Intel and Micron Will Cash In on 3-Bit NAND -- Eventually
Memory chips take over the world -- that's actually the subtext for this story. In the world of NAND versus NAND, 3-bit memory seems to be the current bet for manufacturers to continue the march to dropping cost of production and -- eventually -- making some money. The current path of choice seems to be 3-bit NAND, as Intel and Micron yesterday announced their new 3-bit-per-cell 34nm multi-level cell (MLC) NAND chips which will eventually be available from their joint venture, IM Flash Technologies.
Wait, did I hear a yawn? Before you become disinterested, realize that Intel and Micron, as well as a joint venture of Toshiba and SanDisk, expect this technology to be part of a long-term strategy for flash-style memory to take over much of the overall memory market from other technologies, driving demand and, in the end, prices so all the companies involved make out like bandits.
The memory industry has long been the dockside brawl of tech competition. In fact, it's easy to forget that Intel was originally a memory chip company that turned to microprocessors when it became obvious how much of a pain that business was -- huge competition, low margins, pummeled prices. According to Jim Handy of semiconductor market research firm Objective Analysis, that has continued to the point where the memory industry in general, and NAND flash in particular, go through long brutal periods of losing money, each trying to get market share and wait for one of those times when a jump in technology lets them make the big bucks.
Right now, Handy estimates that 1GB of the common 55nm MLC NAND costs about $2.50 to manufacture, and sells for, well, significantly less. Early this year, flash was selling for about $1 a gigabyte, much due to an inventory glut. Now it's selling at about $2.25 per gigabyte, so most NAND business is just about breaking even. That's hardly something to write home about. What they bet on is salvation from the cost reduction curve, because they've been dropping at about 40 percent a year, which allows the pricing brinksmanship. "As long as you're the one that has the lowest cost, you're going to profit at the expense of the other people," says Handy.
When the industry moved to two-bits-per-memory-cell technology, manufacturing costs dropped by nearly half. In the long run, when the manufacturing processes get yield up to reasonable levels, 3-bit memory could trim another 30 percent. Combine that with the 34nm form factor (or 32nm for Toshiba/SanDisk), and you're putting 32GB of data on a square that's about 11mm (0.43 inches for the metrically challenged) on a side.
The cost savings aren't available yet because it takes time to get the yield percentage of good chips up high enough, but they'll end up there, and the manufacturing cost for memory will be around 70 to 75 cents a gigabyte. And now to reveal the grand plan. Handy calls NAND "the fastest growing market in the history of semiconductors," which, if you know the industry at all, is saying a lot. It went from zero to $10 billion annually in under ten years. NAND has been pushing out all sorts of other memory, including DRAM, and that gets us to Intel's interest. The company doesn't want back into the memory game, but it does see NAND flash as strategic to its platform vision.
Eventually the size and cost advantages displace almost everything else -- including hard drive. Now demand skyrockets, memory companies can bid farewell to the concept of over capacity, and buyers can wave bye-bye to constantly falling prices. The companies still left in place get to make boatloads of cash. So the game for the vendors becomes holding out long enough that the weaker players fall to the side. Or, as Handy remembers from an industry dinner back in the 90s, Micron CEO Steve Appleton once said that he's happy if he loses money every year, just as long as his competitors lose a lot more.
Image via stock.xchng user svilen001, site standard license.