Intel has a NAND problem. Its IMFT NAND venture with Micron hasn’t lived up to expectations. IMFT has neither the scale nor cost structure to compete in these dark times.
The struggle for NAND market share is on. Stakes are high and the going is tough. Samsung is dug in at #1. Toshiba/SanDisk are making their move at #2- heads barely above water. Hynix is reeling at #3. IMFT is down a notch from there at #4.
Eli put it well on analyst day: It is tough to be number 4 in a rough neighborhood
Although NAND chips account for only a fraction of Intel’s business, NAND chips are having a big impact on Intel’s bottom line- and not in a good way. Intel’s NAND business is doing so poorly financially that Intel was recently forced to warn that the company’s Q1’s gross margins would be down a couple of percentage points due to NAND alone.
This is not the kind of drama that Intel wants or needs as its stock price bumps along at multi-year lows.
Intel’s CEO Paul Otellini has vowed to fix the problem. His solution- SSDs:
“We’re going to move aggressively into solid state devices and out of the commodity market.”
What is Mr. Otellini thinking? SSDs look really really tough. Hardly a high margin market segment. Competition is going to be fierce, with margins suffering accordingly.
Sure, SSDs are going to be big. This is an open secret. Samsung, Toshiba and SanDisk have been getting ready for this one for years. 2009 appears to be the year when their best efforts will be revealed.
Intel appears to believe it has a technical edge over these NAND heavyweights, that will set its SSDs above and beyond the competition. Time will tell. Personally, am not convinced.
That said, Intel appears ready to roll out its product line in the next few months and the specs look pretty good. A friend passed along an Intel marketing presentation on its NAND/ SSD strategy containing lots of interesting stuff that hasn’t hit the mainstream media yet.
Intel plans to “rapidly achieve and maintain technology leadership; Ramp capacity to achieve scale and efficiency; and Deliver value to Intel Computing Platforms via innovative NAND SSD and Caching solutions.” Intel plans to “ waterfall these SSD solutions into embedded applications.”
In very general terms, Intel divides the SSD market into three segments: Removable Drives such as USB/ SD and CF cards which have “lowest cost, lower density, MLC, limited reliability testing” and are “disposable”; Value Solid State Drives which offer greater “reliability over removable, strong performance, low power”, and “higher density”; and Performance Solid State Drives which offer “ultimate performance” and “assured reliability and compute validation.”
Removable Drives are deemed the lowest in cost and performance. Value Solid State Drives are next up the food chain. Performance SSDs are considered to have both best performance and highest cost. Although Intel appears to be planning on offering at least some products in each segment, clearly Intel is most excited about Performance SSDs, or SSDs as HDD replacements.
Intel High Performance SSDs- Code Name “Ephraim”
Intel SATA SSDs are scheduled to go into production in Q3 2008 using 50nm NAND in both 1.8” & 2.5 form factors. These High Performance SSDs purportedly will be available using both SLC and MLC. SLC SSDs will be available in capacities up to 64GB and MLC SSDs will reach 160 GB.
Intel SLC SSDs are advertised as having 250/130 MB/s sustained R/W throughput. Intel MLC SSDs purportedly will have sustained R/W of 250/50 MB/s. Power for both Intel SLC and MLC SSDs is rated as 1.7W in active and 0.095W in standby.
Intel brags on its drives’ performances vs both a Samsung 32GB SLC SSD and a SanDisk 32 GB SLC SSD. Feels like marketing hype- comparing products not yet in production to products already in the marketplace. We’ll have a better sense of how Intel stacks up towards the end of this year when SanDisk delivers its next-gen 43nm MLC SSDs.
For perspective, current Toshiba MLC SSDs (most likely 56nm) are already rated at 100/40 MB/s R/W. And its worth keeping in mind that independent third party testing have yet to run all these drives through their paces.
Last, but not least, Intel makes no mention of SSD price points in its presentation. Price is going to the huge deal once performance points are reached. SanDisk, Toshiba and Samsung will be happy with SSD product GMs @ 35%. For gaining market share, 20 to 25% GMs will probably suffice for these companies. I have my doubts whether Intel shareholders used to the 55%± for GMs will have the stomach to for this SSD battle ahead.
Intel’s SSD roadmap for 2009 shows <40nm NAND for 2009.
IMFT’s transition to 3Xnm promises to be an interesting story to watch. Their strategy appears to be to jump a generation (40nm) entirely and move from 50nm to 3Xnm. If IMFT can pull this one off smoothly and profitably, they will justly deserve a tip of the hat. If not, trouble looms. Personally I don’t find it auspicious that production from their Singapore state-of-the-art fab was recently pushed back.
SanDisk/Toshiba, probably the current leader in process geometry technology, is taking what appears to be the safer path. From 56nm San-shiba is moving to 43nm for 2008/2009 and then on to 3Xnm for 2009/2010.
My guess is that a key to these technology transitions will be lithography and related equipment. At analyst day, Eli took time to explain that SanDisk/Toshiba made the move to immersion scanners at 56nm while the competition, including IMFT, has continued to use dry scanners. In Eli’s mind, SanDisk/Toshiba has the advantage for the next leg down to 4Xnm, in that they already understand this proven technology.
Moving below 4Xnm appears to require next generation lithographic technology. Double-patterning appears to be the consensus pick. Extreme ultravoilet (EUV) lithography appears promising. This next-gen equipment is both expensive and unproven and there-in appears to be the risk for IMFT.
SSD performance appears related to both system and chip expertise. I suspect a subplot will emerge as ONFI competes with SanDisk/Toshiba’s All-Bit-Line Architecture. I’m going to leave that for another post.
IP for SSDs is another related angle best left for later. I have posted in the past about Solid State Storage Solutions, LLC. FWIW, for those following this obscure angle, SSSS LLC looks to be a Renesas IP angle that SanDisk has acquired. (Thanks DanR)
Another wild card is Rambus which is clearly working with at least one (big) flash company. Given recent litigious events, it seems a safe bet that Micron, Samsung and Hynix can be eliminated. That leaves Intel, and SanDisk/Toshiba. Intel, partnered with Micron, would seem the less likely.
Posted by savolainen 
