Dan,<<<So, of your $15 Billion in quarterly server sales and a million units, 3/4 of the units are systems already with Intel inside for $350 million in sales, a good part of the remaining sales are for S/390, RS6000, AS400, SPARC, etc. sales which are going to be gained very slowly, if at all, but Intel has a good shot at the Alpha and HP sales over the next few years, and a good shot at the SUN sales if SUN doesn't shape up.>>>
According to your worst case server processor sales scenario, you hypothesize the following:
1. Low end servers make up 75% of the 4M servers sold each year.
2. 2M of those server have only one processor. 1M has two processors. Therefore, 4M processor sell for ASP of $250, giving the low-end server processor market of about $1B.
3. The ASP for low end servers is less than $5K. The low end server market including software, other hardware, support and services is $15B (3M X $5K).
Okay, let's go with your worst case scenario.
The remaining 1M servers worth about $45B dollars including (in your mind) software, support, services, memory, Disk Drives (although I don't really think they include disk drives from a fractory shipment in the large server category - I think they ship disk drive from EMC, IBM, and where ever else, separately - whatever), and DBMS (same is true here - I don't think they prepackage Oracle, DB2, Informix, Sybase, etc) are currently made up of S390, RS6000, AS400, SPARC from IBM, SUN, and HP.
<<<Remember that when you budget for a large server, you have to include memory for it, disk drives (which may be procured as part of a SAN, Tony, but you still have to buy them and they still go into the budget as server costs :-),>>>
This is a giant logical leap on your part as to how corporations budget and allocate expenses and how that becomes a part of the total dollars that is associated with a server market. You make up this "server cost" category and throw a lot of stuff into it and then relate that number to the server market in some form of association. This is pure nonsense, but it does sound good.
Anyhow, let's continue.
<<<If you want to try your hand at configuring one of the more expensive servers, I think you'll find that Citrix or Oracle oriented server, with 4 large cache Xeons, will run at least $30K to $60K, with $4k of that going from the OEM to Intel>>>
Let's assume that the mid range server market is 75% of the remaining 25% of the server market (about 750k servers) using your mid-range server configuration is worth about $3B (750 X 4 X $1K).
In your scenario, Intels share of the market is at best potentially $4B.
1. Low end servers = $1B (4M processors @ $250) from a total 3M servers and $15B total low end server market.
2. Mid range servers = $3B (3M processors @ $1000) from a total 750K servers and $15B total mid range server market.
That is your take and that is what makes a market.
In my opinion, the entire server market (S390, RS6000, AS400, and SPARC) is up for grabs.
I will agree however that Intel will not make any dent in the upgrade market. Itaniums can compete in any new applications that would have gone to S390, AS400, RS6000, and SPARC. Anyone designing a new application on AS400 has to be examined.
Even if you discount Barrett's contention that only 5% of servers that are needed have been implemented, a 15% growth rate in this market is not an outrageous expectation. Compounded over three years at 15% the server market will be about $90B.
My guess (is that what they call a straw man?) is that Itanium will reach sales within 3 years of between 500K to 1M servers worth anywhere from $6B to $24B to Intel in a completely new market.
Mary |
