[DigitalToday reporter Chi-gyu Hwang] Shares of memory chip companies that announced surprise results have instead fallen in a series of cases.

In a column for the Financial Times, Simon Edelsten (사이먼 에델스텐), a fund manager at Goshawk Asset Management, viewed the recent situation through the lens that the semiconductor industry is inherently cyclical, repeatedly swinging between booms and slumps.

Samsung Electronics recently forecast that its second-quarter operating profit will rise 19-fold from a year earlier. It was 19-fold, not 19 percent, but the shares fell 10 percent immediately after the announcement. U.S. memory chipmaker Micron also reported that its quarterly net profit rose 15-fold. The shares initially climbed 15 percent, but two weeks later fell to below the pre-announcement level.

According to Edelsten, Micron’s share price pattern is clear. February 2016, May 2019 and December 2022 were good times to buy. Those points coincided with periods when expected earnings were negative or the price-to-earnings ratio was close to infinity. By contrast, May 2018 and December 2021 looked like times of low P/E, but were in fact times to sell.

“When profits peak, the P/E looks low, and when profits hit bottom, the P/E looks high,” he said.

The point was that investors should sell when results are good and buy when results are poor.

There are also concerns about oversupply. SK hynix recently raised $28 billion through American depositary receipts. The aim is large-scale new facility investment. Edelsten said it “brings to mind past memory chip cycles”.

Some argue that AI demand could make this cycle different. Micron Chief Executive Sanjay Mehrotra (산제이 메흐로트라) argued that the cycle is over this time thanks to AI demand, but the market does not yet appear to agree.

Edelsten points to revenue growth at AI hyperscalers as a key variable. “It is natural for AI revenue to grow strongly. The question is whether that growth is sufficient to cover the enormous investment costs,” he said, adding that “the market is now waiting for more concrete evidence of AI performance”.

He also mentioned the dotcom bubble in 2000, saying that “after internet portal shares turned down, investment fever spread to networking equipment companies such as Cisco, but their shares ultimately failed to set new highs and the bubble ended.”