We are in the midst of a worldwide NAND flash shortage, which is raising the cost of SSDs. This comes as bad news to members of the Disk is Dead Marching Band and Chowder Society who have been predicting for the past two years that SSDs would be cheaper per gigabyte than HDDs any day now.
Let's take a look at what caused this NAND flash shortage, and what it means for people who buy storage systems.
What causes flash price fluctuations?
Too many casual observers of the semiconductor industry think of Moore's law as a basic fact of nature, like Newton's law of universal gravitation or the three laws of thermodynamics. Actually, Gordon Moore simply observed the current rate at which semiconductor makers could shrink transistors and predicted it would continue.
What chip makers actually charge for their wares is driven at least as much by the law of supply and demand as it is by any advancement in chip density. Making chips requires not just deep technical skills, but deep pockets. A dynamic RAM or NAND fab costs billions of dollars and takes several years from the time of investment to actual chip production.
The interrelationship between research, equipment development and the need for periodic large capital investments has led the memory business into a cycle of supply gluts and shortages.
At the shortage point in the cycle, vendors turn profits by selling as much product as they can at a constant, or even increasing price per unit. The profitable companies borrow to build new fabs. When those fabs come on line, the new supply drives prices down, and vendors have to sell their product at a loss. That's because, while it may cost $1.10 to make a chip you can only sell for $1, much of that cost is the interest on the money you borrowed to build the plant. That means you lose less money running the plant than mothballing it. Demand eventually grows to meet the new supply, and the cycle repeats.
Flash has only been a big enough factor in the storage business for NAND prices to affect the average user for the past seven or eight years. The closest we've come to a NAND flash shortage until now was in 2012, when NAND prices rose approximately 10% and, more significantly, the NAND foundries put customers on allocation.
While that price blip didn't have many casualties, SSD vendor OCZ found itself without sufficient allocation. It then went to the spot market and ended up with inferior product at a higher cost. That ultimately led to OCZ's acquisition by Toshiba in 2013.
NAND flash shortage tip: Place orders early
I'm hearing about the usual early signs of a shortage, with vendors extending deliveries on some SSDs from the usual 30 days to 45 days to 8 weeks to 12 weeks. The good news for the enterprise storage world is that vendors would rather sell their higher-margin enterprise SSDs than the lower-margin SSDs for clients.
So what does a NAND flash shortage mean for storage buyers? First, you should expect delays, and get your orders in for expansion flash 90 days before you're going to need the capacity. Expansion orders tend to get delayed when a vendor has 200 SSDs, and they can either be an expansion order or part of an order with a million dollars of array controllers or hyper-converged infrastructure (HCI) nodes.
As the shortage worsens, some vendors -- like OCZ in 2012 -- will find themselves without the supply of flash they need. Large vendors with long-term supply contracts will get SSDs before a local value-added reseller that is custom-integrating HCI offerings.
I've taken the precaution of adding questions about flash supply to my usual briefing questions, and I suggest you do the same.
It might be time for object storage to move to SSDs
How flash caching can benefit your storage structure
Cracking open Intel's Optane memory technology