Micron customers won't have to wait long for 3D flash memory
CEO Mark Durcan tells CNET that the company will start providing samples of the advanced memory technology to customers in the first quarter of 2014.
Micron CEO Mark Durcan said his company likely will start providing samples of 3D NAND to customers in the first quarter of 2014.
"We could probably do it sooner, but I don't know that we will," Durcan told CNET on Thursday. He add that's because of company strategy, not due to any technological issues.
Durcan declined to specify what end devices will first use the technology but said it will likely be high-end products. 3D NAND promises improved performance, including greater reliability and endurance, and a "significant cost reduction" compared to traditional planar flash memory, Durcan said. Basically, what it means for consumers are gadgets that are snappier, longer-lasting, and cheaper.
He added that the industry's transition to 3D NAND won't happen overnight because it will cost a lot of money to purchase new tools and change current factories to accommodate the 3D NAND production. In addition, it takes time to develop end uses for the new memory because the interfaces to the outside world will likely be different, Durcan said.
"It takes awhile to enable customers to use memory like that," he said. "I don't think it will be material to the industry in terms of impacting supply in a significant way until 2015."
The news about Micron's efforts in 3D NAND come only a couple days after larger rival Samsung unveiled that it has begun production on its own 3D Vertical NAND, which it calls V-NAND. The South Korean company said its memory will be used in a wide range of consumer electronics and business products, including solid state drives.
NAND flash memory has been popular in portable devices for years, and demand has soared as more smartphones, tablets, and other devices are developed. The global NAND flash memory market is expected to total about $30.8 billion in revenue by the end of 2016, up from about $23.6 billion this year, according to IHS iSuppli.
However, NAND is starting to reach its limits in terms of scaling to more advanced process technologies. 3D NAND has emerged as the most promising way to keep flash memory advances going. Unlike regular NAND, where memory cells are laid out flat on the surface, 3D NAND involves stacking the cells, sort of like creating skyscrapers to pack greater density into a smaller area.
Typically, moving to more advanced process nodes, such as building chips at 20 nanometers versus 32 nanometers, comes with significant performance and cost benefits. However, regular, planar NAND flash will no longer see the same benefits, and it's becoming technologically impossible to scale the technology much more, Durcan said. Micron currently is producing flash memory at 16 nanometers, or billionths of a meter.
"We have hit the limit in the mid-teens of nanometers for really effectively scaling planar NAND flash," he said. "One can argue whether you can build a 15-nanometer part or a 14-nanometer part, but there's a quality/cost tradeoff in there. It's right there somewhere. For significant improvement going forward, it's got to be something different. It can't be planar NAND."
Dynamic random access memory, which is commonly used in PCs as well as mobile devices, also will face changes in the coming years. Durcan predicted that the industry will become less volatile as companies move more slowly to advanced process nodes. Typically, DRAM pricing swings wildly based on supply and demand in the market, and that has caused Micron to frequently post losses over the past decade.
"I do believe the memory industry itself, going forward, is going to be different than it has been historically," Durcan said. "I believe this memory business is going to start the process of de-commoditizing."
Historically, chipmakers shrink their chips by about 15 percent to 20 percent every year to 18 months, which provides energy consumption and performance benefits. Now, the stretch between 20 nanometers and 16 nanometers for DRAM could be more than two years, Durcan said. And between 16 nanometers and whatever comes after that could be more than two-and-a-half to three years.
"It's clearly possible to keep scaling DRAM, but it's becoming prohibitively expensive," Durcan said. "There's not an economic driver to do it once you get significantly below 20 nanometers."
He added that some pieces of the market may want smaller geometries to drive performance or other factors, but it's unlikely that companies will quickly change entire fabs to the newer nodes after 20 nanometers. Rather, Micron and its rivals will look to advanced DRAM technologies, such as one similar to 3D NAND, to provide performance and power benefits.
"The industry now has the opportunity to deliver more value-added, differentiated products to the customers because that new, value-added product doesn't need to compete with a commodity one that's 40 percent cheaper next year," Durcan said. "It has to compete with a commodity product that's 15 percent to 20 percent cheaper next year, but that's a smaller hurdle for that differentiated product to get over."