Prices give us valuable information about supply and demand – or market power – if we watch them. Commodity storage prices – disks, DRAM and NAND flash – are particularly interesting since they often make up a significant percent of a product’s total cost.
Drive prices are sensitive to demand, while semiconductor prices gyrate based more on supply. Contract prices move slower than spot prices, but the two tend to move in tandem.
Flash dichotomy
But lately that hasn’t been true of NAND flash. While cheap thumb drive prices have fallen off a cliff in the last 12-20 weeks – they’d been at ≈$2/GB for about a year and just tumbled to ≈$1.50/GB – vendors results show stable prices.
As flash market guru Jim Handy of Objective Analysis put it:
The manufacturers’ earnings statements and their statistics (WSTS – the World Semiconductor Trade Statistics) all show prices remaining stable, while the brokers show a steady decline that began in May. This inconsistency really cannot be explained . . . .
Fools rush in
But that won’t stop StorageMojo’s crack team from trying.
Let’s look at the difference between contract prices – what big vendors who need an assured supply pay – and spot prices – what most thumb drive vendors pay for whatever is available. DRAMexchange.com has a large spread. Contract prices are almost $2/GB for 8GB NAND while spot prices are less than $1.50.
No new plants have come on line lately, traditionally a major reason for massive price reductions. Smart phones, iPads, iPods and SSDs seem to be selling well. (Apple’s Q3 results will tell us more as they buy half the world’s NAND flash supply.)
But the spot market demand isn’t what it used to be. Why?
The StorageMojo take
I see three likely causes. In no particular order:
- Thumb drive capacity growth has slowed. Partly a result of stable prices, but also because USB 2.0 isn’t fast enough to make thumb drives over 16GB very useful. USB 3.0 is bullish for flash consumption.
- SD/Compact flash capacity growth has also slowed. The megapixel race is winding down in the consumer market, so file sizes aren’t growing. The spread of RAW formats will mitigate the stagnation, but most consumers will never leave jpeg.
- Netbooks. The iPad knocked the air out of the netbook market. With many models flash-based a promising source of demand took a dive. Apple doesn’t buy on the spot market if they can help it.
Mr. Handy predicts that
. . . the capital spending that kicked in during the second half of 2009 will create such an oversupply in the second half of 2011. I wouldn’t anticipate any significant price declines before then. By this time next year price per gigabyte is likely to be closing in on 50 cents.
That means consumer SSDs will be less than $1/GB so every gamer will be able to afford one. That will drive SSD volumes for some time to come.
Courteous comments welcome, of course. Jim has a nifty report – “Understanding the NAND Market” if you need to know more.
Possible that data reporting for contract prices is inaccurate? How does dramexchange discover the price of private contracts?
Of Note. AOL uses Violin Systems for 50 TB Flash SAN with YadaYada virtualization.
http://www.computerworld.com/s/article/print/9191079/AOL_installs_50TB_of_SSD_boosts_DB_performance_by_4X?taxonomyName=Storage&taxonomyId=19
Why $2/GB is not-affordable to Gamers and 1$/GB is affordable?
IMHO the SSD-unit price is stable enough for already few years, just the performance (especially sata 6gig) and capacity grows. I cannot think about any game that needs more than few GB. Never mind current sweetspot 64GB. The future SSD sweetspot price will not be halved just the capacity doubled. Also more-bits-per-cell usually means latency will be longer and longer.
Why is USB 2.0 too slow for thumb drives over 16 GB, but okay for spinning 2.5″ 500 GB drives or 3.5″ 2 TB drives?
Steve,
The use cases differ. Thumb drives replace convenient low-capacity removable media for data transfer. 500GB USB drives replace tape for longer term storage.
Yes, there is overlap, but the low speed of USB 2 limits the convenience of thumb drives for larger files.
The real question is: when USB 3.0 is ubiquitous, who needs SATA?
Robin
@Robin: I agree that USB 3.0 has potential to speed up flash storage just as it is already speeding up large transfers to spinning rust, but I have rarely seen a USB 2.0 flash drive that reads or writes at anywhere near the speed that I regularly see from USB 2.0 hard disks. I can regularly get 25-30 MB/sec for large file access on hard disks, whereas most of the cheap flash drives I see are about 8MB/sec write and 15MB/sec read.
As for USB 3.0 displacing SATA, it depends if you were only thinking of external applications when you posed this question. eSATA will certainly have a run for its money if USB 3.0 hubs are efficient enough to display SATA port multipliers; for some reason there seem to be very few standalone port multiplier boxes available, which would allow access to multiple standalone eSATA drives from the usual single port on, say, a laptop. If USB 3.0 can do this well, I’ll be a convert.
For internal connections, you can draw 900mA@5V in total from a USB 3.0 connector, while each power pin in a SATA power connector can nominally supply 1.5A, and there are three for each of 12V, 5V, and 3.3V, so that’s a big difference in power available for spinning rust. SATA also has some command queueing advantages over the old-style USB mass storage protocols, but if USB Attached SCSI gets traction, that advantage would disappear.
-Jason