There are worse problems to have.
According to the Q2 conference call, SanDisk can sell every chip it can make and then some. And for tidy profits.
Demand is outstripping supply. But supply is limited. SanDisk is supply constrained.
Exactly what this means and in what time frame isn’t as obvious as it might seem at first take. Supply constrained means growth limited by supply. In this case the limiting supply factor itself is growing.
How much or how fast SanDisk’s supply is growing, is what this post is about.
Wafers vs Bits
There is a big difference between wafers and bits.
Projected growth rates for each are very different. Wafer supply growth for SanDisk is ho-hum over the near term. Bit supply adds the juice.
For 2010, wafer production for SanDisk will likely grow only around 10%. In other words, not much.
On the other hand, bit growth should be 70% to 80%. Big difference. Especially since revenues are driven by bits.
The graphic below illustrates Merrill Lynch’s estimate of SanDisk- NAND JV fab capacity from 2009 through 2012 in thousands (K) of Wafers per Month (WPM). The WPM’s shown are the combined share of Toshiba and SanDisk.
As shown Fab 3 was maxed out already in Q1 2009. It was running and continues to run at 150 K WPM. So far so good for ML, though that’s about it.
ML seems to have completely blown it for Fab 4. Q1 2009 through Q2 2010 is shown as 110 to 113 K WPM. During this time, Fab 4 has been running at closer to 160 K WPM which when combined with Fab 3’s 150 K WPM gives a total JV capacity of 310 K WPM.
By the middle of 2011 Fab 4 will be maxed out at 210 K WPM giving a total JV capacity of 360 K WPM.
As far as SNDK’s share of the today’s JV total capacity: after the restructuring in early 2009, the split between SNDK and Toshiba is about 40/60. SanDisk’s share has been about 125 K WPM and Toshiba’s 185 K WPM.
We know SanDisk’s share because at Investor Day 2010 (slide 61) SanDisk said that in 2009 its yearly wafer output was 1.5 million 300mm wafers. 1.5m/12= 125 K.
Today SanDisk and Toshiba are fitting out the last of the remaining space in Fab 4 and are going to split that capacity of 50 WPM 50/50. SanDisk and Toshiba expect to reach full capacity of Fab 4 by the start of production in Fab 5- which means Q3-ish 2011. About 1/2 of this additional shared production should be available by the end of 2010- SanDisk and Toshiba should each pick up another 12.5 WPM.
So by the end of 2010 SanDisk will have 137.5 WPM (+10% for the year) and by Q3 2011 should be up to 150 WPM.
And then Fab 5 kicks in.
SanDisk has said that in 2011, Fab 5 will account for no more than 10% 2011 wafer (production). SanDisk also has said that Fab 5 will only contribute single digits in terms of percentage of its overall supply in the second half.
I’m going to guesstimate this means another 15 K WPM by the end of 2011, bringing SanDisk’s total up to 165 K WPM.
If this is how it works out, by the end of 2011, SanDisk could have 20%± more wafers than at the end of 2010 or 32%± more than today.
SanDisk’s Lost Wafer Capacity
In early 2009 SanDisk restructured its flash manufacturing joint ventures with Toshiba. As part of the agreement, more than 20 percent of the joint ventures’ capacity was transferred to Toshiba. The total value to SanDisk was approximately $890 million.
This 20% of capacity translated to about 30 K WPM which SanDisk lost.
I suppose, if Toshiba were agreeable, SanDisk could buy back this capacity. I think this unlikely though, because first, no one has been talking about this possibility, and second, Toshiba reportedly already has its customers on allocation.
One Toshiba customer in particular would probably have to be squeezed, and Apple isn’t likely to be too happy about that.
NAND chips are produced from wafers. These days the standard wafer is 300mm in diameter. As the graphic below illustrates, lots of chips (the little squares in the grid) are produced from each wafer.
How many chips will fit in that 300mm circle depends on how small the grid. The smaller the better- if bits per square (or chip) are equal.
Both process node and the number of bits per cell are key for chip size. The smaller the node, and the more bits/cell- the smaller the chip can be. And, the more chips/bits can be produced from each wafer.
For example SNDK’s X3 32 Gb 32nm die size- 113 mm2. Intel’s 34nm 32 Gb MLC X2- 172 mm2.
Another example- SNDK’s 32nm, 32Gb X2 is 140 mm2 in size. SNDK’s 32nm 32Gb X3 is 113 mm2. X3 provides ~20% more Gigabytes/wafer which translates directly to 20% more profit where X3 can do the job of X2. And hence the importance of X3.
Yields too, come into play. The more chips are good per wafer, the bigger the bit supply.
In the recent Q2 conference call, SanDisk noted that its captive bit supply is growing at approximately 75% for 2010. 2011 looks like more of the same: “75 to 80% bit growth for 2011 over 2010.”
Given that wafer growth in 2011 looks to be double that of 2010 and 24 nanometer will in full ramp in 2011, I suspect that 80% bit growth will prove low.
Other Silicon Sources
As far as I know, no specifics have ever been released on the details of SanDisk’s non-captive supply agreements with Samsung and Hynix. It has been announced that both are legally obligated to set aside supply for SanDisk under certain conditions.
Right now SanDisk isn’t using any of this potential non-captive supply, but that’s about to change.
In the Q2 conference call, SanDisk revealed that it has placed orders for non-captive supply for both Q3 and Q4. Sanjay Mehrotra:
“Daniel, as you know, our demand has grown rapidly during the course of the year. In fact, we have raised the revenue estimate second time giving – compared to the start of the year now. So we are basically running as fast as we can in terms of really fulfilling the demand, and that includes maximizing our captive supply, but, yes, also tapping into the NAND non-captive sources. So we saw the demand increasing during the second quarter. We placed orders for the non-captive supply. We’ll be receiving it in third quarter. So in terms of shipments, the impact of non-captive in third quarter will be small, and in the fourth quarter, it will be somewhat larger. But as Judy said, overall, for the second half, still, it will be in single digit percentage.
So there are several factors that we look at in managing our non-captive business. We want to make sure that we are managing it prudently. And some of the considerations include margins, the mix of the product as well as basically the timing of the supply that’s available. So looking at all of that, our picture ahead is single digit percentage in the second half of non-captive utilization.”
From SanDisk SEC filings:
“The Company’s contracts with its other sources of silicon wafers generally require the Company to provide purchase order commitments based on nine month rolling forecasts. The purchase orders placed under these arrangements relating to the first three months of the nine month forecast are generally binding and cannot be canceled”
Non-captive doesn’t work for OEM or specialized/optimized uses like SSDs.
To my mind non-captive supply has the potential to be a Q4 buffer, and maybe suitable for picking up a few bucks. Hopefully SanDisk can profit from its non-captive supply agreements in these tight times.
I had hoped to get to free cash flow, Intel and the implications of Eli’s departure, but time’s running out so I’m going to call it a day here.
Personally, I’m not worried about SanDisk’s supply. Heck, captive bit growth looks like its going to be north of 75% through the end of 2011 with guaranteed additional non-captive supply waiting in the wings.
If current demand keeps pace, SanDisk should bank a pretty penny.