The Disconnect

On August 12, 2009, Eli delivered a keynote address at the Flash Memory Summit in Santa Clara, California.

I wasn’t at Eli’s presentation, but SanDisk has posted the slides. Many are the same as those that Eli used the day before at his Oppenheimer Conference Presentation (OPCO).

Much of the same material appears to have been covered.

One of Eli’s main points at the Flash Memory Summit was that over the next three years, the NAND business is not going to play out along the lines currently modeled by analysts.

There is a disconnect.


Analysts project huge growth in NAND demand in 2010-2013. To keep up with this demand will take incremental Capex of about $25B to 30B in new NAND capacity. Analysts project cumulative incremental NAND revenues of only about $20B in 2010-2013.

Analysts project NAND price reductions that are not consistent with slower cost reductions due to slowing NAND technology transitions beyond 2010.

EE Times picked up the story with:

“The NAND industry is at the ”crossroads,” as there is a ”disconnect” between future capacity requirements and demand, warned the top executive at SanDisk Corp.”

Here is a link to the full article.

Here we are two months later and there does seem to be a disconnect.  More demand than supply. Not enough capacity. Predictably NAND prices continue to be strong.

Here is a slide from Sanjay’s Deutsche Bank (DB) presentation. Eli used an older version of the same slide at the Flash Memory Summit. Currently prices are up from this slide. 8 October DRAM eXchange 16 Gb average contract price= $4.98.

Such NAND pricing strength is unprecedented.


Eli’s OPCO comments on this same slide:

“You can see the positive impact for us of pricing rising. What I’m showing you here is every quarter for the last five years, 4 1/2 years. The ASP declines that SanDisk experienced on a quarterly basis and the red line is on an annual basis.

And you can see, its always negative, every quarter price reductions coming in, supported to a large extent by cost reductions, but not completely.

In 2008 you can see that price reductions are really cumulatively adding up to 62%. Deprived us of all profitability. Things started improving, though it is difficult to see here in the first quarter of 2009 and definitely on the upswing in Q2 of 2009.

So the trend is definitely moving in the right direction.

Of course the flip side of higher prices is that demand in terms of MBs/unit and even the number of units slows down. We know that elasticity works both ways. But we are very focused on profitability. This is our #1 concern right now.”

When SanDisk reports Q3 earnings on October 20, we’ll find out whether higher NAND prices have dampened demand. I doubt it, given that Sanjay noted at DB that bit demand growth would be upwards of 100% for 2009. Basically this means that SanDisk is selling all their NAND production and burning through inventory.

The slide below was used at both the Flash Summit and OPCO. SanDisk left off 2010e. Nonetheless, from his comments, Eli appears to think 2010 will look like 2009, with supply bit growth of 50%±.


Eli’s commentary at OPCO:

“One of the most important developments that the industry has gone through in the last 6 months is a significant deceleration in supply and that has occurred really in the first and second quarters through production cutbacks. Which heretofore, have been unheard of.

As well as delay or defer, basically nobody is talking about new wafer fabs for NAND. So I think things, have in Q4 of last year, have become very very difficult in the industry, and there was really no other direction to go, but to cutback.”

In OPCO Q&A Eli elaborated, pointing out that without adding new wafer capacity, bit growth won’t go above 50% for any NAND supplier:

“ We have said that for this year we expect our total bit growth, 2009, to be under 50%. And that is through technology transitions. 43nm and going to three bits per cell. The next technology transition will be 32nm both two and three bits per cell.

Typically in a technology transition like that you can expect between 30 and 50% more bits. You cannot expect 100%. You have to put in place new capacity for new wafers if you want to go above 50%. That’s true not just for SanDisk, its true for any other supplier.”

This last week the DRAM eXchange released their estimates for NAND demand and supply for 2010. For Q1 and Q2 2010 supply is expected to slightly exceed demand, then in Q3 and Q4 demand is forecast to exceed supply. Bottom line- Shortage in 2010.

See chart below.


I kept looking at the chart and something seemed wrong. Finally it dawned on me. Estimated supply growth looked remarkably strong. Given Eli’s comments I was expecting 50%± since industry wafer adds are expected to be minimal at best.

I went back to the commentary and sure enough the DRAM eXchange is expecting NAND supply bit growth of 79%.

Sure seems like a disconnect somewhere.

NAND supply could get really tight next year. No wonder Apple is locking in all they can.


5 Responses to The Disconnect

  1. Jexzz says:

    Savo – Hi, remember me? I posted on Yahoo a few years back. One question: Eli said technology advances would bring bit growth to 50%. True. I see the disconnect there. But couldn’t he have been leaving out other ways of bringing about bit growth, like opening up new lines – not new fabs – just new lines that had been idle. Or increasing to 12″ wafers maybe. These might contribute to higher than 50% bit growth, though maybe not as high as 79%.

    Also, here Eli suggests only 50% bit growth for 2009, yet in a recent conference Sanjay already stated 100% plus growth for 2009 – the disconnect is where?

    What do you think? Hope all well, and enjoy your articles.

  2. savolainen says:


    How are you doing? Its always nice to see familiar names.

    As I see it Eli made a clear distinction between technology transitions and new wafer capacity. There wasn’t any grey in between.

    Technology transitions includes moving to smaller geometries such as 43nm to 32nm and includes moving to more bits per cell. X2 to X3 and X4. Eli said that the best that can be gained (including both) is between 30 and 50% more bits without adding new capacity for new wafers.

    New wafer capacity is new wafer capacity whether from new lines in existing fabs or new fabs. No one is as far as I’m aware is adding new lines for wafer adds or even talking publicly about doing so. Certainly no one is talking new fabs.

    Clearly those (like SNDK/Toshiba) who have the space in existing fabs to add wafer capacity are at an advantage compared to those (like MU/INTC) whose fab space is maxed out.

    As far as I’m aware all idle lines are now back in production. The transitions to 12” wafers are pretty well done too.

    Sanjay’s 100% growth comments were bit demand growth. In other words, he was referring to the other side of the equation- Demand. To meet this unexpected demand SanDisk is planning to ship all of its production bits (<50% bit growth) as well as inventory (the other 50+%).

    I thought the DRAM eXchange comments on 2010 supply were very revealing:

    “According to DRAMeXchange, 2010 NAND Flash total wafer output is expected to mildly exceed the amount in 2009. 2010 CAPEX will be primary applied for technology migration. We expect mass production for 2Xnm technology products will be scheduled at 2H10 since most vendors just adjust up the portion in 3Xnm technology in 4Q09. Also, some vendors just formally announced the mass production plan for 3-bit/cell MLC products in 4Q09 and we expect that the portion will be apparently adjusted up in 2H10 since it takes time to tune the yield and promote the products. Therefore, according to DRAMeXchange, 2010 NAND Flash supply bit growth will grow 79% in YoY basis to 10,759M GB.”

    Basically they agree with Eli in that wafer starts are not expected to increase significantly. They disagree with Eli in that they think technology transitions (geometry shrink and increased bits/cell) will add up to 79% growth (vs Eli’s 50%).

    Time will tell whether that is a disconnect or not. Big difference though.

    DRAMeXchange is forecasting 2010 worldwide NAND Flash demand bit growth to be 81% yoy to 10,986M GB. Unless we get an economic crash, this seems about right. Mobile is driving demand now. On top of that 2010 looks like the year of the tablet. Those are going soak up a lot of NAND.


    • savolainen says:


      A few more thoughts.

      When I started this “Disconnect” post, my intention was to work through the big picture including capex. It became clear very quickly that this was going to be too big a project for one Sunday.

      Maybe I’ll come back to it at some point.

      Analysts seem to be modeling the NAND business based on the last few years, where ASP/ GB dropped faster than costs. The result was that NAND markets were in essence subsidized by the NAND producers.

      Kind of like cheap oil, cheap, cheap NAND was a blessing for many, but not for the suppliers. I believe Eli is saying that these days are over- simply because the massive capex outlays needed for new fabs currently can’t be justified on the business models of the past few years.

      Also limits of physics are coming into play.

      It will be interesting to watch how the next few years play out. Hopefully it will be profitable and profitable for SNDK investors. I think SNDK has a real shot at something special.

      That said, lots of pieces have to fall into place.

      It seems pretty obvious that if capex stays low and demand stays high, NAND pricing should continue to be strong. I think what we are seeing now could be a preview of 2010 at least.

      I am guessing that the DRAM eXchange along with many others, is looking at demand and assuming that (somewhat magically) supply will keep up. After all it has in recent memory.

      Supply/demand is clearly the name of the game. Demand outstripping supply will mean strong pricing. Strong pricing will mean good GMs.

      A significant Flash Summit slide I never got to is #10 which shows a line at 30+ product GMs (non GAAP). Above this line “ROI become attractive”

      Eli from OPCO:

      “ On the supply side there is no question that the supply is limited to what’s in place right now. And the profitability is still not there to warrant going out on a capex spree and building the next fab. I don’t think that that is in the cards anytime in the next several quarters.

      I can tell you for ourselves we certainly don’t feel that today the ROI is there to build the next fab.

        It will take several good quarters of improved corporate profitability beyond what it is today

      before we would be even thinking about that.”

      It appears that sometime in 2010, if product GMs can at least approach the 30s for several quarters, capex $$ could start flowing again.

      Then its going to be at least 2011 before significant new wafer capacity will be on line. I suspect this is the same timeframe envisioned by other NAND suppliers.

      The other factor playing into this is the end of NAND as we know it. Interesting times.


  3. jexzz says:


    Thanks for your comments and clarifications. I see what you mean. Especially when you highlight out these “dualing” statements. Yes, a mighty disconnect. I would tend to believe Eli over DRAM Exchange, but that’s just me…. And the stock price isn’t reflecting much of my belief at the present moment 😦 ….but that’s what we’re used to, n’est pas? Have a great weekend, and glad to be keeping in touch.

    All best,

  4. jexzz says:

    Very interesting, by the way, about the GM needed for “to even think about” capex expansion…. I see it now – in other words, AAPL et al better start financing the suppliers’ capex, or they are up a creek w/o a paddle. Cheers.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: