Slowing Capacity Adds

August 24, 2008

Q2 was yet another quarter characterized by excess supply, soft demand and aggressive pricing. SanDisk had been hoping for more. After all Q2 has become something of a second Xmas driven by Mothers’ Day, Fathers’ Day and school graduations. This year, Springtime Santa only brought lumps of coal.

As discussed in my last post, Q2 was a terrible quarter. Q3 doesn’t look much better. Time will tell about Q4, though analysts don’t expect much.

In his 2008 Flash Memory Summit Keynote on 13 August, Eli characterized the challenges facing the NAND industry as threefold: Falling US Consumer Confidence; Excess Supply/ Commodity Pricing; and 3rd year of ~60% Annual ASP decline.

Eli’s slides from the 2008 Flash Memory Summit Keynote.

Falling US Consumer Confidence is softening demand and is certainly beyond the control of the NAND industry. On the other hand these three years of 60% annual ASP declines is industry-wide, self-inflicted pain.

A 60% annual ASP decline exceeds the rate of cost declines of even the most efficient fabs. Demand continues to grow, but supply is growing even faster as mega-fabs pump out NAND at record volumes and at record efficiencies.

As described graphically in the slide below SanDisk believes that new markets taking off will drive demand back into balance with supply. Profitability will be restored.

The slide above purports to show that the self-correction of slowing capacity adds, will lead to growth with profitability.

I find this slide confusing. Below is my revised slide of what I think SanDisk is trying to say.

Basically both supply and demand are growing, but right now there is more supply than demand resulting in excess capacity. By slowing the rate of growth of supply, until new markets take off, the industry can minimize the near-term pain while moving up the transition for profitability.


One would think that these NAND companies would already have cranked the spigots down.

Fortunately or unfortunately it’s not this simple and isn’t going to happen near-term for several reasons. First, Mega-fabs are like ocean liners, they don’t turn on a dime. Second, market share is a big deal, seemingly worth fighting for.

And last but not least, huge demand seemingly looms around the corner (2010ish). If and when this SSD/Mobile demand kicks in, those with the supply at their fingertips will be sitting pretty. Winners may well be determined by those who can stomach this near term pain.

Dreams of 2010ish profits may be nothing more than a siren’s song luring the captains of this industry towards the rocks of a capital crisis. On the other hand, this could be the real deal.

Stakes are high as solid state storage goes mainstream in the digital age.

Slowing Capacity Adds

While the spigots aren’t being cranked down, they’re not being entirely opened either. Pain has reached the point where players are being forced into slowing the ramp. Capacity is being delayed.

SanDisk finally announced its decision to slow capacity adds in the Q2 cc:

“For SanDisk, we are implementing the following decisive actions to curtail our rate of supply growth until market conditions improve markedly:

First. We will push out the start of the next major phase of Fab 4 wafer output increase to no sooner than April of 2009. This will moderate the growth in our captive capacity starting in early 2009.

Second. We will put on hold our investment decision for Fab 5 until market conditions improve.

Third. Over the next 1-3 years we plan to increase our non-captive purchases of Flash memory from currently close to zero to 15%-20%. This will give us more flexibility to better manage our inventory.”

SanDisk is already committed to a Fab 4 ramp to 110,000 wafers a month. There isn’t much to be done about that. The equipment is either already installed or being installed and the employees are there.  This is why SanDisk’s efforts to slow capacity will only start early in 2009. Fab 4 is a very large structure that is modular. SanDisk is delaying additional capacity in 2009 for equipment that has not yet been ordered.

Fab 5 will be an interesting story to watch. SanDisk and Toshiba have already shown creativity in Fab 5 funding arrangements. I expect even more ahead given the size of the prize.

In the slide below also from the 2008 Flash Memory Summit Keynote, SanDisk notes that a huge rise in Flash demand [driven by SSDs] is expected by 2012/2013. Such demand, if 10% of the HDD market, will require 10 new Flash Mega-fabs.

If this is how the next few years come to pass, planning for Fab 5 will look visionary, even if on-hold, as of today.

One of the big challenges a NAND producer faces is dealing with the seasonality of demand. The year starts with a relatively soft first quarter ramping through a very strong fourth quarter. But then its back to a soft first quarter. A new Fab, such as SanDisk’s Fab 4, isn’t seasonal. Production just goes up.

As a result, Q1 typically is over-supplied, whereas Q4 comes up short. Were it so simple, the excess supply of Q1 could just be carried until it was consumed in Q4. Unfortunately, relentless ASP decline makes carrying such inventory prohibitive.

SanDisk’s proposed solution is to dial back production and then purchase the difference rather than carry inventory. Makes a lot of sense. It will be interesting to see where the NAND come from. When all is said and done would not be surprised if SanDisk and Toshiba work something creative out on this one too.

S-Curve: Memory Scaling Limits

Memory scaling limits are now looming in the NAND business. Its been business as usual from 210nm to 32nm and from 1 bit/cell to 3 bits/cell. It looks like 2xnm is doable as is 4 bits/cell (at least for SNDK/Toshiba), but that might be it and hitting these targets will take more time than usual.

The slide below, also from the 2008 Flash Memory Summit Keynote, shows the issues on an S-curve.

Below is my version of what I think SanDisk was trying to say:

NAND has had a great run but will likely be eclipsed in the next decade. 3D R/W memory looks poised as a promising candidate to continue Moore’s law.

Curiously at the very time capacity adds are slowing, huge demand is looming around the bend.  The pace of NAND scaling itself is slowing at the very time billions of dollars will soon be pouring into Mega-fabs producing what might be a technology reaching its limits.

If SanDisk has a post-NAND winner in 3D R/W, it promises to be not only timely, but most lucrative.

[It appears that this site is now adding links that I cannot control. The link that follows is such an example]