves waiting to be sold. Excess inventory might also be on the shelves of their customers, OEMs usually, but even so, it means they aren’t selling products due to lower demand for PCs, smartphones and, less frequently, data center servers.
In one shining exception, automotive chips have been selling, based on March quarter results from a number of chipmakers. With some chip nodes, ironically, there are still shortages.
Some distilled history: The prognosticators back in 2021 noticed that with the pandemic and work and study from home, there would be a need for more PCs, especially, but also chips for servers in the cloud to support internet access as well. And, of course, when those orders were fulfilled, often with just-in-time methods, the apple cart was upended when the work from home patterns eased. With demand down, chipmakers had plenty of chips in inventory and then, too many.
So today’s correction is real and has been dragging on far too long already. Is the fabled inventory correction corrected? Fierce Electronics asked an industry expert recently.
“No,” the expert answered. Emphatically. That expert, Jennifer Strawn, head of global solutions and sourcing for Rand Technology, should know. She and her team hear constantly from companies in various nations wrestling with component and supply chain problems. Rand offers them resources to help (like finding needed parts in another country or from another supplier or finding customers for another company’s excess inventory), but some of the global supply chain inventory correction concerns defy correctability (if that is even a word).
As an indication the concern is widespread, GlobalFoundries CEO Tom Caulfield told investors on an earnings call May 8 that excess semiconductor inventories are “coming down more slowly than expected” with a rebalancing of inventories with demand to extend at least through the second quarter. More to the point, he projected “the return to more normalized inventory and demand levels…will mostly likely take well into the second half of 2023.”
RELATED: Familiar earnings news: GlobalFoundries latest to signal caution
Rand has seen companies in multiple industries sitting on the sidelines waiting to spend more on capital expenditures because the current growth cycle is so uncertain. “Everybody has their hand on pause,” Strawn said. The inventory correction is partly (sometimes, mainly) to blame, yes, but also the geopolitical climate is wonky, even scary, interest rates are high and recession worries loom.
“Companies can’t spend more money now and it’s hard to bleed through [inventory]. We thought by the end of ’23 we’d start to see most of it absorbed but for the economics,” she said.
The automotive sector, which got hit in 2021 with a shortage of needed chips, now faces the complexity of planning for electric vehicles. Auto executives don’t know when the charging infrastructure to support their EVs will be sufficiently advanced to warrant full EV production. “Can the infrastructure be ready by 2030? That’s very uncertain, so auto still has to invest in gas models, and which ones and how many?”
The converse for auto chipmakers is finding capacity to support chips needed for older gas models as well as supply chips for newer EVs and charging infrastructure. “Not one customer has said we think we are in the clear,” Strawn said. She included in that group Texas instruments, NXP, ST Microelectronics and Infinion.
A related capacity issue has developed because some chipmakers are focused on advanced chips, even while some in the auto or medical industry still rely on older chips that are bigger and take more room on a wafer.
“There’s still demand for older tech and industries such as auto and medical cannot move or redesign to a smaller wafer [or chip]. They say the new design is not a proven chip yet and it is costly to redesign,” Strawn said. A trend to replace MOSFiTs with smaller MCUs is one example: four MCUs can be produced on a wafer taking up the same space as a single MOSFiT. Obviously, chipmakers find the MCUs far more efficient with greater revenue potential.
While industries such as automotive and medical may struggle to balance the right chips for their designs, a potentially bigger challenge faces generative AI app engineers and cloud providers who will demand access to powerful GPUs for AI training and inference.
“Nvidia is the GPU leader, no question, but they can’t produce them fast enough to support demand” for AI needs, Strawn said. “The demand is completely unprecedented. We’ve not seen this advancement with so much technology at one time... It can’t happen as fast you’d like.”
With GPUs facing such demand, an AI chip shortage could certainly emerge and “we’d find ourselves back to 2021,” she said.
Against the backdrop of economic and geopolitical confusion and an uncorrected inventory correction, Strawn nonetheless remains an optimist about a robust global supply chain, at least in the mid-term but not really in the short term, (and who knows about the long term.)
“If a company has the appetite to make it through 2024, we’ll start to see significant change,” she said.