Semiconductor cycle – The splurge

Spending splurge in full swing.

• Intel has announced that it will spend at least $33bn on leading-edge capacity in Europe in yet another sign that we are in the upward leg of one of the biggest cycles seen for many years.
• Intel will spend around $20bn for a leading-edge fab in Magdeburg in Germany, $13bn on expanding capacity in Ireland as well as long term research commitments in a series of EU countries.
• The carrot was the $47bn in state aid recently approved by the EU for leading-edge semiconductor investments of which the Magdeburg factory will soak up a good portion.
• This is why European semiconductor companies are against Intel’s investment as it will reduce the amount of subsidies that are left for them.
• This comes on top of new fabs in Arizona and Ohio and round out Pat Gelsinger’s make or break strategy to return Intel to the pinnacle of semiconductor manufacturing.
• Its rivals are not standing still and both Samsung and TSMC are intending to spend a mind-boggling $40bn each on capital expenditure this year.
• This comes on top of substantial increases last year meaning that a lot of capacity is likely to come on stream over the next 3 years just as demand is starting to soften.
• This is particularly the case because many of these investments are being made for geopolitical reasons rather than economic ones.
• Currently, 100% of the world’s leading-edge semiconductor manufacturing capacity is to be found either in Taiwan or in South Korea.
• Both of these are considered to be within striking distance of China and so there has been a strong desire to diversify leading edge more widely across the globe.
• This means that decisions are being made on more than just economics, which inevitably means that when the down cycle comes, it will be brutal and very painful.
• The causes of every semiconductor cycle are different each time but the result is always the same.
• This is because this industry has a fundamental mismatch between supply which is lumpy (fabs) and demand which is smooth.
• Consequently, the industry is often in a state where supply and demand do not match which is what creates the characteristic semiconductor cycle.
• This time around we are in what I would characterise as a supercycle (larger than usual) as a series of positive drivers of demand and constrictions upon supply have all occurred at the same time.
• These are the acceleration of digitisation triggered by the pandemic, supply chain disruption caused by lockdowns, war in Ukraine as well as China’s zero-COVID policy and geopolitical concerns causing more leading-edge fabs to be built in North America and Europe.
• Furthermore, the supply chain concerns have led to inventory building by device manufacturers fearful of having to cease production due to a lack of silicon chips.
• Hence, demand for chips has, in all probability, continued to outstrip the underlying demand potentially setting up a meaningful reversal of the current situation.
• SARS-Cov2 is slowly being recognised as an endemic virus which, in all likelihood, has now competed out the influenza virus and will take its place.
• The combination of mass vaccination and the milder nature of the Omicron variant means that hospitalisation and death are greatly reduced as a percentage of total cases.
• Hence, once the current wave passes, there a settlement in Ukraine and China abandons its doomed zero-COVID policy the outlook for a full return to normality will improve greatly.
• Normality is likely to be followed by a slowing or reversal of inventory building as well as a pause in end demand as the world goes back to the office and back to school properly.
• China is now dealing with its biggest outbreak since the pandemic began and I suspect that its efforts to fight it will prove futile and within 6 months, it will also bow to the inevitable.
• Hence, I think by the end of this year we might start to see geopolitical risks begin to wane and inventories normalise.
• At the same time, inflation is going to start biting during this year and I think that it is here to stay for several years.
• This is because central banks are refusing to increase interest rates to realistic levels because of what it would do to government debt repayments, the stock market and the real estate market.
• Furthermore, high inflation is exactly what governments who borrowed heavily during the pandemic need as it is the only practical way to raise taxes without getting the blame for it.
• Hence, governments have no incentive to tame inflation outside of the damage it will do to their electoral performance meaning that I think we have at least three years of this ahead.
• Hence, during 2023, the impact of inflation could crimp demand while at the same time inventories begin to wind down and new supply comes online.
• The down cycle is likely to be greatly exacerbated given the amount of capacity being built now and the demand factors that could all bit at the same time.
• Hence, if one is to be in semis it should be at the value end of the market where one can find growth companies like Qualcomm and MediaTek which trade on multiples that are way below much of the technology sector.