Q4 2008 Wrap

Q4 was awful. Seemed to match expectations though.

Anticipating this, SanDisk seems to have decided to get as much bad news out of the way as possible. Hard to fault such a strategy.

The wrinkle seems to be the announcement of the shelf registration. Initial panic seemed overdone and has proven exactly that.

Before news of the shelf (conference call), on 2 February  SanDisk closed a shade over $11. The next day price at the close was $8 and change. A week later SanDisk was once again pushing $11.

Updates on x3 and especially x4 from the International Solid State Circuits Conference (ISSCC) probably helped restore confidence. Worked for me.

My tentative plan is put up some analyst excerpts next week. So I’m going to skip the numbers stuff today and focus instead on the shelf and x4.

The New Universal Shelf

Judy announced the shelf on the conference call in a brief dry statement- something to the effect that it might be nice for SanDisk to have a “liquidity cushion.” Hardly reassuring.

The problem for us existing shareholders is, of course, dilution. At the current share price, SanDisk’s target of $300 million to $500 million translates to dilution of 12% to 20%. Not a pretty picture.

But as is usually the case with SanDisk, there are many angles to be considered.

First and foremost is timing. While it may have felt imminent, SanDisk did not say they were going forward with such equity financing next week, next month or even this year.

This last week at the 02.11.09 Thomas Weisel Conference, Judy made an effort to put the shelf registration in a bigger, less-threatening context:

“We also announced last week on our earning release that we have filed a new universal shelf. Our previous shelf was due to expire in May of 2009. We filed a new three year shelf. We have not made any plans at this point, definitive plans, but we wanted to have that in place. We are planning conservatively, not knowing the duration of the down cycle to try to insure that we have the right path in place should we decide that we want to add a liquidity cushion to the balance sheet. If we do choose to do something, I’ll tell you that it would most likely be in some form of equity as opposed to debt, because we have $1.2B of convertible debt on our balance sheet and we also have large outstanding off-balance sheet lease obligations.”

The shelf doesn’t seem to be driven by a near-term cash crunch. SanDisk has enough cash for 2009. SanDisk said as much in the conference call.

Personally, I don’t think SanDisk is going to sell shares at these prices- certainly not in the near term.

While it is all fine and well to talk about a liquidity cushion, I wouldn’t be at all surprised if  SanDisk also intended to send Samsung a message.

With the shelf in place, SanDisk can negotiate harder with Samsung for the renewal of the patent-license agreement.  It shows Samsung that SanDisk is going to do whatever it takes to get a fair deal.

It’s a way of telling them that SanDisk can raise the money needed to survive long enough to stick it to Samsung in court- if Samsung is considering not signing by August.

SanDisk’s existing shelf is due to expire in May 2009 and as Judy said, it may be prudent to raise money at some point or at least have the paperwork in place.

And, realistically, if Samsung doesn’t resign, SanDisk will need the extra cash.

Then there is also the issue of SanDisk’s  convertible debt that is selling at a deep discount- 47%. This next stretch would be a good time for SanDisk to cleanup/ strengthen their balance sheet by retiring debt at roughly 50 cents on the dollar.

SanDisk may have enough cash to do this without raising money, but it is good to have options. Especially given current economic uncertainties.



Eli and Sanjay played it cool on the conference call, but when details of x3 and x4 were revealed this last week at the the 2009 International Solid State Circuits Conference (ISSCC), they did not disappoint.

x4, SanDisk’s 4-bits-per-cell memory, looks really good. To be honest, better than I expected. Probably far better than the competition expected.

It was widely known that the chip would be 64-gigabit in a single die, making it the world’s largest capacity flash memory chip. The big unknown was whether it would measure up on performance. Hints from November didn’t look promising.

An EE Times teaser article said x4 write speed would only be up to 5.6 Mbit/second which translates to .7 MB/s (less than 1 MB/s). In other words, it looked like x4 would be really s…l…o…w compared to run-of-the-mill generic 2 bits per cell (x2) MLC which clocks in at around 10 MB/s.

SanDisk’s x4 press release from 10 February boasts of 7.8 MB/s x4 write performance- about a 1000% better than expected. I felt like standing up and cheering. In fact I did.

What happened between November and February to boost x4’s performance?  Certainly nothing design-wise. I suppose EE Times could have screwed up with the numbers. Somehow I doubt it.

A more likely explanation is that SanDisk decided to set up the competition for a game of whiplash. Set them up with expected sub-par specs, and then hit them with their worst fears.

I wouldn’t be at all surprised if this was just another gambit in the Samsung re-sign saga. Intel couldn’t be too thrilled either. Neither seems to have an answer to x4.

An interesting angle to watch will be how SanDisk chooses to play its x4 controller card. The x4 controller is really special. There is no substitute. An x4 chip without its matching controller is worthless.

Way back, when msystems owned x4, Dov Moran msystems’ CEO was planning on keeping partners/licensees honest by controlling the controllers. It will be interesting to see how SanDisk proceeds.

There are lots of options for SanDisk- and options are good.

I suppose I should start referring to x4 as X4. This seems to be the SanDisk way. Old habits are hard to break, though.

Adiós MegaSIM

Back in 2007 I wrote a post titled MegaSIM or Bust.

Bust it is.

In the Q4 conference call SanDisk announced that it has stopped development of MegaSIM cards.

Sometimes things just don’t work out. In tough times like these, best to move on when prospects aren’t promising.

I have put up a transcription of the Q4 conference call. Here is a link. The transcription is also accessible from the “Pages” box to the left.

Apparently China is blocking my blog. Seems bizarre, but stranger things have happened. Pretty amusing in any case.


14 Responses to Q4 2008 Wrap

  1. Pam Smith says:

    Savo, with reference to the Write speed for the 64Gb x4 NAND at 43nm, I beg to differ with your explanations.

    The EETimes article by Rick Merritt was obviously a mistake as Rick, like many others before him, had used bits and bytes interchangeably, either carelessly or without a proper understanding of the difference between a bit and a byte! In Rick’s case, I would assume it was the former. I had read the ISSCC 2009 Advance Program before Rick’s article appeared in EETimes and the abstract of the paper that was presented last week had clearly rated the Write speed at 5.6MB/s for their upcoming 64Gb x4 at 43nm!

    As for the Sandisk’s press release, I feel someone forgot to proofread the press release before it hit the wires! I think, the 32Gb x3 NAND at 32nm should have a Write speed of 7.8MB/s and this would be consistent with their previous x3 NAND that was announced at ISSCC 2008, the 16Gb x3 at 56nm with a Write speed of 8MB/s which just won the Lewis Winner Outsanding Paper Award last week! So, I don’t think Sandisk was trying to setup the competition for a game of whiplash, but an amusing thought anyways!

  2. savolainen says:

    Greetings Pam,

    Yep you are probably right. Carnival of errors does seem more likely than clever coherence.

    Same effect though. 😉

    I’ll contact SanDisk about the X4 speed and post what I find out. I’m still pulling for 7.8 MB/s, but nothing wrong with 5.6 MB/s if that’s what we have to live with.

    Before posting, I suspected something along the lines that you outline, so I went back and looked carefully through the SanDisk PR on X3 at 32nm looking specifically for speed, but I didn’t see it. Found that curious. Oh well.

    We still have the issue of X4 reliability (somewhat) outstanding. Found it reassuring to hear directly from Judy that x4 is suitable (at this time) for pretty much all SanDisk products other than SSDs. From Thomas Weisel:

    [Question] “I’m intrigued by the 3 bits per cell and 4 bits per cell driving up capacity per cell and driving down price. Can you say more about [pause] would that would fit into an SSD for example? or is the reliability not there for SSDs? If not what is the target market for these products?

    [Judy] Well as I indicated this new SSD called the G3 that we just announced is now going to be produced on MLC which is 2 bits per cell as opposed to single level cell.

    So we are just moving from SLC to 2 bits per cell for SSDs. I don’t expect SSDs will go to X3 anytime soon. However X3 and X4, 3 bits per cell and 4 bits per cell, really can be used in pretty much all of our other products.

    The products that you would expect them not to be used in are the very high performance, high capacity SSDs, maybe a few other very high performance products, but for the most part we are actually shipping many of our products on X3. We have about 15% of our supply, our output, in Q4 was X3, three bits per cell.”

    BTW for anyone out there who is interested, the Weisel presentation is still available on the SanDisk site. It’s just not easy to find. One would think that it would be in “Presentations and Audio Archives” but no, its under “Past Events” accessed through the investment calendar.


  3. Pam Smith says:

    Just a few years ago, NAND MLC (2 b/c) had Write speeds lower than 5.5MB/s, so 5.6MB/s and 7.8MB/s for x4 and x3 NAND respectively, are both decent Write speeds for most applications (except maybe high end cards and SSDs.) x3 in particular with 7.8MB/s Write speed, comes fairy close to 10MB/s Write speed of 56nm x2 MLC nand from Sandisk/Toshiba! Thanks to ABL, x3 NAND is working at Write speeds close to that of x2 NAND without ABL!

    More important question in my mind is the endurance of these chips. How many Write cycles can one have with x3 and x4 based NAND before the cells become unusable?

  4. Poofypuppy says:

    Hi Savo,

    Thanks as always, lots of interesting stuff you covered in this (and the prior) post.

    Regarding StartKey, it makes sense that Microsoft would want to develop technology to enable more “on the go” computing. Running a Windows OS from a flash key would be a huge step forward, although Microsoft still has to contend with the threat and perhaps dominance of Google’s Software as a Service (SaaS) and cloud computing ambitions. I think Microsoft would have to promote StartKey (against Google) for its data privacy advantages (i.e. users physically keep their data with them vs. leaving it “out there” with Google to safeguard).

    Regarding the X4 controller as the required secret ingredient to make it all work, what are your thoughts on who is making the controller (and where)? Is SanDisk producing it at the same NAND fabs with Toshiba, or doing it at/with Tower Semiconductor (Israel), or someplace entirely different (e.g. Taiwan?)? Given how critical this controller is, any sense of whether there will be any supply contraints or single-source production risks?

    Regarding the shelf registration, if SanDisk is going to issue more shares to bolster its cash position, I really wish (but doubt) that SanDisk would sell shares to some strategic partners rather than to the general public. Something along the lines of selling 3%-5% stakes to (2 or 3 out of the following companies) Seagate, Intel, Toshiba and Microsoft. Obviously, SanDisk and Eli wouldn’t/shouldn’t sell a 15% stake to a single company lest that company acquire undue influence over SanDisk. BUT… selling 3%-5% to Seagate (and establishing a NAND or SSD supply agreement with them) and also 3%-5% to Intel (along with licensing agreements for X3 and X4 and ensuring that Micron is not included in the licensing rights) would bolster SNDK’s cash position, boost investor confidence, and also put Samsung on notice that it either needs to renew its licensing agreement or get locked out of the NAND X-files club.

    Regarding the death of MegaSIM, looks like it ended up donating most of its organs to the Service Delivery Card announced today. I wonder if it will be successful, but in any event am very glad to see that the whole MegaSIM effort wasn’t just time/effort/money that was flushed down the drain (at least not yet).

    Really wanted to replace the HDD on one (or more) of my computers with an SSD. Seeing as how the SanDisk G3 SSD isn’t available yet, I’ve ordered a Patriot 32GB SSD to try out. Not sure who makes their NAND (Toshiba or Samsung or ??), but am excited to see how it goes.

    Regarding China blocking your blog, the issue is very clear. Since SanDisk has operations in China, your posts on this blog now mean that you are disclosing state secrets, which cannot be permitted. 😉 LOL


  5. savolainen says:

    Greetings Pam,

    I heard back from SanDisk. Good news and an explanation.

    It turns out that the 5.6 MB/sec as published for the X4 chip in the ISSCC 2009 Advance Program was correct- at that time. These results were from early testing. When paper was due, this was the X4 speed.

    Since then, the X4 chip has been “optimized for higher speed” resulting in substantial improvement. The SanDisk X4 press release is correct- today. We are up to 7.8MB/s- Now.

    X4 must be very interesting. Some “Optimization” and the speed jumps 40%.


  6. Pam Smith says:

    The Write speed optimization is just wonderful news!! Sandisk/Toshiba can now deliver twice the capacity at same price, assuming similar yields on x4, at almost the same Write speeds as provided by x2 chips from competitors.

    This makes me wonder what is the correct Write speed for their 32nm 32Gb x3? Their last x3, 16Gb at 56nm was rated at 8MB/s. IMFT’s 34nm 32Gb is at 9MB/s.

  7. Poofypuppy says:

    Came across this article discussing X3 and X4 speeds. Some of the performance data seems to have been confirmed, while other numbers appear to be the author’s extrapolations…

    Sandisk to commercialize X3 and X4 MLC Flash this year

  8. savolainen says:

    Hi Poofy,

    I agree with your comments about StartKey and Microsoft- Its not cloud computing, but a big step in that direction.

    It would be nice to get at least confirmation that the project is still alive. I suspect it is, but still even a flicker of news would be reassuring.

    Whether slotMusic will have a tie-in with MSFT remains to be seen, but as I recall at CES this year, one of the slotMusic demos was on a Windows Mobile device. This is the beginning of what could evolve into DVD replacement. Along with this could come software distribution which seems a natural for StartKey.

    And StartKey mobile could be another big angle. Whereas U3 was strictly USB drives, StartKey will also be targeting cards. The obvious play would be cards for mobile which could evolve into something of a back door to another side of the mobile market for MSFT.

    As far as where SanDisk X4 controllers will be fabricated. I don’t know. Probably at SanDisk’s regular trusted sources.

    Back at the 2006 msystems’ analyst day, Dov said that the controller, though very complicated, was basically a standard controller which could be manufactured in any fab that makes controllers. For a while I suspected TSEM would be manufacturing some, but now I’m not so sure. When Eli stepped down from the TSEM board, I took it as a negative signal. As you said, it probably makes sense to have several sources.

    The idea of selling shares to some strategic partners makes a lot of sense. I don’t see Seagate or Intel, but Toshiba and Microsoft make sense. I think ultimately Intel will back out of the NAND chip business. Margins are too low.

    Microsoft could well be interested if relations are good. There could be many things going on behind the scenes with SanDisk and Microsoft, but its just speculation now. MSFT is more interested in software that hardware. With consumer devices its the opposite for SanDisk.

    MegaSIM morphing into the service delivery card is interesting and promising.

    One of the great X3 and X4 angles is cards. Apparently eight X3 chips can fit in a MicroSD card. and eight X4 chips in an SDHC card. At upcoming capacities this will mean 32GB MicroSD and 64GB SDHC cards. Those without X3 and X4 will be out of luck for those capacities (I think).


  9. savolainen says:

    Hi Pam,

    The write speed for the X3 32nm is 7MB/s (per SNDK). This is the speed published in the ISSCC technical paper. As 32nm “is in the early development stages”, SanDisk expects the speed to improve.

    Curiously this means X4 @ 43nm is faster than X3 @ 32nm right now.


  10. Paul Rice says:

    Hello Savo

    Thanks as always for all the excellent work and insight provided

    I have been hoping the drop in flash prices would launch a new market namely content distribution and software application distribution

    of course piracy and/or multiple users have dogged both software developers and music and film – content providers

    The music industry is still searching for the good old days when piracy was a minor problem (comparatively speaking)

    The film industry watches as the sales of DVD’s decline while their catalog of film titles have been completely exploited and now produces much less revenue

    How do they insure revenue from movie distribution?

    It seems a flash device (not unlike slot music) could provide a more robust security methodology insuring content could not be copied and was tied to the computer – appliance processor serial number where it is installed – resides on

    Although it could be de-installed and used on a different appliance…
    titles and songs could be rearranged on a computer and then transported to another device
    but the content could not be transferred with out the original slot device

    The same would insure application revenue … apps costing upwards of $200 to $2000
    i.e. Adobe Photoshop .
    The delivery on flash would insure one authorized computer at a time

  11. savolainen says:

    Hi b9,

    I agree with you that flash will likely soon be used for both content distribution and software application distribution. It seems its only a matter of time now with flash prices where they are and trending ever down.

    Hopefully SanDisk will have figured out a way to profit and not just end up as the flash equivalent of a supplier of CDs and DVDs. One key would seem to be adding intelligence to the flash device- whether it be card, USB drive or whatever.

    I am really happy that slotMusic has moved away from DRM. I believe the original concept included DRM, but that is not where the business is headed. DRM was value added from the perspective of the music industry, but value lost from the perspective of the consumer.

    I am also somewhat more optimistic about the potential for slotMovies. Here I think the movie studios would be very very interested in both cooperating and helping to drive a flash-based content distribution system. Right now I don’t see DRM as a big problem for consumers, but it is huge plus for the film studios. The value added for the consumer might very well turn out the small size of the flash device.

    Mobile devices are coming into their own as platforms for watching movies. Cards seem just about perfect for delivering movies to mobile devices.

    Another angle (and value-added) is the (potential) death of the dedicated video player- likely this is the last generation of DVD players as we know them.


  12. hapa says:

    Greetings, Savo.

    Regarding 2X-nm node Toshiba said recently, ” …it is currently trying to determine whether the floating gate structure can still be applied or a new structure like a nitride trap (MONOS) will be required.”

    If FG runs out of gas and CT becomes reality, will the current Sandisk’s IP still be applicable to NAND with this new cell?

    Thanks, as always.

  13. savolainen says:

    Hi hapa,

    Yep the 2X-nm node will be an interesting story to watch. It would make an interesting post, though at this point, for the most part, it would be riddled with wild speculation.

    Nothing new there. 😉

    Aside from the IP question, there is the question of what Toshiba/SanDisk will do. I suspect that there is a slight difference of opinion between the two companies right now on how to proceed below 32nm.

    Toshiba has been dropping hints about Charge Trapping (CTF) NAND while SanDisk has said that Floating Gate (FG) NAND will likely be scalable down to 2Xnm.

    Regardless of which fork the companies go down, it is going to be an expensive transition and far slower than others. New expensive unproven advanced lithography equipment will be required. and then there are the not insignificant challenges that simply come with pushing the physical, electrical and reliability limitations of what these chips can do.

    My guess is that SanDisk/Toshiba won’t have 24 nm into significant production before 2011/2012. I don’t expect the competition to do any better.

    My guess is that SanDisk will be proven right and FG NAND will reach 2Xnm- though not without some grief. Error correction and so forth will become ever more important as the interference between adjacent floating gates increases. And of course such system solutions are SanDisk’s strength.

    I wouldn’t be at all surprised if some of the issues of 2Xnm NAND aren’t similar to the challenges that X4 has successfully dealt with.

    The big problem with CTF NAND is that it is an unknown unproven technology with its own risks- the biggest of which is how it will handle X3 and X4.

    For context’s sake it is worth remembering that only two years ago CTF NAND was being trumpeted by Samsung and Hynix as the sub-40nm solution to the significant difficulties of cell-to-cell interference. We haven’t seen it yet from either. Probably for good reason.

    Floating gate NAND has proven remarkably resilient.

    Back to your question about whether SanDisk’s IP is relevant to CTF NAND, I really have no idea, but would guess that if CTF NAND at 2Xnm is capable of X2 and above SanDisk’s system level IP would come into play.

    The next transition to the 1Xnm node might even be more interesting. 3D read/write, is the goal. Let’s hope it’s ready for prime time.

  14. hapa says:

    Thank you very much for a very thoughtful response.

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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: