Demand for data is set to increase as Internet of Things and cloud services continue to gain popularity. If so, will memory chips makers benefit from that trend?
Troubled Japanese conglomerate Toshiba Corp. put its memory chip business up for sale earlier this year, drawing bidding interest from US, Japan, South Korea and Taiwan investors.
Meanwhile, prices of DRAM and NAND chips have kept rising this year. So the answer to the question seems to be yes.
But on a closer look at the dynamics of the industry over a longer horizon, the answer is actually negative.
Hardware is not an easy business. Let’s take a look at the share prices of some giants in the sector. Intel share prices peaked in 2000. International Business Machines Corp. hit the peak in 2013, even though it has been actively branching into AI and cloud business.
True, data usage is expected to jump — from 4.4 zettabytes (ZB) in 2013 to a projected 44 ZB in 2020, and 180 ZB by 2025, according to IDC. However, that would most likely come along with declining hardware prices.
Moore’s Law, an observation from Intel founder Gordon Moore, says that chip prices halve every 18 months. That means chip prices would slump by 75 percent every three years.
Therefore, the whole market will only expand by 2.5 times even if chip demand soars 10 times within five years. That would translate into an annual growth rate of below 20 percent, not such an attractive growth rate in the internet world.
In fact, memory chip is essentially a commodity product; chip makers largely compete on prices, and most of the time, product prices are falling.
Worse, to stay in the game, chip makers need to set aside a large portion of their operating cash inflows for reinvestment, as we’ve seen in the case of leading chip makers Micron Technology and SK Hynix Inc.
Price wars happen from time to time and keen competition is the norm, leading to swings and erosion of the profitability of the chips industry.
To benefit from a surge in data usage, I would prefer instead to invest in companies that can gain from managing the growing volume of data and turning them into something valuable.
In other words, I like platform-as-a-service and software-as-a-service models far more than infrastructure-as-a-service.
This article appeared in the Hong Kong Economic Journal on July 18
Translation by Julie Zhu
[Chinese version 中文版]
– Contact us at [email protected]