Comment Did you ever! Fusion-io shares opened at $25, $6 above the $19 IPO price and valuing the company at $1.9bn. Is this a flash bubble?
The Fusion-io IPO was priced to value the start-up at $1.5bn, a sum that has now been exceeded, and which seems out of all proportion to the value of Fusion's relatively old technology. Unless new technology is introduced soon the shares seem set for a fall.
How can this view be justified?
A comparison between Fusion's ioDrive technology and Micron's latest P320h product illustrates the gaping performance chasm between the two.
Both companies supply PCIe-connected solid state memory using fast single level cell (SLC) NAND chips. These provide a layer of solid state memory between a server's DRAM and its direct or externally-attached disk drives. They enable I/Os to complete in microseconds instead of the milliseconds needed by spinning disks.
Fusion-io introduced its ioDrive technology three years or so ago and has become the market leader and glamour boy of the flash industry, eclipsing early leader STEC. It partook in several million IOPS demos, with IBM for example, and recruited Apple co-founder Steve Wozniak to be its chief scientist, a very astute marketing move even if the Woz is not that great at storage science. It has got Facebook as a high profile customer, investment from Michael Dell and also Samsung.
There have been worries about a concentration of sales in a very few customers and OEMs, making revenues susceptible to a sudden decline if a single customer turns off the order tap. These have been shrugged off over the past few weeks as the IPO share price has been ratcheted up to take advantage of positive investor sentiment.
Arguably that sentiment is too positive, and not justified by Fusion-io's technology, as well as being undermined by the slight number of customers and OEMs.
The ioDrive comes in various formats, some using slower multi-level cell (MLC) flash. A comparison of two SLC products at the 300-320GB capacity point, Fusion-io's ioDrive Duo and the Micron P320h is instructive.
Let's look at random read and write IOPS. The 320GB SLC ioDrive Duo does 261,000 random read IOPS with 512 byte blocks and 262,000 random write IOPS. Micron's 350GB SLC card, introduced earlier this month, does 750,000 random read IOPS with 4K blocks and 298,000 random write IOPS. That's a very large read difference indeed.
The ioDrive does sequential reads and writes at 1.5GB/sec, using 64KB chunks of data, while the P320h does sequential reads of 128KB data chunks at 3GB/sec and sequential writes at 2GB/sec, outclassing the ioDrive Duo.
Several suppliers are in the PCIe flash card market with performance under or roughly equal to the ioDrive, such as LSI, OCZ and Virident. TMS has introduced a 900GB SLC RamSan-70 offering more than 330,000 random read IOPS, 400,000 random writes, and more than 2GB/sec for sequential reads. STEC is entering the PCIe flash card space and it knows that Micron has set a bar it has to beat.
Assuming Micron doesn't screw up its pricing then server vendors and end-users looking for server application acceleration will go for Micron's faster card.
It's obvious that the competition has caught up with Fusion-io and can offer cheaper and slower flash, equivalent performance or radically better performance.
Unless Fusion-io has much faster ioDrive flash technology in its development labs then it is hard to see how its sales can continue to grow or even remain at their current level in the face of this competition. If they flatline or decline then the share price will likely decline and Fusion-io's glamour stock status will vanish as a Fusion-io bubble bursts. So, can it produce the technology goods to sustain its growth, and end bubble speculations or not?
Only it knows the answer to that question, and it isn't telling. Here's the response from Matt Young, Fusion-io's senior director of marketing for EMEA: "Would love to talk you about these and other things but due to the IPO we are under media blackout for the next 30-45 days." ®