Why does Samsung need SanDisk?

There’s been a lot of speculation and analysis on why Samsung would acquire SanDisk. Here’s my take.

1. Samsung needs the capacity. – Samsung has enough capacity, especially in this market.

2. Samsung is attracted by SanDisk’s retail reach. – Samsung has a strong retail brand and could have leveraged its own retail channels to push flash cards through its distribution network if it wanted to. Why do it now?

3. Samsung saves on royalty fees and the savings justify an acquisition. – Pay $3 billion (SNDK’s market cap) to save on $400 million a year? Sounds pretty good. >10% ROI. But the level of that royalty stream is not guaranteed and I don’t think Samsung would do a deal just to save on royalty payments.

4. Samsung wants the IP. – Sometime in the next few years, FG NAND flash is likely to hit a brick wall and an alternative technology will be required. CTF has its problems and one of the potential options is 3D memory (see http://forward-insights.blogspot.com/2008/06/sandisk-and-toshiba-to-jointly-develop.html). SanDisk appears to have some fundamental IP in that area. Sure, Samsung could negotiate an extension to their existing cross licensing agreement with SanDisk to include 3D memory but they don’t get the knowhow. The knowhow is what is going to allow the leaders to scale the brick wall.

Samsung is not known for big acquisitions and its last big acquisition of AST Research 20 years ago is not considered a success. And it’s hard to imagine SanDisk’s manufacturing and joint development partner, Toshiba standing idly by as Samsung takes ownership of SanDisk’s most prized assets.

Seagate to Purchase SanDisk? Definitely, Maybe Not

Rumors are flying about Seagate buying SanDisk or Intel's stake in Intel-Micron Flash Technologies (http://www.eetimes.com/news/latest/showArticle.jhtml?articleID=210004140). Does it make sense? Seagate does need access to low cost NAND flash to fuel its SSD ambitions. Here are some ways to think about it.

1. If Seagate were to purchase SanDisk or Intel's portion of IMFT, what are they going to do with all that capacity? SSDs are not going to be able to eat up all that capacity, at least not for the next couple of years. Does Seagate want to manufacture flash memory cards or supply to MP3 players? SanDisk's operating margin in 2007 was 8% and Seagate's 6.8% in FY08. However, if you strip out income from royalties and licensing, SanDisk's operating margins were -5%. Combining a primarily OEM HDD business with a low margin retail business could be challenging.

2. Seagate is primarily interested in enterprise SSDs which uses exclusively SLC NAND flash technology. Any SanDisk acquisition provides Segate with 150k 300mm wafers per month of MLC NAND technology. Spending $4 billion (market cap of SanDisk) to obtain MLC NAND flash technology for enterprise SSDs? (SanDisk and partner Toshiba currently do not manufacture any high density SLC NAND parts, although that could change in the near future.)

3. Seagate spends almost $1 billion in capex ($930m in FY08) p.a. in their HDD business. SanDisk in 2007 spent roughly $1.6 billion and Intel $1 billion in NAND flash. Looks like Seagate's capex would have to at least double if it wants to become a NAND flash manufacturer and it's balance sheet would be strained supporting such a high level of investment.

4. Intel is not ready to exit NAND flash, at least not in the near term. IMFT has typically been behind the technology leaders by 1-2 process generations. With the 34nm announcement, they are about to leap ahead. With the technology lead, IMFT will presumably have the lowest costs. In addition, Intel will be releasing a slew of competitive SSD offerings in the next months. If, with the lowest costs in the industry and a very strong product lineup, Intel still can't make money, then it will consider whether it makes sense to be in this business, but they're not going to quit before giving it their best shot.

Seagate needs access to low cost NAND flash and they don't need to obtain that access by getting into the retail flash card business or doubling their annual capex. They could achieve the same means by either investing a small equity stake in a NAND flash vendor or placing an upfront payment to secure NAND flash capacity at preferential pricing. Apple did something similar when it introduced its first flash-based iPOD. Such an arrangement also affords Seagate sourcing flexibility especially if those wild hockey stick projections of SSD shipments don't pan out.

Finally! A Simple Metric for Solid State Drive Endurance

We've all seen the impressive MTBF (mean time between failures) specifications of SSDs versus HDDs of one to two million hours versus 600 thousand hours in a notebook pc class application. However, unlike HDDs where a mechanical failure can make the entire HDD unusable, the main failure mechanism in a SSD relates to the memory cells becoming unusable. MTBF is a statistical calculation that unfortunately, does not capture the effect of write endurance.

The write endurance of a SSD is a function of the type of flash memory used (SLC, MLC), storage capacity, frequency of writes, data sizes of the writes and the amount of static data which relates to the wear leveling algorithm. System level endurance of SSDs in the industrial, enterprise and miliary space have ranged anywhere from one million to five million program/erase cycles. There are three main problems with this.

1. The endurance will depend on how the SSD is used in the application. As a result, SSD vendors can tweak the parameters for their own marketing purposes.

2. Not all SSD vendors use the same formula for calculating the endurance and the formulas can be overly complex.

3. A straight endurance metric is nebulous and difficult to grasp its implications, especially for OEMs who've never had to grapple with endurance issues in HDDs. For example: What's the effect on the lifetime of the drive?

SanDisk aims to solve this with the Longterm Data Endurance (LDE) metric. LDE is simply defined as the total amount of data writes allowed in the lifespan of the SSD. The metric is based on the Bapco write usage pattern for a typical business user and assumes the data is written equally over the lifetime of the drive and that data is retained for one year once the LDE specification is reached.

LDE allows OEMs a simple way to compare SSDs and determine, based on the applications usage patterns which drives are suitable for a particular application. For example, a drive with an 80TBW (teraByte write) LDE can support 20GB writes per day for 10 years (equivalent to 73TBW). For an application requiring support for only half the number of writes per day (10GB), a 40TBW rated drive would be sufficient.

The beauty of LDE is that it captures endurance in one single, understandable figure. A common metric is necessary to facilitate SSD adoption moving forward. Now comes the hard part: garnering support from other SSD vendors and OEMs.

I've uploaded a copy of Don Barnetson's presentation on LDE, "Solid State Drives: The MLC Challenge" on my website at http://www.forward-insights.com/.

Samsung Introduces Server-Grade NAND flash for SSDs

Samsung Electronics announced today that it collaborated with Sun Microsystems to develop a SLC NAND flash memory with five times the endurance of conventional NAND flash memory. This would put the "ultra-endurance server-grade" flash memory at 500k program/erase cycles.

To extend the endurance, Samsung likely tweaked the underlying cell process and/or modified the programming algorithm. The net result is that the improved endurance comes at the expense of reduced performance or retention. However, by targeting high transactional enterprise applications with this device, the degraded retention should not be an issue.

SanDisk and Toshiba to Jointly Develop 3D Memory

In a SEC filing today, SanDisk disclosed that it had signed a collaboration agreement with Toshiba for re-writeable 3D memory. Both companies will cross-license related IP to each other and SanDisk will receive licensing payments from Toshiba.

Investment for SanDisk-owned equipment which includes tooling for 3D R/W and 3D OTP memory is forecast to amount to $400 million in 2009 and $200 million in 2010.

The 3D R/W memory which consists of stacked vertical diode arrays shares most of the process modules and design architecture concepts with the 3D OTP memory. In the case of 3D OTP memory, four layers of stacked memory cells are in volume production at the 80nm node with 45nm currently under development.

A four level 3D R/W memory will have to at least catch up with NAND flash on process technology to be considered competitive with x4 NAND flash. This would put 3D R/W memory at least 3-4 years out. The other issue is whether an eight level stack is manufacturable at high yields. An eight-level stack was demonstrated by Matrix in 2003 on a much less advanced 0.25um geometry. Producing eight level memory stacks at leading edge technology is another matter.

Hynix Develops x3

Hynix announced it has developed a 32Gb 3-bit per cell NAND flash slated for production in October. Hynix stated that the 3-bit per cell technology would enable cost reductions of 30% over 2-bit per cell technology. The 32Gb chip is based on 48nm process technology.

Based on my estimates, the die size of this device should be over 200mm2, much larger than the 172mm2 of the recently announced 34nm 32Gb MLC NAND flash from Intel/Micron. It is unlikely the chip is using the All-Bitline architecture developed by SanDisk and Toshiba implying a fairly low program performance.

The 32Gb chip will probably be employed as a learning vehicle for applications enablement until a more competitive 41nm offering comes out next year.

Intel and Micron Leapfrog the Competition

Intel and Micron announced today that it will be sampling a 34nm 32Gb MLC NAND flash to customers in June with production slated for the second half of 2008. Owing to the aggressive gate half-pitch, immersion lithography with self-aligned double patterning employing spacers is most likely being used. Also expect changes in the bitline and wordline materials as well as a higher k interpoly dielectric in comparison to the 50nm generation.

At 172mm2, Intel-Micron's 32Gb product will be the only 32Gb monolithic MLC device capable of fitting in a TSOP package. If the ramp of IMFT's 50nm 16Gb device is any guide, we should expect to see volume in December or in early Q1/09. It's quite remarkable that Intel-Micron have managed to catch up and surpass the other NAND flash vendors on process technology in the short span of three years. IMFT achieved this milestone by skipping the 6xnm and 4xnm nodes. However, any cost advantage could be short-lived if IMFT fails to ramp up the technology smoothly and SanDisk/Toshiba ramps its 43nm 32Gb x3 in Q1/09 as planned.

Macronix: After the Qimonda Divorce

Just 3.5 months after signing the agreement, Macronix International and Qimonda AG announced that Qimonda terminated the Flash technology license and joint development agreement. Having lost Euro 482 million in the most recent quarter as well as its technology partner, Nanya Technologies to Micron Technology, Qimonda is in survival mode. By cutting back its flash technology development, Qimonda intends to re-deploy resources to focus on ensuring the success of its new buried wordline DRAM technology.

Macronix, on the other hand, will need to either develop flash technology internally or with another partner. With only $487 million cash on its balance sheet at the end of 2007, Macronix will not be able to fund technology development as well as invest in a leading edge 300mm fab. It will clearly require partners.

However, the number of potential partners is limited as all NAND flash vendors with the exception of Samsung are already aligned with others. An earlier agreement to collaborate with Powerchip Semiconductor on flash memory development and foundry services at fab 12M ended due to a fight between the two companies over control of the Macronix board at last year's shareholders' meeting.

One long-shot possibility is Spansion which is rumored to be developing a SONOS-based memory in a NAND architecture dubbed "ORNAND2". However, the ORNAND2 cell is based on Mirrorbit technology and is therefore different than the BE-SONOS technology championed by Macronix which is based on electron tunneling for programming. The fact that ORNAND2 is based on Mirrorbit technology means it is not really a true NAND. Other than the diverging technology strategies, lingering negative sentiment over Spansion's 2006 trademark infringement lawsuit against Macronix makes any collaboration unlikely.

Despite its collaboration with Numonyx, Hynix may be the most attractive partner. Hynix and Numonyx jointly develop product designs, however, technology development rests mainly with Hynix. Normally a technology follower, Hynix will no longer to be able to just copy the technology of others if it is to become a technology leader. BE-SONOS offers a viable scaling path for sub-40nm NAND and a combination of Macronix's strong development team and Hynix's low cost 300mm wafer manufacturing could make a potent team.

Seagate Fires First Volley Against SSD maker

Hard drive maker Seagate a lawsuit against solid state drive manufacturer STEC in the Northern District of California claiming STEC infringed four of Seagate’s patents related to how a SSD interfaces with computers. This is the first time a HDD manufacturer has sued a SSD maker.

It’s hard to see the financial motivation behind such a move. STEC’s enterprise SSD revenues were only $11 million in 2007, although it could reach 4-10x more this year depending on the ramp of the ZeusIOPS and Mach8 MLC product lines. With a market share of over 50% in enterprise HDDs, Seagate clearly sees a longer-term threat from the leading maker of enterprise SSDs.

However, a bigger motivation would be to send a signal to flash memory makers about the value of its intellectual property. It’s no secret that Seagate has been courting NAND flash vendors to secure NAND flash for a SSD it’s planning for the end of the year. A rumored JV with Micron fell apart last year and it is apparently in discussions with two of the three largest vendors. One of the main stumbling blocks is what Seagate could bring to any cooperation as all NAND flash vendors have ambitions to develop and market SSDs. This litigation could be a validation of the IP which Seagate can offer. Western Digital, which along with Seagate, was an early investor in SanDisk (then SunDisk) apparently holds some critical controller and wear leveling IP and could be next to enforce its IP.

SanDisk is taking no chances. In Q2/07, it set up Solid State Storage Solutions LLC with unknown partners that will license IP, presumably SSD-related IP. In July 2007, Solid State Storage Solutions LLC invested $42.5 million for the acquisition of intellectual property. It has apparently purchased relevant SSD and controller IP from Renesas Technology. This IP could prove effective in extracting royalties from flash memory card manufacturers and controller makers entering the SSD space, but it is questionable whether it is enough to counter any future claims by Seagate and WD.

Hynix Signs Another Deal

Hynix signed another deal, this time with Grandis to jointly develop spin-torque RAM. Unlike the current MRAM products marketed by Freescale, STT-RAM uses the angular momentum derived from the spin of the electrical current to alter the magnetic orientation of a free layer. By having the current directly pass through the magnetic tunnel unction, the fields required to switch a bit can be scaled down as process geometries shrink.

This is the third in a string of technology deals starting last October with Ovonyx on phase change memory and with Nanosys on nanocrystal memory. Hynix has traditionally been a technology follower, mimicking Samsung’s advances on the technology front. However, as conventional DRAM and flash memories encounter scaling challenges alternative approaches may be required. STT-RAM is one of them.