The Inquirer-Home

Itanic shell game continues

No numbers that matter once again
Tue Jul 18 2006, 20:28
ONE OF THE main reasons that the Itanic is the running joke of the IT industry is not anything silicon related, it is simply an accounting issue. The problem is that Intel simply won't come clean on the numbers, and worse yet, they toss out weasel-words when asked.

If you look a the projections, results, and other ways of telling the overwhelming superiority of Itanic, Intel always quotes system revenue. Last time I checked, the count for Itanic systems that Intel sells on the open market is zero, up from zero on the last Itanic variant. Intel sells chips, occasionally chipsets, mobos and other widgets, but they are de-emphasizing that as quickly as possible in this segment. So the quoting of server revenue is several shades beyond disingenuous.

alt='itanna'This dodge is getting old, and it hides the real problem, the Itanic revenue model is broken. Every other competitor is vertically integrated, the competition makes chips, boards, systems, OSes, sometimes software, and service/support. Intel makes chips. Their partners may make some or all of the pieces, but none make the entire things because they don't make the chip.

IBM can give away servers and make it up on software or support. Ditto for Sun. If you are sitting in a negotiation session, the carrot dangled of 'this server is free with a 10-year support contract' is quite enticing. Ask SGI about this one, IBM beats them up with it all the time. Now, the end result is the same, it costs what it costs on average, but IBM can tune any single part of the system to make a bid, Itanium vendors can not.

This leaves Intel with a small piece of the pie, and no wiggle room, what a wonderful business model. As the Olympic class footgun skeet championship that is Itanic continues with startling speed, Intel has been retreating from markets faster than you can say 'circle the wagons sheriff', but luckily, they are retreating upmarket instead of down.

I say luckily not because it is good for the chip, far from it, but it is good for PR. When you sell a server with 64 CPUs for $2500 each in a system, Intel gets $160,000 out of the deal. If that server costs a hair under $6 million, well, there is a lot of 'other', and Intel's share of that 'other' is zero.

If you are charitable and and say they got the full $3692 out of HP for the CPUs above, that means Intel gets (64 * $3692)/$5,937,555, or a whopping, lets be charitable and round up, 4%! Wow, what a business model there. Now, you can say that TPC-C is an extreme case, so lets double Intel's share of the system revenue for the sake of argument. If P4s sold for $100, that would mean a PC would have to have to cost $1250 or so to make the same low margins, and last time I checked, Intel had margins in the 50% range.

What can Intel do to fix this? Euthanasia would stop the bleeding, but there is not much else you can do. The competition can sum up the parts of the system and service, and as long as them make an average profit, they are in the right business. Intel has a chain where each step has to make money or they all fail. IBM is laughing all the way to the bank.

That brings us back to weasels, and the numbers they fling. You can see that when Intel quotes the triumphant number for system revenue, they are hiding a hollow core, pun intended. They will never give you CPU sales numbers, or even CPU shipment numbers, they are quite different in this case, you only get system revenue.

This game is getting tiring, and what is worse, some analysts keep falling for it. When are they going to ask the real questions about the Itanic line and stop playing kissy-face? Nothing will fix the broken model, but the sooner we can get the real numbers, the sooner we can figure out what is really going on. Oh wait, that is exactly what they don't want. µ

 

Share this:

blog comments powered by Disqus
Advertisement
Subscribe to INQ newsletters

Sign up for INQbot – a weekly roundup of the best from the INQ

Advertisement
INQ Poll

Heartbleed bug discovered in OpenSSL

Have you reacted to Heartbleed?