Will Seagate build a flash SSD foundry?
How big will Seagate's dash for flash splash be?
The ebullient Seagate CEO Bill Watkins has made it clear that Seagate thinks flash solid state drives (SSDs) are interesting and Seagate will sell them - but not yet. When Seagate does sell a Seagate-brand SSD how will it make it? Will there be a Seagate flash foundry?
A customer buys a hard disk drive (HDD) array from an OEM or channel source and the innards consist of a controller from LSI or someone similar and disk drives from Seagate or someone else similar. Seagate makes drives but not controllers.
A flash SSD array or array component will consist of a flash array controller and flash SSDs, which consist of flash chips and the SSD-level controller, roughly equivalent to the HDD integrated circuit board sold and supplied by Seagate as an intrinsic part of the hard drive.
So, logically, Seagate would sell to OEMs a flash SSD equivalent to a HDD unit; that is to say, a bunch of flash chips and control electronics. Will Seagate make the flash chips?
That is a $100m question, possibly even a $250m question. Flash foundries don't come cheap. Seagate could build its own fab, participate in a joint-venture with, hypothetically, a firm like SanDisk, or simply source its chips from an existing fab owner like STEC or Mtron.
The company isn't saying what it will do. Kevin O'Dwyer, Seagate's regional sales director for the northern Europe region, said: "We think it's (flash SSD) a real opportunity. It's a market we'll participate in. How we will build them hasn't trickled down yet. Bill may have some ideas."
What are the apparent pros and cons of each chip-building approach?
If Seagate sources chips from an existing foundry then it has to add value somewhere. Unless it has chips designed and built specifically and only for itself then it would have to add value in the SSD controller area.
Same question: will it build these or buy them in? If it buys in chips and buys in controllers then its added value is its brand and its distribution channel. The company could face the same kind of problems as Imation has as it has diversified beyond tape media manufacturing to things like USB thumb drive sales: a lot of competition and less strong branding.
There is a cost-saving if Seagate enters into a joint venture but there are also the complications of partnership and of someone else using the flash chips for products that could compete with Seagate. To my mind this is less desirable to Seagate than owing its own fab outright.
It could build one or it could buy one.
Building one needs considerable cojones, not to say considerable amounts of cash. Seagate understands, it really understands, the deep nuts and bolts aspects of building disk drives, but it knows little about the deep nuts and bolts of flash fabrication. Certainly it could learn, but the time involved in that could persuade Mr. Watkins and his team to think about buying an existing fab - because that could be productive from day one.
Recently Seagate has bought companies that are active in markets it wants to get into, with Maxtor the most obvious example. It has to be a reasonable possibility that Seagate will try to acquire an existing NAND flash chip manufacturer.
This will also require lots of money, or shares, but will get Seagate up and running with a flash SSD distribution channel as well. Say Seagate went out and bought somebody like STEC or Mtron - is this likely?
The back-of-an-envelope calculations above make this look attractive, more attractive than a JV. The sheer 'bigness' of it seems to be the sort of thing that would appeal to Seagate execs too.
If you're gonna make a splash in your dash for flash then make it a big one.
Copyright © 2008, Blocks & Files.com
Sponsored: IBM FlashSystem V9000 product guide