Post

There have been two big news items about memory chips in the past couple of days, and it's quite interesting to look at them together. On August 21, Samsung approved a shareholder return plan for 2026, ranging from 90 trillion to 110 trillion Korean won, equivalent to 65 billion to 80 billion USD, the largest scale in South Korean corporate history, five times the previous record set in 2020. Where did the money come from? The AI memory super cycle and chip price increases. Almost simultaneously, Micron announced it will invest 10 billion USD over the next decade to build a research lab in Boise, focusing on next-generation memory, advanced computing architectures, and packaging technologies. Note, this 10 billion is additional and not included in the previously promised 250 billion USD US manufacturing investment. One is making a lot of money and distributing dividends, the other is making money and heavily investing in R&D. Both are memory giants but with completely different approaches. Samsung's logic is clear: having benefited from the AI memory cycle, they first distribute money to shareholders to stabilize valuation. One in five adults in South Korea holds Samsung shares, so there is political pressure as well. Also, Samsung is large, leading in process technology and capacity, so maintaining the status quo is sufficient. Micron's logic is: I am smaller than Samsung and can't compete on capacity, so I bet on technology. The rules of memory in the AI era are changing; HBM, advanced packaging, and next-generation computing architectures are the future moats. If we don't invest now, we won't have a place later. So these two news items are essentially the same thing: AI is reshaping the competitive landscape of memory chips. Previously, it was about who had larger capacity and more advanced processes; in the future, it will be about who bets correctly on the technology path. Samsung tells the market with dividends that it can make a lot of money, while Micron tells the market with R&D that it can make money for longer. Whose approach is right will be clear in ten years.