Electronic connections: Economic fears stalk the chip business

Malcolm Penn, president of market analyst firm Future Horizons, has for many years characterised the wild oscillations of the semiconductor market as being under the influence of the “four horsemen”. If they all falter, a massive slump looms. If they ride together, the industry can look forward to boom times as it did in 2017 and 2018. This year, one in particular looks quite unhealthy, but the others are still cantering well.

The four horsemen are: the economy, fab capacity, unit demand and pricing. Over the years, the industry has apparently tamed at least one of its horses. In the 1980s and 1990s, overcapacity was a recurrent problem largely because nation states regarded semiconductors as being strategic and helped favoured companies build up their manufacturing. Often, the spending peaked during a boom that would then quickly come to a sudden, jarring halt. There was the home computer and gaming boom and bust of the mid-1980s, followed a little over a decade later by the initial internet boom and a collision with the bursting of the Asian-Tiger bubble.

The boom meant downstream customers started building up their inventory to ward off production problems caused by extended lead times. No-one wanted to hear the dreaded word “allocation”, which often meant they were not the ones being allocated the next batch of parts.

As unit demand eased towards the end of each boom, customers discovered they could take advantage of a glut of wafer capacity and renegotiate much better prices. Overcapacity would often go hand-in-hand with a slump in both unit demand and pricing. If the general economy turned downwards as well, that was the point that even large players threw in the towel and got out of their weakest businesses.

For about a decade, no-one has complained of overcapacity. If anything, the industry has teetered on the edge of severe shortages several times, peaking in the middle of last year. One of the reasons is the interaction between technology and commerce. Building and operating leading-edge fabs is now so outrageously expensive that only those with the deepest pockets can invest. Companies that used to boast of making all their own chips now routinely buy their more advanced chip production from foundries and even then the choice is now extremely limited. It’s TSMC as the most likely choice at the leading edge and possibly Samsung for those who can order high volumes. At this stage, no-one is ramping up production unless they have a very good idea that they have a customer for it.

The collapse of the Asian Tiger bubble saw the end of a period of rising average selling prices for chips. Through boom and bust, prices kept going down. This was not as the result of that bust but probably more due to the lingering effects of an alignment between private equity and the rise of the foundry. Fabless chip startups need a lot of cash but not nearly as much as when companies needed their own fabs. The FOMO that afflicts the investment community when the outlook is broadly good let to a glut of startups competing in narrow fields such as Bluetooth and WiFi.

It was not until 2016 that average selling prices finally began to trend upwards and they have been doing so since then, at least up to now. They have been helped by shortages caused by an alignment between steady unit demand and its conflict with a lack of capacity, but overall the chip market seems to have broken loose of its capacity for self-destruction.

The bad news is that it’s the wider economy that has embraced self-destructive behaviour. Donald Trump’s tariffs and threatened trade war have taken the shine off the global economy, though they might only be the spark rather than the fundamental cause of a bigger crash.

It’s perhaps no surprise that an industry that is so fundamental to so many others, from cars to phones, should have a much stronger connection to the global economy than it did in its early days when military and business-computer spending dominated sales. IC Insights, another market analyst firm, reckons the correlation between semiconductors and the economy has increased dramatically even since the turn of the millennium. From 2000 to 2009, the correlation coefficient was 0.63. If you exclude memories – which is prone to bouts of overcapacity through irrational exuberance – the coefficient rose to 0.91 from 2010 to 2018. It was 0.86 if you factor in memories. Even with memories, IC Insights the coefficient will rise to 0.93 over the next five years.

It’s the economy that is the great unknown in the latest forecasts. Whatever happens, it’s unlikely to be a boom year like 2018, which saw growth of close to 20 per cent. Future Horizons reckons 1 per cent in 2019 is the most probable outcome.

“2019 is difficult to predict, but you know it's not going to be a double-digit growth year. Can it get to negative? It could happen but it’s got to get pretty bad,” Penn said at the company’s January forecast seminar in London this week (22 January 2019). There is sufficient momentum behind both the global economy and the chip industry itself to prevent a crash, short of a major financial crisis emerging.

However, the combination of a tax-cut sugar rush in the US running out of steam, tariffs choking off demand and the unwinding of massive amounts of corporate debt in countries such as China and things could get ugly quite quickly.