Flash is winning a major portion of the storage market, but its success is bringing supply issues.
With the performance boost that flash and solid state drives (SSDs) give to computer systems, the takeup of the new technology has been slower than expected. For that, we can blame the recession and a fear of wear-out issues.
Now, though, production is moving into high gear. Many initial concerns have been put to rest, and demand is taking off.
Both tablet computers and smartphones are using huge volumes of flash chips today. While the phone business is starting to slow, demand remains strong in the sector. Meanwhile, the tablet space is taking off, as tablets supplant PCs. The only major offset is that digital camera sales are dropping, as phones get capabilities exceeding most dedicated camera products.
All of these shifts add up to a rapid increase in demand for flash chips. We are running into supply deficits, and the price reductions of recent years are slowing.
The NOR flash market doesn't get much interest these days, but production of these chips does eat up foundry space. Most of the issues circle around NAND flash, which the new electronics all use. NAND can be segregated into Single Level Cell (SLC), Multiple Level Cell (MLC), and some new technologies.
Today, the bulk of flash and SSD units use MLC. The enterprise-class product space demands SLC devices, which can tolerate many more write operations. The durability is required for write-heavy applications. This market was the slowest to respond to SSD and flash, as the large storage vendors overcame their inertia and recognized both the challenges of the technology and the negative implications to the traditional hard drive side of their business.
MLC's lower durability is adequate for phones and tablets, which are price sensitive. Next year, we'll see the addition of wearable computers, such as the smartwatch, computerized glasses, and the like. These will put further pressure on MLC demand.
Just over the horizon, the Internet-of-Things will create a tsunami of flash demand. Many of the "dumb" electronics in our lives will get smart. From microwaves that read bar codes to fridges that know what food is inside, the domestic scene is changing.
But most environments will evolve fast, too. Touch-screen kiosks will proliferate where anything is sold. All cars will have phone and monitoring capability. The European Union is looking to make those applications mandatory.
All of these trends will add to demand. Electronics OEMs are going to be looking for device manufacturers to step up and deliver more flash.
The supply picture is complicated by a bunch of other factors, too. There are really several distinct chip types, and their uses are differentiated enough that market dynamics vary considerably.
I predict that next year will see a reversal of the inertia of the enterprise-class space as all flash and hybrid appliances take off. The market for high-performance SLC-based SSD is also growing.
SLC drives, compared with MLC flash that can store two bits per cell, are universally expensive, since more chips are used to get the same data. The result is that most enterprise offerings come with slower, but cheaper, MLC versions, easing the pressure on the SLC supply side somewhat.
The supply side is moving forward, but it takes a while to build a foundry. Complicating this is the desire to make denser chips. Samsung is looking at 3D cell stacking, and this is almost a certainty, since the difficulties of making smaller cells with adequate reliability are starting to slow product evolution. The downside of stacking is it adds more process steps for each extra layer, so it impacts supply.
A recent explosion at a Hynix facility is probably just a bump in the road, but even an announcement by Toshiba that it is building more fab facilities looks inadequate to meet growth.
The implications are that demand will outstrip supply, at least over the next two years.
What's an OEM to do?
In the face of high demand and limited capacity, OEMs will need to compromise some when using flash in products. These same trends bode well for deduplication technology as a way to get more from less. Prices will remain relatively flat, and may even increase over the period.
Beyond 2015, the equation will depend on how much more foundry space hits the market. Other denser technologies, such as Resistive RAM may enter the game. However, they face foundry space issues, too, as well as the acceptance process for any new technology, so their impact will likely wait until 2016 and beyond.