The Memory Market, Short and Long Term

A quiet battle is currently raging over computer memory. The combatants, many of whom are also allies, include Japan and South Korea, DRAM and NAND, solid-state and hard drives. Although the short-term results are always in question, the long-term war is virtually a foregone conclusion -- because it's part of the revolution of mobile over desktop, and there's no question which side will eventually triumph.

This particular battle is over Elpida, the Japanese DRAM manufacturer that declared bankruptcy in February 2012. On one hand, Elpida is a ripe plum; bidders such as Micron, SK Hyrix, and a pair of private equity firms look at Elpida's valuable 2011 R&D work on DRAM, ReRAM, and especially the hot property, NAND.

It's because of NAND that DRAM itself is much less tantalizing than it used to be. In addition to Toshiba recently bowing out of the bidding, DRAM market leaders Samsung, Nenya, and Powerchip were conspicuously absent from the start. The latter two have market share troubles of their own (Powerchip especially feels Elpida's pain), and Samsung may not believe that an Elpida acquisition would assist them enough against the other top NAND competitors.

Micron, with Intel as its partner in NAND domination, is much more interested in Elpida's potential. As of May 10, the company is officially the front-runner in acquisition talks -- and one may wonder why, as Micron already has a considerable hold on the DRAM market. In fact, the company's recent debut of next-generation DDR4 memory may signal one final golden gasp for DRAM before the NAND age is fully established.   

So why Elpida, why NAND, and why now? Let's back up a little for those just joining the conversation...

From time immemorial, PCs were the center of our computing experience, and they were all built the same way: for storage, hard drives made up of spinning platters; for random-access memory, DRAM; and when you wanted to take your data somewhere, you used floppy disks or optical disks (CD, DVD, etc.). Throughout nearly three decades of loose talk about revolutionary advances in computer technology, this hardware paradigm persisted.

Until NAND-based solid-state drives appeared on the horizon, that is. Already, the comparatively pokey flash NAND used in USB drives has driven the floppy disk to (or over) the brink of extinction, and they're currently working with cloud storage and other agent provocateurs to topple the once-mighty optical disk regime. Virtually the only straw propping up the hard drive is the cost per byte ratio -- although increasingly less expensive all the time, larger-capacity SSDs are still exorbitant compared to cheap Terabyte-plus HDDs.

This protection is not available for DRAM memory, however -- NAND prices have been dropping so quickly that a July 2011 study concluded that NAND was already the cheaper choice. In fact, NAND may even be too cheap; both Samsung and Toshiba have geared back production, feeling that the market was in danger of being oversaturated. But the same issue affects DRAM, which many analysts feel is a silver lining to Elpida's collapse. Consolidation, says one theory, will provide more control over supply and demand, and bring profitable stability back to an infamously volatile memory market.

Elpida was an early symbol of this consolidation. Formed by the DRAM portion of NEC and Hitachi in the final days of the 20th Century, the company also picked up Mitsubishi's memory business a few years later. With the well-known Kingston Memory brand as a major shareholder, Elpida made a number of advances in DDR2, DD3, and GDDR, and was just barely investigating the NAND side of things in 2007 -- when DRAM prices and demand simultaneously plummeted along with the global economy as a whole.

"DRAM makers will have to do some serious thinking about their future," cautions Jim Handy, author of the aforementioned 2011 study. "Since the middle-1990s, the market has consolidated from 17 important players down to seven, and when PCs move to NAND, the pace of this consolidation will increase."

Plus, It's just not a PC world anymore. NAND is the memory format of choice for mobile devices, especially market dominator Apple, whose iPhones and iPads rely on Toshiba's NAND memory. Toshiba quit the manufacturing end of the DRAM business over a decade ago, selling its US facilities to Micron and dedicating itself to pioneering NAND development. When investors heard that Toshiba was considering purchasing Elpida (alone and then jointly with Hynix), the company's stocks took a tumble; as soon as Toshiba left the ring, share price immediately began to rise once more. Clearly, the market favors a focus on NAND -- just as Toshiba's NAND fortunes largely rest with Apple's continuing patronage.

SK Hyrix may be the wild card in the picture. Although healthy and resolute enough to continue without worrying about following Elpida into bankruptcy and consolidation, the company's finances will likely discourage any significant growth in the near future. This makes the acquisition of Elpida more than unlikely, although perhaps less than impossible. On the whole, consolidation may indirectly assist the South Korean manufacturer, as its own stability is intimately dependent on that of the memory market as a whole.

Which brings us back to Micron, which could conceivably leverage a more profitable and stable DRAM market -- not to mention Elpida's relationship with Kingston -- to invest more fully in NAND research and development. Instead of a disorienting transition, the "lighter" DDR4 is ideal to cushion the existing mobile market (i.e., traditional DRAM for tablets, especially Windows 8 devices), while increasingly positioning its NAND production to compete against Samsung in the next generation of devices. Ironically, the push for NAND was partially motivated by the unstable DRAM market in the first placet, so consolidation will also lessen the impetus for an hurried NAND rollout.

Nobody is questioning the ultimate results. Even if PCs persist in a Post-PC world, they'll migrate to NAND for efficiency's sake. And even if the memory market is shaken a few times, Samsung, Toshiba, and Micron will undoubtedly rise to the top -- and perhaps be the only manufacturers left standing.