Unlock the Editor’s Digest for free
Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.
Over the past couple of weeks, the world’s largest memory chipmakers have produced knockout financial results yet seen their shares fall sharply. What’s going on?
On Tuesday, Samsung, Korea’s largest company, predicted its second-quarter profits would rise 19-fold from a year earlier. Not 19 per cent — 19 times.
Its share price dropped 10 per cent.
Samsung and another Korean memory chipmaker, SK Hynix, each make up about 8 per cent of the MSCI Emerging Markets index, while Taiwan Semiconductor Manufacturing Company makes up another 15 per cent — I wonder how many investors in that index realise it is so dominated by three highly correlated global semiconductor stocks.
A week earlier, Micron announced a 15-fold rise in quarterly profits. The shares were initially marked 15 per cent higher to $1,240, offsetting a similar fall ahead of the results. Two weeks later, the shares were at $948.
You would be forgiven for thinking this all seems rather unfair — to publish blowout profit projections and be rewarded with a tanking share price. But investors need to be aware of those sectors that are prone to just this type of topsy-turvy behaviour.
The semiconductor industry is known to be highly cyclical. Booms have been followed by busts. The head of Micron, Sanjay Mehrotra, tried to argue that this cycle has ended due to the AI boom in demand for faster memory chips. So far, the market thinks otherwise.
Looking at Micron’s share price over the past 15 years, the cyclicality becomes evident. The best dates to buy the shares were February 2016 ($10), May 2019 ($32) and December 2022 ($52). In 2016 and in 2022 the expected earnings for the year coming up were negative and so the price-to-earnings ratio was infinite. In 2019, the price/earnings ratio was 13 times.
Between these dates the best times for stock traders to sell were May 2018 ($57) when the stock was on a price/earnings ratio of 4.6 times and December 2021 ($91) on a ratio of 11.6 times. The price-earnings ratios look low when you need to sell and look high (even infinite) when you should think about buying the stock.
Some content could not load. Check your internet connection or browser settings.
This is logical, if counterintuitive. The peak earnings of highly cyclical companies make price-earnings ratios look low. Then when the earnings turn down (and in this case go into loss) that ratio of share price to earnings rises ever higher.
Other sectors that show these fundamental characteristics include the cement industry. The market power of manufacturers tends to be local rather than global. There are few viable suppliers of cement in any location as it costs a lot to ship about. However, cement companies have a record of building too much plant when prices are high, leading to excess capacity and cost when the downturn comes.
The semiconductor industry has a similar reputation, especially in capacity for memory chips. Around 2005, flash memory capacity was in excess — following a boom in demand from smartphones and digital cameras. Prices fell so far they ended lower than DRam prices — DRam is the older sort of memory you have in your laptop, which is cheaper but slower. So they were using up the excess capacity, but at a cost to DRam volumes.
Samsung tends to be seen as building much of the excess flash memory capacity, perhaps pursuing sales despite the risk to profit margins. SK Hynix has just raised $28bn in an American depositary receipts placement — a form of overseas company listing in the US that makes the shares easier for American investors to hold. The objective is to fund a large build of new capacity — reviving memories of previous memory chip cycles.
One last item jumps out from the Micron report and accounts — management has chosen September as its financial year-end. Most companies opt for December. A September year-end always raises questions. It seems to be timed to show Micron’s financial position at its least stretched.
Most US electronic goods that might need these chips are sold around the Thanksgiving holiday, when the Black Friday retail discount offers kick in. Micron will have shipped out much of its chip inventory to manufacturers some time ahead of this date, so the shops are well stocked. Its financial statements, therefore, will tend to show lower inventory, higher cash and lower working capital than the company’s typical levels over the whole year.
For a company in such a cyclical industry, this suggests its true financial risk is higher than perceived without detailed analysis. Leverage makes good times look even better but also makes bad times awful.
Where do these shares go from here? I’m afraid I won’t make any prediction. If the AI boom does last for many years, these high-performance memory makers will be coining it for a longer period than normal and the new capacity build will be justified.
However, if the end buyers of all these chips — AI hyperscalers — fail to increase revenues enough to afford all this kit, the boom could turn to bust. Having strong revenue growth in AI is to be expected: the difference will be between “strong” and “enough to pay the bills” — and the bills are huge.
Right now the market wants to see more meat on the AI story before further funding is instantly given.
Those with long memories will not just recall cycles in memory chip stocks. This might also raise echoes of 2000. After the internet portal stocks peaked (Freeserve, Tiscali and others) the heat passed to the network builders — Cisco, Baltimore Technologies, Colt.
When they stopped making new share price highs, it was all over.
Simon Edelsten is a fund manager at Goshawk Asset Management.
