| | | As long as the memory makers are making lots of money, they will over invest in order to gain share. That's how the memory market has always worked. If the memory guys expect the market to grow 30% per year for the next 5 years, they invest to grow 35% per year. If they memory market is expected to grow 5% per year for the next 5 years, they invest to grow 10%. This is basic business strategy in memory.
Your rendition of how the market worked historically is misleading. What happened historically in the Bad Old Days was that a dozen or more firms would invest in fabs with relatively cheap money because their home countries (primarily Japan, Taiwan and Korea) were convinced that being in DRAM was necessary and strategic and a great way to make money. That was the strategy that led to the gargantuan losses over the first decade of the 21st century. The countries and the firms learned their lesson in the hard way and stopped doing that by 2010 and the sector began to consolidate both through mergers and bankruptcies. When a dozen companies build fabs that have to run 24 hours a day, that produces a whole lot of chips. Especially when the technology is relative easy and the equipment companies do most of the work. Samsung managed to become the most advanced DRAM firm technologically and could sell their chips at the low prices that an oversupplied market led to and still make profits.
Nowadays, though, there are only three companies of any size left and they all have their own IP and process technology that is no longer easy and supplied by the equipment companies. The customers of these companies don't want any of them to go under which would create a non-competitive situation, which means that as long as they keep up technologically, they will get their share of business. Samsung's incentive in the past when there were many chip companies and they weren't vertically integrated as they are today with their mobile businesses was to ramp up and sell for low prices. Today, though, their mobile business is much larger than their chip business, at least in theory. Their incentive today is to keep prices as high as they can in order to injure their mobile competition. This keeps their margins in both their chip and their mobile business high.
That is the supply side of the business. Others have noted the changes in the demand side. I would just add, in order to make this explicit, that in the past PCs accounted for around 95% of the DRAM demand. Today, PCs account for between around 15 and 20% of DRAM demand and that is dropping. There are far more uses for it now than ever before and until a technology that is both as fast and as reasonably priced as DRAM, that trend will remain intact. The next 5-10 years will see a dramatic rise in demand for both DRAM and NAND and, IMHO, it will be difficult for companies to keep up with that demand especially when autonomous autos and various other AI related technologies mature.
So, yeah, IMHO, TTID, even though at some point in the next year or two I think that there could easily be a down period because the AI technologies aren't ready yet for mass production.
But they will be. |
|
