I’ve been researching this industry. On my first pass at it, I’m not very excited about these stocks at the moment. I think it would be prudent to patiently wait for Mr. Market to have one of his mood swings. Valuations are not compelling at the moment.
There is a lot of hype surrounding flash technology. Many companies making flash-based products have very high short interest (e.g. FIO, STEC). While these stocks will likely go down, they may not be compelling shorts due to expensive put options and high short interest affecting the borrow on the common stock. Basically… nothing in this sector is really compelling to me at the moment.
How I think this industry works
The technical side of things is very complicated and so is the process of buying these products. The vendors will have sales engineers to help figure out the technical requirements of its customers. Of course, the problem with a vendor’s information is that it is biased. Buyers will rely on value added resellers and opinions from friends in the industry to aid them in their purchasing decisions. The buyer’s fear is that they may pay too much or that the storage solution is insufficient for their needs (unreliable, not fast enough). The big picture is that understanding the technical side of this industry is very difficult when the people who work in the industry have a hard time with it themselves. Unlike other industries (e.g. cars), there are no standard benchmarks to help buyers compare products.
For every customer, there is some uncertainty as to which solution is the best and whether or not a product will perform according to the vendor’s claims. Smaller customers do not have the luxury of evaluating demonstration units. The established companies have a small advantage (with both large and small companies) as it is more likely that customers will trust them to meet their claims and stick around to provide support and spare parts in the future.
There is no one-size-fits-all storage solution. Every approach to storage tends to excel at one particular type of problem. Effectively, there are many different markets for all the different storage needs out there. There is minor overlap between the different markets as a customer may have needs that fall in-between two different approaches to storage. The major storage companies generally do not compete in every niche.
How the products work
Storage solutions are generally a combination of software and hardware working together. Most of the hardware is the same between vendors. Everybody uses the same drives. Different vendors’ hardware controllers do very similar things. The biggest area of differentiation and value-add is in the software. The software may implement various tricks and techniques to improve data integrity and performance:
- Checksums detect data errors.
- Disc scrubbing detects errors early.
- Write caching improves write performance. The better solutions will not lose data if there is a power outage.
- Read caching improves reads.
- De-duplication frees up space.
I see the storage vendors as being very similar to software companies. Like software companies, most of what they sell is intellectual property. A small part of the business is in providing services to customers to help them plan and integrate the storage into their overall system.
Open source/free software
Google created its own in-house storage solution which inspired the open-source Hadoop project. There are many companies adopting Hadoop as its storage solution. The threat is that these companies do not need to buy anything from the enterprise storage vendors.
Hadoop actually consists of many parts, one of which is the Hadoop Distributed File System (HDFS). Many enterprise storage vendors are offering Hadoop-related services/products as they think that their storage solution is more appropriate than HDFS for many applications.
Flash will increasingly become a part of enterprise storage. It has major advantages (vastly higher performance in terms of IOPS) and major disadvantages (high cost per GB) compared to traditional hard drives with spinning platters. Some applications will benefit from a mix of flash and hard drives to get the best of both worlds. Some applications may be best served entirely with flash drives and others entirely with hard drives. The storage vendors will need to improve their software to leverage flash technology and to handle its idiosyncrasies.
I am skeptical about many suppliers of flash drives. Many hardware companies have little room to add value and are basically in a commodity industry. Historically, companies in commodity hardware tend to fare poorly (including the companies that make high-end commodity hardware such as OCZ).
As far as enterprise flash products go (e.g. FIO), I am also skeptical. To achieve comparable reliability as an enterprise drive, one could use a larger number of consumer flash drives and use RAID (or other) techniques to achieve similar (or higher) reliability. This factor will limit the margins on enterprise products.
I do think that flash drives will put a huge dent in the enterprise hard drive market. There are some applications where IOPS (input/output operations per second) is the most important factor. In these markets, flash drives will dominate because they are a magnitude faster than enterprise hard drives.
I’m not sure how this will play out. On the low-end, commoditization will be a problem. Some users are figuring out how to meet their storage needs with commodity hardware. The Stack Exchange sysadmin blog lays out their decision to go with commodity flash drives over some type of SAN or a FusionIO card (an enterprise flash product). Throwing more cheap commodity hardware can solve certain problems. More drives in a RAID array increases performance. More drives dedicated to parity or replication safeguards against drive failure. Enterprise products have to add enough value to justify their high margins… and they need high margins to make back their fixed R&D costs from small volumes of product sold.
The value gap between enterprise and commodity products could continue to close. The trend has been for enterprise features to migrate to commodity hardware. One such feature is Error-correcting code memory (ECC memory). This allowed Google to use commodity servers (now with ECC memory) to control its hard drives instead of specialized RAID controllers with ECC memory. Because the servers had ECC memory, they could reliably perform checks for data integrity and it was ok to use (relatively) unreliable consumer hard drives. Google does not need to buy enterprise software or hardware.
The future of enterprise storage
I think that the future enterprise storage market will look very different than what it looks like today. Commodity flash and commodity hard drives will dominate over enterprise versions of these technologies. The substitution is already happening as many people are incorporating consumer SATA hard drives and consumer flash drives into their storage solution. I think that this trend will continue. On the software side, the winners will be the companies that develop software that works in tandem with consumer drives. One obvious area of innovation is to effectively use a mix of flash and traditional hard drives to get the best of both worlds (high performance/$ and high capacity/$). This is an open ended problem as there are different ways to use the flash drives to cache reads and writes. I am not sure who will have the best software and who the winners will be.
It is unclear to me if there will be more open source software that could serve enterprise storage needs (e.g. needs not met by Hadoop’s HDFS). Such open source software would likely lower profits as many customers will simply use free software, as is currently the case. However, some of these customers may pay for IT services to effectively use the software or they may pay for updates. This is the business model employed by Red Hat (RHT).
The future of enterprise storage stocks
I’m not exactly sure which companies will grow their businesses the most over the long run. I don’t see any of these companies as having an unusual competitive advantage like Microsoft or Intel. EMC’s reputation gives it a very small advantage but this advantage could erode very quickly if a competitor has superior products. My wild guess is that Netapp will grow its business the most, followed by EMC. Dell/Compellent and HP/3Par may lose market share to these giants. Many companies have paid ridiculous P/E multiples for emerging companies such as Data Domain (see this short thesis on VIC that in hindsight is a bad idea), Compellent, and 3Par. This suggests that the management teams at EMC, Netapp, Dell, and HP have flawed capital allocation skills. I find it curious that they all imitate each other in overpaying for emerging companies. Poor capital allocation will make it more difficult for shareholders to make money when it is being funneled to the VC firms behind these startups (who will be encouraged to fund more storage startups, increasing competition). At current multiples, I do not find any of these stocks compelling.
FusionIO (FIO) and STEC are involved in enterprise flash. FIO would be a good short if short interest wasn’t so high (34.4% of float). STEC has high short interest (17.7%) and has been pretty beat down so it doesn’t look like a compelling short to me.
Where to get information on this industry
The best resource that I have found is the excellent Storage Mojo blog by Robin Harris. The technical posts are excellent and there are many good comments on the blog. However, I find it interesting that the blog’s author was long STEC right at the peak in August 2009:
[Robin Harris] Disclosure: I’ve done work for Fusio-io and wish I owned stock. I haven’t done work for STEC, but do own their stock. Neo, what should bake your noodle is: do I own STEC stock because of my analysis; or is the analysis due to my owning their stock? Which came first?
Incidentally, I knew very little about enterprise storage back in 2009 but made a killing from STEC put options. The CEO had a tendency of selling shares right when earnings peaked. I got lucky. Anyways, I’m not sure if you should listen to experts with domain knowledge for stock picks.
*Disclosure: No position in any of these stocks.