Flip side to the rise of flash

Electronic News, Feb 3, 1997 by Walt Lahti

1996 will likely be remembered for some time as the year DRAM prices fell from grace to the gutter in one easy step. Millions of profit dollars evaporated into thin air, or at least took up residence outside of the semiconductor industry. DRAM manufacturers were hammered unmercifully. Too bad it is not all over yet.

There are still new market entrants that see promise in the DRAM. Additional capacity is being made available that will continue the pressure on pricing. No DRAM ASP is truly safe yet.

What is worse is that the malady may spread. The idea that because DRAM prices are down, all memory prices should be down can probably be countered by reason and holding the line. Memory prices are really set by supply and demand, even if there is a little emotion thrown in for good measure. So what is happening to supply?

Several of the established DRAM manufacturers have been eyeing flash. Over the past few years flash has experienced dramatic growth. Flash worldwide dollar shipments for 1996 will exceed both EPROM (the mother of flash) and mask ROM combined. Some DRAM manufacturers see the goose that lays the golden egg and are planning a barbecue.

Recently, some DRAM manufacturers have announced that they will convert a portion of their DRAM capacity to flash. Remember, DRAM is the elephant and flash is the flea. Even a small portion of DRAM capacity will give flash, at least the variety accessible to the interested DRAM manufacturers, a significant boost.

To make the scenario even more interesting, Intel is building mega-capacity for its flash product line. An existing logic fab is being converted to flash and a new state-of-the-art fab is being built exclusively for flash. Other heavy hitters in the flash world are following suit. It will take some time for the DRAM capacity to be converted, probably two to four quarters. Flash has not proven to be the easiest process to bring up or to maintain. Some DRAM manufacturers are ahead of the game and have been working on flash for a couple of years. The new Intel flash fab will not be on-line until next year. Some very astute industry sages have stated that flash growth in recent years has been stunted by capacity shortages. By mid-1997, I think it will be safe to say that the industry will not be plagued by any more flash capacity shortages. In fact, flash capacity will likely be on the verge of--if not wallowing in--excess.

This will lead to faltering flash ASPs. Not the dramatic free-fall experienced by DRAM, but a fall nonetheless. 1998 should be a year of intense competition in the flash world. And it does not stop there. A huge effort has been and is continuing to be undertaken by flash manufacturers to wring the most out of flash processes, reduce design rules and to improve yields. This is not generally true of the EPROM market. There are a few manufacturers, SGS-Thomson Microelectronics for instance, that are still giving it their all for EPROM. The downside is that there are more than 30 flash manufacturers and only a handful of EPROM supporters left. Flash manufacturers will simply be able to run costs down more quickly than the EPROM manufacturers. As flash prices fall, relative to EPROM, flash will look like the better option in many instances. This will be most evident in the higher densities where the ability to easily reprogram is a desirable and often required feature. UVEPROMs are really not that easy to reprogram, particularly if they happen to be installed in the end product when the need is discovered.

OTPs, probably the most popular form of EPROM at the moment, are not reprogrammable at all. What are you willing to pay for insurance? The EPROM manufacturers will be pushed to the point of losing a little on every deal and making it up with volume. EPROM will tend to be forced into the few sockets the flash marketeers do not want. The child may well kill off the parent in its rush to be king.

Wow, I hope I am wrong!