Thanks, Dave. Thanks to Jay also.
"Now don't overreact to this, but on the issue of MR improvement with KM we finally reached the point where I heard "similar" to inductive. If so this could be good news. Why do I say "could be"? Because Bramson said that additional testing is still being completed and that perhaps by the end of the quarter he could more fully address the issue. I would sum it up as positive, but not wildly so. Remember, somebody still has to buy this to make it a commercial success."
I think it was in the 1996 1st quarter 10Q that Ampex started disclosing the kind of numbers they were getting with KM+TFI. If I remember correctly, they started reporting 33-38% improvement in linear density (with 2 different kinds of proximity inductive heads) and they stated that they think they could get further improvements in track density and even more improvements by optimizing the read channel electronics (PRML integrated circuit + matching preamplifier). As you know, areal capacity is increased by improving track density (number of circular tracks in a 3.5" disk), linear density (number of bits in each track), and reducing inductance. Ampex subsequently reported capacity increases in the range of 20% with newer and improved inductive heads. I think it is fair to assume that Ampex is somewhere in the 20-40% capacity improvement range with KM+MR. It is certainly "similar" to the early progress they made with KM+TFI.
I would gladly trade all the KM+TFI revenues in 1997 for the kind of progress that I think Ampex seems to be making with KM+MR at this point of the year not only because it is very early in the MR era, but most importantly because Ampex, I think, has been trying to shoehorn KM during the remaining life of the TFI read and write head by focusing heavily on its benefit of increased capacity at nominal cost. With the progress to date of KM+MR, Ampex now has a chance to pitch the broader benefits of KM like "significant capacity improvement," thermal decay, and cheaper head design. The latter, in particular, can be traced directly to this excerpt from Patent No. 5041922
"It has also been unexpectedly found that signal transfers are improved at shorter wavelengths through use of the media of the present invention. Empirical plots (signal strength vs wavelength) for playback signal transfers of data indicate the presence of the differentially permeable layer in the media of the present invention causes a favorable shift towards shorter wavelength at which the first order, wavelength dependent null in response occurs (commonly referred to as "gap null"). This indicates the length of the magnetically formed virtual gap or transducing zone in the differentially permeable layer can be controlled to be smaller than the physical gap of the magnetic transducer.
The physical gap between the poles of the typical yoke-shape magnetic head determines the linear density of the head. Reducing the physical gap is difficult and expensive, but is necessary in order to improve the linear density (again, the number of bits on a track). Whatever the cutting edge in technology is when it comes to reducing the physical gap of a recording head, KM provides the benefit of improved linear density beyond that cutting edge! I don't think that benefit was fully explored as the TFI head era was winding down. Again, it is useful to remember that a MR head is really two heads in one: a MR writer (a TFI head optimized for writing) and a MR reader (a shielded and biased MR element that senses the transitions written by the TFI head using the magnetoresistive effect).
I think that IBM, not Western Digital, is the horse we want to keeper with because they have the best MR technology. In response to Dollar Bill's still unanswered question, Roger Wood, who specializes in channel electronics BTW and who is one of the inventors of Patent No. 5,041,922 (the KM patent), is still with IBM, although, he is on some kind of sabbatical and is working with a Singaporean Data Storage think tank. Also, as recent as 1994 IBM collaborated with Ampex, among many others, on the HPSS (High Performance Storage Systems) software project which is now the current software standard in the supercomputing field. We all know how those supercomputing geeks just love the capacity of the DSTs.
Next, with the kind of roadmap Ampex is developing for KM+MR, I think it makes sense for Maxtor to abide by the minimum purchase clause of its deal with Ampex in order to maintain its preferred pricing status for the next 3/6 years. I am skeptical that Maxtor can get all the MR heads it needs to go 100% MR this year because component makers, like the disk drive makers, have this tendency to become bombastic, especially since the safe harbor law was enacted in late 1995, coming off what has historically been the 2 strongest quarters of the year and facing what has historically been the 2 weakest quarters of the year. Even if you give them the benefit of the doubt, the logical move is to allocate all premium priced MR heads to the highest margin products, i.e., server drives, laptop drives, high-end desktop drives, before it reaches the highest volume (read: cheapest) PC segments. Remember, nobody expects ample supply of MR heads AND practical cost-parity until the end of 1998.
For example, I think Maxtor continues to be a preferred supplier to ACER which is the OEM for a computer maker named Legend which earned the distinction of coming out of nowhere to become the leading computer maker in China after having sold 200,000 mostly $800 PCs in the 1996 4th quarter. China has over 3 billion people and a PC penetration rate of less than 5% of all households, which contrasts sharply with the 35-40% PC penetration rate of the rich North American market. Like most things related to China, the kind of quarter to quarter growth rate that is possible is usually breathtaking. North America may go totally MR by the end of this year, but extremely price-sensitive markets like China will not.
H&Q has projections of global PC sales that might be helpful since 1.5 disk drives are sold for every computer sold:
1997 Estimate of PC sales
North American - Corporate ........................19.9 M units
North American - Consumer .........................11.8 M units
Europe ............................................17.7 M units
Japan .............................................15.0 M units
Rest of the world .................................18.6 M units
Total PC sales ....................................83.0 M units |
