More Volatility Ahead

The entire semiconductor industry had a wild ride on the stock market this week, plunging on Wednesday and recovering on Thursday. This is just a sign of things to come.

The cause of this week’s volatility can be tied directly to a Morgan Stanley report, which said that NAND prices have peaked and will begin dropping at the beginning of 2018 because supply has caught up with demand. The report specifically targeted Samsung, TSMC and Western Digital, but the whole semiconductor industry felt the impact as investors reached for the “sell” button.

The NAND shortage is largely a function of a shift from spinning media to solid-state drives inside of data centers. Prices have dropped to the point where it is cheaper to store data on solid-state drives than on spinning media, and demand for NAND created a shortage that began over the summer. SSDs use less energy for a couple of reasons. For one thing, there are no moving parts. And second, SSDs are significantly faster at data retrieval, which means there is less wait time for servers. There are other benefits, too. SSDs can be packed together much more tightly, particularly with 3D NAND, and 3D NAND generates very little heat, which means less energy required for cooling.

The fact that one component can have such a broad impact is noteworthy, though, and it is likely to happen again—and not just for NAND—across a wide swath of the semiconductor industry. For starters, there are so many changes underway in technology right now that supply and demand imbalances will pop up unexpectedly in a variety of areas. In the past, it was much easier to predict demand because semiconductor companies reported their unit volumes. There are now fewer big pure-play semiconductor companies. Companies like Apple, Google, Facebook and Amazon are developing their own chips, and they’re not telling the rest of the world how many they’re making or how they’re being used. As a result, it’s hard to predict what demand will be like for components that are required across a broad range of devices, such as memory, passives, discretes or even wafers and materials, or for services such manufacturing or packaging.

Consequently, the whole market is getting drawn into what normally would be viewed as a problem in a particular sector. With so many unknowns, the reaction by investors is to hit the panic button first and ask questions later. It’s becoming much more difficult to figure out how all the pieces fit together and what else may be affected, so the whole industry feels every bump in the road.

The supply and demand equation becomes even murkier as new markets such as artificial intelligence, machine learning, virtual reality, automotive, IoT and medical begin ramping demand for components and services. Unlike data centers and cloud, which are at least well-known and localized, demand in these new markets will come from hundreds or thousands of companies, many of which are not public. Understanding how they contribute to demand ahead of time, how likely these companies are to survive, and the long-term impact of these new markets is pure guesswork.

Consolidation across the industry doesn’t help matters, either. There is an obvious impact when a systems company such as Apple shifts a contract from one processor vendor to another, or when a chipmaker like Qualcomm swaps from one foundry to another. The less-obvious effect is when smaller companies cannot get necessary components, or they have to wait several months to get on a foundry run to prove their concept in silicon—or pay significantly more while demand gyrates during a series of mini-bubbles.

Starting in 2001, when the giant dot-com bubble exploded, the semiconductor industry began cleaning up its supply chain, eliminating double- and triple-ordering and sharply reducing inventory. By the time of the Great Recession in 2008, inventory was at an all-time low. Even though the semiconductor industry felt the pain of the larger economic forces, recovery was much quicker.

As we head into 2018, all bets are off on inventory management. Supply and demand are becoming opaque on every level, from components to manufacturing, And while consolidation will continue in some areas, such as mobile phones, the explosion of companies pouring into all parts of the supply chain will only accelerate as new markets gain steam. Predicting how that will unfold will be nearly impossible, making volatility much more common and leaving the whole industry longing for the days when predictability and supply chain management were a rather mundane exercise.