Global semiconductor sales fell 21.1% year-on-year in May, similar to the pace of contraction to April and March. But a look at the month-on-month change suggests the market has reached, or passed, a trough. Chip sales in 3m-avg. terms have now been rising since March, albeit slowly (Exhibit 1).
Exhibit 1: The semiconductor cycle appears to have reached a trough, but recoveries take time
While end-use demand for traditional electronic goods such as PCs and smartphones remain weak, anecdotal evidence suggests demand will start to pick up soon as devices bought during the pandemic are replaced. Trade data are showing a change in trend, with Asian exports of semiconductors having flattened off in recent months (Exhibit 2).
Exhibit 2: Trade data have flattened
Don’t expect a quick recovery
Those expecting the gradient of the upturn to match that of the downturn, however, are likely to be disappointed. Semiconductor cycles are rarely symmetrical, with steep falls generally followed by more drawn-out recoveries (Exhibit 3). And sales rarely return to the trend of the preceding upcycle. Rather, there is a level shift down in the trend.
Exhibit 3: The before and after of a trough in chip sales is usually not symmetrical
The chief reason for this is the industry’s very high fixed costs, with huge capital investment needed to build fabrication plants. The cost of producing one extra chip is therefore much lower than the average cost. Producers will keep making chips as long as they make marginal profit (i.e., marginal costs are lower than the price). Sustained high output amid falling demand creates inventories and means it takes time for prices and sales values to recover.
There’s no reason why this time should be different. Admittedly, AI is already translating into greater demand for physical chips. But AI will not be an immediate panacea for world chip demand. It’s a relatively nascent slice of the market, so it is growing from a small base. Widespread adoption will likely supercharge AI chips’ importance, but we’re not there yet and it may not come as quickly as many think. For now, AI is just one small subset of many types of end-use demand, which will generally face headwinds from the following factors:
Weakening global growth. The chip recovery will likely coincide with a slowdown in global growth as the lagged impact of tighter monetary policy and other financial conditions bites. Resilience in western economies in the first half of 2023 is set to mean higher for longer policy rates, which is clouding the growth outlook for 2024.
End of auto chip outperformance. Auto chips followed a different cycle to the rest after automakers cancelled orders en-masse in early 2020, leading to severe shortages until recently. But with the auto chip market coming back into balance and those order backlogs fading, this current pocket of strength will likely cool.
Slow recovery in PC and smartphone demand. Replacement demand should rise as the many consumer devices bought in 2020 age. But given the size of the last upgrade cycle (supercharged by 5G-enabled phones) we expect the coming one to be more muted.
No big boost to GDP growth likely
Given our view on the outlook for the chip cycle, we doubt it will be a strong tailwind to growth in Asia’s electronics-producing economies (Exhibit 4) over the next few quarters.
Exhibit 4: Electronics production is around 10% of GDP in some Asian countries
That is not a commonly shared view, though. Some governments and central banks have talked of the uplift from an electronics recovery later this year. Meanwhile, consensus GDP forecasts imply a strong pickup in growth in the second half of the year.
Exhibit 5 shows the annualised rates of growth that would be needed over the rest of the year to hit the consensus forecasts for 2023. For Malaysia, Taiwan and Singapore it’s about 6%. Those rates of growth are unlikely without a strong rebound in electronics.
Exhibit 5: Consensus estimates imply a strong growth pickup over the rest of this year
As we’ve explained in recent research, we don’t agree, and expect growth over the rest of the year to be much more subdued than the consensus (Table 1).
Table 1: 2023 GDP growth forecasts
Terms of trade will remain weaker than before the downturn for some time
Given their dominance in the export mix of some Asian economies, tumbling chip prices have affected the terms of trade. Of course, chips have not been the only game in town—the steep rise in global commodity prices since Russia invaded Ukraine in early 2022 has also hit Asian economies (apart from Malaysia, which is a net energy exporter). Still, we can see a noticeable fall in the terms of trade in Taiwan and South Korea.
With the recovery in chips likely to be very gradual, chip prices are unlikely to drive a big reversal in terms of trade in the coming quarters. Indeed, prices could be expected to lag any recovery in underlying volumes given current inventory levels and excess capacity in the industry.
Don’t expect a currency rebound
The deterioration in terms of trade driven by the chip cycle has weighed on currencies. Admittedly, there is not a noticeable underperformance among electronics exporters relative to other Asian currencies since world chip sales began falling in the second quarter of 2022 (Exhibit 6).
But again, that is because there have been many other influences. Singapore manages its currency and has had strong capital inflows. Taiwan also intervenes in FX. Malaysia has had a positive terms of trade shock from oil prices, while interest rates in South Korea have been hiked more sharply than elsewhere.
Exhibit 6: Currencies of chip exporters have not underperformed, but there’s a lot going on
Nonetheless, there is evidence of a relationship between terms of trade and currencies. That drives our view that an initially tepid electronics recovery will not be a big plus for Asian currencies over the coming quarters. We remain most downbeat on the outlook for the Korean won, as we don’t think the terms of trade or the currency will reach pre-pandemic levels anytime soon.
Structural factors drive long term growth
The short-term outlook for the semiconductor industry is mainly a cyclical story. But the medium to long term will be governed by structural factors and these point to a strong recovery further ahead.
The world will increasingly demand more chips as they become ubiquitous in everyday products (an electric car can have more than double the number of chips as a regular car). And the production of these products will also likely become more chip-intensive as industrial automation continues.
As well as goods, semiconductors are an increasingly large input into many of the services we consume. AI has the potential to supercharge that trend and will require many physical chips. It took 10,000 Nvidia GPUs to train the Chat-GPT model, for instance.
The main risk factor to the long-run outlook is geopolitics. Trade restrictions between the US (and its allies) and China threatens to fragment semiconductor supply chains.
The sustained exponential rise in chip processing power over the past was not just driven by technological advancement, but innovation in manufacturing and supply chain organisation. Therefore, the fragmentation of the chip industry could prove a substantial headwind to productivity growth.
Alex Holmes is focused on the economies of South and Southeast Asia, including detailed country-level analysis as well as research into broader regional trends. Alex has several years of experience in macroeconomic research, with more than half of that time focused on Asia. Prior to Oxford Economics, he spent much of his career at Capital Economics, following stints at ICAP, Barclays Bank and the UK Government Economics Service.