HP Labs has pioneered a new non-volatile memory they call the memristor - a form of resistance RAM (ReRAM) that promises to fix many of the problems of NAND flash: limited endurance; future technical viability; and slow write speeds.
They even think it might replace DRAM. That is aggressive.
But can HP - which is still doesn't know if it wants to be in the PC business or not - really drive the entire industry to a new storage technology? Maybe, but it will be an uphill battle.
- Marketing. HP buys half the world's disk drives and a big chunk of the world's DRAM, but that isn't the problem. Getting people to buy an unproven technology is. And HP hasn't broken trail on anything rad for years.
- Cost. No one has ever built memristors in high volume before. While HP knows how to do it in the lab, pounding out tens of millions of devices will take some tweaking. If yields ramp slowly the costs will hurt.
- Software. It's taken years to get flash to look like disks. When will Windows be ready for memristors?
- Hardware. Unless it is a drop-in DIMM replacement there will be hardware changes that are low-volume and thus more expensive.
Consumers Most of these issues impact cost. Consumers care deeply about cost - that's why they buy PCs - and not so much about technology. Can HP convince consumers that memristors are a good deal and not just another gee-whiz gadget from Silicon Valley?
More important, can HP convince Microsoft to invest in changes to Windows 8 to take advantage of the memristor's unique features? Historically software companies have been no faster than enterprise users to adapt their software to a new technology.
Enterprise Margins are higher in servers, but so are buyer expectations. No one is going to put a new storage technology into production until it has been out and tested for 3-5 years. It's taken flash drives that long.
The Storage Bits take Chicken, meet egg. Chips need volume, volume needs sales, sales needs customers and customers need a reason to buy.
NAND flash sidestepped this problem by winning the mobile device market where disks couldn't compete. That market justified the factories and economies of scale that drove NAND flash prices below DRAM and woke up the rest of the industry.
For all the hype around flash SSDs they consume less than 10% of all flash production. No cell phones, no SSDs.
HP's strategy to license the memristor to multiple semi houses makes sense. HP buys a lot of semiconductors and can certainly jump start the market.
But will that be enough to overcome the lead that flash has today and will increase over the next 2 years? I hope ReRAM succeeds - we need something better than NAND flash - but the critical piece isn't the technology: it's the volume-building strategy that will make or break flash competitors.
Comments welcome, of course. Best bet: get iPhone 7 as a launch customer. What do you think?