If you listen to the proclamations of outspoken flash peddlers, we're to believe that hard drives are dead, near death, or at least reading over hospice pamphlets. With the ballooning data sets found in the enterprise though, it's easy to scoff at such claims declaring them the propaganda of an overactive marketing guy. The thing is, flash guy might be right. The primary stated objection to flash adoption in the enterprise is cost per GB, which is interesting because it means a few things; the most obvious being that many IT shops are being run by decision makers who are inherently short-sighted and perhaps uneducated about modern technologies (seems impossible given the number of sales people in the flash business). Flash prices are dropping remarkably fast though, and according to the data we've seen, within the next 3-4 years mainstream enterprise flash might realistically be in the .50/GB area, encroaching mightily on the domain of hard drives.
If you listen to the proclamations of outspoken flash peddlers, we're to believe that hard drives are dead, near death, or at least reading over hospice pamphlets. With the ballooning data sets found in the enterprise though, it's easy to scoff at such claims declaring them the propaganda of an overactive marketing guy. The thing is, flash guy might be right. The primary stated objection to flash adoption in the enterprise is cost per GB, which is interesting because it means a few things; the most obvious being that many IT shops are being run by decision makers who are inherently short-sighted and perhaps uneducated about modern technologies (seems impossible given the number of sales people in the flash business). Flash prices are dropping remarkably fast though, and according to the data we've seen, within the next 3-4 years mainstream enterprise flash might realistically be in the .50/GB area, encroaching mightily on the domain of hard drives.
The next common argument in favor of hard drives is capacity. Today we have 6TB hard drives and with HAMR and new technologies en route, the HDD market will have us over 10TB per disk in the near future. There are many hurdles before HAMR is commercialized though and we'd be foolish to think that SSDs will not make similar leaps in capacity. Die shrinks and NAND quality improvements are certainly happening faster than platter density growth (on a percentage growth basis) and let's not forget that Micron, SanDisk/Toshiba, Hynix, and Samsung are already well on the path to next-gen flash technologies that will enable SSDs to make a major leap (capacity, endurance and write performance) as early as next year. While traditional SSDs currently top out at roughly under 2TB, there are some major steps forward coming that will continue to pressure the hard drive business on the capacity front, one SSD vendor sees a 16TB drive possible in 2016. This assumes sticking to the 2.5" form factor too, which allows for much higher drive density than large capacity 3.5" hard drives.
Of the three hard drive players left, Toshiba is the only one who has an investment in a NAND fab through joint venture with SanDisk. WD has made roughly a billion dollars in investments in flash storage companies like Stec and Virident, and has put substantial engineering into their HGST SAS SSD line; however they do not produce their own flash. Seagate is looking to re-energized in the enterprise flash drive market of late, but also does not have a fab. The big problem here is that, as these two evolve their products and try to bring more flash-based storage to market, the harsh reality of NAND supply and demand will rear its ugly head.
The only way to win with flash at a large-scale drive level is to have your own NAND, the client (consumer) SSD guys have learned this the hard way. What was a market of seemingly 100 ex-RAM guys turned SSD vendor, has been whittled down to 8 or 9 of any consequence. Even the bottom half of that list isn't making much profit selling SSDs because they're paying spot prices for NAND from one of the major vendors. As a result, these companies don't get the bonus engineering benefit when it comes time to build the drive, so the end product suffers as a result. It's untenable and unsustainable long term unless you're Micron, SanDisk, Samsung or Toshiba.
On the enterprise side, there's a bit more wiggle room, as companies like Fusion-io can use an engineering advantage to make a drive design that's NAND agnostic. NAND price fluctuations are hard on the margins for a company like that, but given their lower unit volume, they're better off negotiating long-term contracts than investing in their own NAND production. For Seagate and WD, who should intend to be the Ford and Chevy of the flash business however, they simply just can't rely on the existing NAND vendors to supply them long term. It might not even be possible given the growth in demand for SanDisk, Samsung, Toshiba and Micron's own flash products. The fabs would rather sell back to themselves and continue to use the inherent engineering benefit when using one's own NAND as a core competitive advantage while grabbing market share at others' expense.
It sounds trivial right? Go build a fab and voila, problem solved! But of course, it's not trivial; it takes years as well as billions of dollars to get it done. However, it has to be done for the hard drive vendors to stay relevant, and perhaps even in business. Hard drives aren't going away any time soon, but the writing is clearly on the wall and disk drive innovation alone isn't going to solve the problems the storage market faces over the next decade. Yes, a high capacity archival drive that somehow only needs power on the off chance data is requested is great, but it also sounds a lot like an SSD, which can do that already. "SSD for archival? Hahaha, we can't afford that!" Well maybe not today, but let's have a little more foresight; in as few as two or three years, that may very well be the case. The question is, though, whose name is going to be on that drive label?
