Fab Capacity Shortages

Planning fab capacity is a little like parachuting out of an airplane. Your chances of getting hurt are 50-50 every time you jump, and past experience doesn’t necessarily make it safer the next time.

At the leading edge, there is debate about just how much capacity should be added at 16/14nm, or whether the lion’s share of that investment should go to 10nm or even 7nm. At least part of the decision may rest on how much capacity Apple needs and where it needs it, and there are rumors flying around all the time in the manufacturing world about which foundry has secured Apple’s chip business, how much of it they have secured, and what exactly Apple is asking for.

The fact that Apple has enough clout to make a difference in their investment strategies is a big issue. A blip in Apple’s earnings could have a resounding impact in many places outside of Cupertino. And for all the naysayers on the Apple Watch’s future, its success or failure will ripple across the entire global semiconductor supply chain. Fab capacity is merely one aspect of that, although a large and costly one. Where Apple will place its bets also will likely create shortages somewhere else until foundries can determine whether they will get sufficient payback for their very substantial investments. And until those decisions become clearer—if they ever do—they will affect all other chips at those nodes as smaller companies seek to get their chips manufactured.

None of this is happening in a vacuum, of course. Along with Apple’s good fortune there is a big uptick in automotive, industrial, medical, consumer outside of Apple, and in the future that will include the chips used for cloud computing. The Internet of Things, and in particular the Industrial Internet of Things, is gaining steam, and that will create an almost unquenchable demand for new chips for sensors with communications and on-board processing. Demand will continue to escalate for years to come, particularly at established nodes, and so far there are no new foundries being built to provide more capacity (Mentor Graphics’ Jeff Wilson provides insights here here).

That could change, of course. There is talk about three new fabs in India (see Bettina Weiss’ excellent analysis here), and there is a potential for more capacity in China (see related blog by Semico Research’s Joanne Itow here). But all of this takes time—usually measured in years—and lots of money. As companies investing in those fabs question how much to commit and where they are in a demand cycle, the uncertainty will be felt everywhere. And those aren’t the only factors involved in these decisions. There are environmental impact studies and local politics to contend with, questions about the availability of resources such as electricity, water, rare earth elements, and many other issues that require alignment across a global supply chain.

Until then—and no one is quite sure when that will be—you can be certain that shortages in capacity will persist. Chips will take longer to get out the door, and too many respins will kill market windows and perhaps even entire companies. But in an industry filled with national ambitions to take part in this high-tech opportunity, it also is likely to fuel some very significant changes that will last far beyond a single process node. For the first time ever, uncertainty about process technology—where to place multi-billion dollar bets—is emerging as a disruptive force in the global semiconductor business. And while that will create shortages and impact individual companies in the short term, it also could rebound in many directions over the long term that no one has ever seriously considered.