silicon wafer

Chip market’s boom-bust cycle takes an unusual turn

Image credit: Dreamstime

When semiconductors drop off their unit-sales trend line bad things often happen - but 2019 was a bit different.

It’s likely to be a good year for the chipmakers. In fact, it’s on course to be a very good year, according to Malcolm Penn, president of market analysis firm Future Horizons.

One reason is simply that 2019 was a bad year. For the semiconductor business, compared to recent history it was a very bad year indeed, although the real pain was pretty well focused. As has been the case many times before, the strong tides that afflict the memory market turned again. When the semiconductor market catches a cold, the memory makers often go down with the flu because that sector is so much more commoditised.

One surprising aspect of the 2019 downturn is that it was not a lot worse, as it saw an unusual downturn that is pretty rare: unit sales dropped rather than just levelling off. Penn points out that the unit-sales trendline is remarkably consistent, at around 8 per cent per year. “There have only been five times in a 35-year period when we have seen a negative unit growth rate. Negative unit growth is exceptional.”

What is different about last year’s drop in units compared to major semiconductor slumps of the past is that it was mostly about the wider economy and not about self-inflicted wounds caused by over-expansion of capacity. Bill McClean, head of US-based market analyst IC Insights, says he sees the same dominance of the global economy on semiconductors being reflected in recent years more than internal factors. One reason for external influences being so strong is that, since the dot-com bubble burst, the chipmakers have been very careful to not increase their wafer-processing capacity more than they absolutely have to.

Now a major force in chipmaking thanks to the growth of fabless customers such as nVidia, Qualcomm and more recently Apple, TSMC’s capital expenditure has a strong influence on supply. The company makes sure it has customers for what it can produce before it starts building work. Contrast that to the situation in the mid-1990s when the “Asian tiger” economies were spending big to muscle their way into the technology sector, following in the footsteps of Japan in the 1980s. When the initial internet boom petered out and serious cracks appeared in the economies of the Far East, over-capacity led to a rout in pricing as customers realised they had the stronger hand. What followed was a 20-year slump in pricing, which leads us to the surprising aspect of 2019’s downturn.

As news of a slowdown spread, TSMC and others slammed the brakes on their capital expenditure. A secondary factor is the way in which big corporations dominate the picture – most of whom are on long-term supply contracts that are painful for suppliers, but do not lead to sudden renegotiations every time the market slips, Penn argues. The result? Unit sales dropped off but pricing did not vary much beyond its usual behaviour. A chart of long-term average selling prices indicates they probably bottomed out in the summer of 2016 and may show further signs of recovery towards levels last seen at the beginning of the last decade. 

What does this mean for 2020? In all likelihood, very good things for chipmakers. Potentially, very good things indeed. Assuming the sector follows historical trends – and it has almost never bucked this trend other than in the aftermath of the 2008 financial crash – unit growth will bounce back and pricing will portably follow. “After a shrinkage of 7 per cent last year, you will see 10 per cent plus this year. That’s partly just the mathematics of it, but it’s also partly a reaction to the over-correction of last year,” Penn says.

That might lead to some capacity crunches in unexpected places, such as the raw materials themselves, which will tend to force prices up further. Penn notes 2018 saw a shortage of wafers. This was due to the companies who convert sand into ingots and then into sliced, polished wafers have been extremely reticent to invest in capacity themselves simply because the chipmakers say they will need it. They have seen that movie before. They have also been at the wrong end of an intensive cost-cutting programme that, even though they sell wafers for maybe a couple of hundred dollars if they are lucky, the finished product leaves the fab worth multiple thousands. A price cut of $10 means a lot to them, but not so much to the chipmakers, Penn argues. Even so, the chipmakers like to nail them down on price and their natural reaction is not to spend much at all.

Capital expenditure in wafers has gone up a little, but not by much. Supplies could easily run dry there, particularly for those companies who do not need leading-edge silicon. What investment there has been has gone into 300mm wafers, used by the more advanced fabs. The 200mm rump is the one that gets neglected, but this is where a lot of seemingly unimportant parts get made. Shortages of those can just as easily bring a production line to a juddering halt as a more expensive, glamorous SoC made on a 300mm line. If 2020 turns out very good, you can expect a bit of turbulence.

The result of all this is that Future Horizons is putting the case for a serious bull market in semiconductors for 2020. The company normally constructs three scenarios for the annual move in the market: bull, bear, or something in between. “This is the first time I’ve taken the bull forecast as being the forecast. We usually go more in the middle. That bull forecast is about 10 per cent, but frankly I was torn as to whether to say it will be higher. I think that 10 per cent is the absolute minimum we will get. It could be at least 15 per cent.”

It might not hit the 30 per cent of runaway booms in the sector of the past and the calming noises made around the global economy may not last long. The industry also has some structural issues that were obscured by last year’s shenanigans in the wider economy. Concerns are growing as to how much longer the smartphone boom can last without new applications driving demand for higher performance.

Penn sees 5G as being a damp squib in the short term. Sales may pick up on the back on shiny new communications technology, but many of the promised applications for 5G are not doing well. Virtual reality is too power hungry and lacks the kind of push that the games Doom and Quake gave to the PC in the 1990s and those that drove the 2000s’ console market. Boredom has set in on the IoT among consumers, now they have discovered the sheer effort required to make all this stuff work together, and although they are getting greater traction in industry, mass-market robots are showing few signs of life.

It's a testament to the diversity of the modern electronics sector that makes it possible to enjoy solid growth without new killer applications. Sometime soon, that growth is going to stall without something new coming along.

Sign up to the E&T News e-mail to get great stories like this delivered to your inbox every day.

Recent articles