Do You Have Any Chips to Go With That?

Summary

• Trends in the global chip sector symbolize shifts in the broader economy.
• Near-term production difficulties are hindering the rebound, especially in autos.
• Chip inflation has returned for the first time in almost 30 years; chip capex is booming.
• A US-China stand-off over Taiwan could mirror the damage from the Yom Kippur war in 1973.

The semiconductor industry’s global economic significance was well appreciated before the recent Covid-19 crisis. Yet events through the past few quarters have highlighted the industry’s macro importance. And a chip shortage for key consuming sectors now constrains the global recovery, especially for the auto sector.

There is a growing recognition that the shortage will not end quickly. We will see protracted adjustments along several dimensions (price, production, trade, and capital spending), and these will be powerful enough to have sustained macro and sector-specific effects.

I think the chip industry symbolises the broader global economy as it adjusts to the post-Covid-19 world. That means strong demand and short-term inelastic supply will lead to more persistent price pressures over the next couple of years than we are used to. It means solid capital spending, which could add to global cyclicality later in the expansion. And it means growing geographic fragmentation driven by the rising geopolitical tensions between the West and China, reversing a few decades of rapid globalization.

40% of skilled US chip engineers were born abroad. And the most serious global geopolitical risks have shifted from the Middle East to East Asia, where 75% of global chip manufacturing occurs. For example, a brief military stand-off between the United States and China over Taiwan could be far more devastating than even the Yom Kippur war in October 1973, which led to the deep 1974-75 global recession.

The Key Global Commodity of the 21^st Century^[1]

Cotton was the key global commodity of the 19^th century; oil replaced it in the 20^th; semiconductors take that mantle in the 21^st. By ‘key’, I mean (at least historically) commodities that have inter alia sparked major wars.

On the eve of the Covid-19 crisis, the market for chips was (and remains) one of the most globally integrated sectors in the modern economy. Global final sales averaged about $45bn per month in the three months through July 2021, up 30% YoY (Chart 1). Industry sales in 2021 are on track to be about $536bn, up about a quarter over 2020. This is only about 0.54% of global GDP, although that low proportion fails to capture the significance of this key input to the functioning of the modern global economy. The integrated industry plays a more important role in world trade, reflecting the multiple jurisdictions involved in the complex chain of production. The gross chip trade in 2019 was $1.7 trillion, which was almost 9% of world goods trade.

We can broadly split the chip sector into three types (Table 1): logic, used heavily in the PC industry (42% of total); memory (26%); and discrete, analogue and other (DAO) (32%). The auto sector uses this last type relatively heavily. Mobile phones account for about 26% of chip usage; computers and servers 43%; industrials and other consumer electronics the rest (a combined 31%).

The geographical origin of chip demand has three dimensions (Table 2). The first is the domicile of the end user (device maker, e.g., Apple in the United States). The second is device manufacturing location. Here, Asia is far more important, with China accounting for over a third of global demand. The third is the true end user. Unsurprisingly, the pattern here is far more balanced, with these data underlining the tech trade surplus of Asia, and the deficits of the US, Europe, and all other countries.

Finally, we can divide chip companies into four types: integrated device manufacturers (e.g., Intel and Samsung); ‘fabless’ design firms (e.g., AMD and Nvidia – note chip manufacturing facilities are called fabs); ‘foundries’ (e.g., TSMC in Taiwan and SMIC in China); and outsourced assembly and test companies (OSATs). As the industry (and its end users) have developed, firms have become more specialized and so the industry more globally interdependent.

Causes of the 2020-21 Chip Shortage

Quantifying the global chip ‘shortage’ is hard. But it is evident in both delivery delay times and the consequent supply disruptions in some (but not all) sectors using chips. The problem has both general and specific causes.

The general cause has been a classic supply-demand cycle like the ‘cobweb’ model explaining swings in agricultural products, where a long gestation period means that this period’s supply responds to lagged prices. For the chip sector, the capacity growth fell short of the accelerating demand growth ahead of the Covid-19 crisis (Chart 2). Stronger demand reflected the increasing chip-intensity of global aggregate demand. This intensity was due to the advent of 5G, AI, electric vehicles (which use 3-5 times the chip content of conventional new cars) and the internet of things (many more basic items are chip-driven). The result was that manufacturing capacity utilization had risen and was already very high pre-crisis (Chart 3).

Specific problems also arose. The most important occurred before the Covid-19 crisis. In 2019, the US introduced a series of restrictions on exports to China, immediately dislocating the (highly integrated) global industry. In response, sanctioned companies –notably Huawei – ramped up their buying in global markets to build precautionary inventories. The potential ongoing effects that these restrictions, which the new US administration has tightened, are hard to discern.

Covid-19 also produced both positive demand and negative supply shocks for chips. Behavioural changes post-pandemic boosted demand for tech products, while (more recently) the spread of Covid-19 in ASEAN countries has affected production by OSATs, notably in Malaysia. Weather also damaged supply. Ice storms hit Texas in February, and drought limited crucial water supply through the summer in Taiwan. Finally, in February, a serious fire broke out at a Renesas plant in Japan—a leader in the DAO sector (key for autos).

The shortage has most heavily impacted the auto sector in the United States (Chart 4), and beyond. The problem has been particularly acute for Germany’s sophisticated car industry, which was already reeling from the emissions scandal (Chart 5).

The shortage has hit autos so hard for three reasons. First, the short-term supply disruptions have been most problematic for chips in demand by auto companies (e.g., the Renesas fire). Second, auto producers made massive cutbacks to orders in 20Q2 only to see demand rise quickly in 20H2; meanwhile, chip suppliers reoriented to other users, whose demand was rising. Third, automakers rely on just-in-time inventory management to lower working capital needs, leaving them especially vulnerable to supply cutbacks.

The Implications of Resolving the Shortage

In any market that suffers persistent shortages, prices tend to rise, absent administered controls. This has been true of global chip prices, although price increases to date have been insubstantial. The price of semiconductors in the US PPI has risen by about 1.8%, AR, since October 2020 (Chart 6). Over the same period, the US import prices of chips have risen 3.7%, AR. These price measures are ‘quality adjusted’ and so typically show rapid rates of decline, reflecting the high rate of productivity growth seen in the sector since the mid-1970s (Moore’s Law). Recent price developments in the sector are consequently at sharp variance with established trends in the past 30 years, which have been an important aspect of the low inflation theme over that time.

Moreover, the pass through of these higher costs to end products in the CPI is sensitive to effects from the shortage – the most extreme example of which has been the spike in used car prices (Chart 7). In the past 18 months, their price has accelerated the most (relative to the previous trend) versus other key products that use chips, even though the direct chip cost for cars is trivial (was 2%, now heading to 5%) relative to other products (25% for cell phones and 33-50% for computers and servers). Used car prices jumped as they became the main price outlet for consumers’ inability to buy new cars (used car prices are set in auction markets).

The ongoing shortage will inevitably dampen near-term growth (i.e., still a drag on 21H2 growth). This will show up in persistent inventory weakness. As chip production rises and bottlenecks ease, however, goods production is apt to rebound, lifting global growth in 2022 and beyond (Chart 2). Automakers will likely lift all types of inventories.

The global semiconductor sector is very capital intensive (and R&D intensive). The current global shortage has unleashed a wave of plans for higher investment, and these will be an important component of a broader capex boom which I expect in coming years. Government investment is also apt to rise as countries focus on geopolitical interests. At some point (as per the cobweb model), extra supply will come on stream as demand growth steadies, causing a supply glut and renewed price weakness.

Equity prices of semiconductor companies have risen sharply relative to the broader market in the past 18 months (Chart 8). Amid the incipient capex boom, the best performing have been those (such as ASML, the Dutch company) which provide chip-making equipment (Chart 9).

1. The section relies heavily on a report produced by the Boston Consulting Group for the Semiconductor Industry Association (SIA): Strengthening the global semiconductor supply chain in an uncertain era, April 2021 ↑