Replies to post #26757 on Advanced Micro Devices Inc (AMD)

[sgolds]

HailMary, I took wbmw's explanation to exclude initial developement through Merced as money not being recovered. He believes that current costs are being funded by sales of Itanium 2.

I have to be from Missouri on that one (show me!). We know nothing about the yields and total costs of Itanium, so this is guessing. I can believe that engineering R&D is covered by existing sales, but one has to consider manufacturing R&D also. How about production costs? Corporate overhead to support all these people?

He also discounted marketing that is combined with other Intel marketing. This is not how companies typically account for these expenses - they partition it between the products being marketed, and internally calculate the profit/loss on these products separately.

Itanium 2 still seems to be in startup mode (that means operationally losing money) and the jury is still out about whether it will become self-sustaining.

[chipguy]

I find the above statement hard to believe though. 100000 Itaniums at $1000 profit each is only $100M. The research on IPF has to be in the billions by now, isn't it? Or is that another overblown number that gets tossed around?

I estimate that the current level of R&D expenditure on
IPF (at least two full design teams, chipset team, large
chunk of the in-house compiler effort, test chips etc) to
be on the order of $350m to $400m per year.

Barrett estimated 2004 sales of IPF to be 200k units but
I think he is being very conservative. At $2k ASP that
is $400m right there not including white box/chipset sales
to small OEMs and resellers. Replacing a good portion of
PA-RISC, Alpha, and SGI MIPS will bring IPF to well over
500k units per year. SPARC is a juicy target for IPF and
it runs around 1m units per year. One analyst has suggested
IPF could ultimately approach 2.5m units per year just in
the big iron market. All of that is in a market segment
within which AMD is just starting from scratch with Sun's
help (given they are sticking to SPARC big time one has
to question how much help Sun will really provide). Intel
has clearly stated that it thinks IPF will achieve cost
parity with Xeon but at twice the performance by 2007. If
that is even partly achieved then IPF will see additional
big sales well south of the big iron segment.

IMO Intel and HP have invested around $3B in IPF to date.
That is a sunk and expensed cost, there is no big "loan"
that needs to be repaid as some seem to think. Neverthe-
less I think that all those billions invested will be repaid
with return within the next five years. Beyond that Intel
will have a basically unclonable chip family that sells
for >10x cost of manufacture in markets with extremely
high barriers for others to enter. What's there not to like
about IPF if you're an Intel shareholder? Or just somebody
long sick to death of x86 yet embarrassed by the technical
bungling associated with SPARC and PowerPC. ;-)

[wbmw]

HM, Re: 100000 Itaniums at $1000 profit each is only $100M. The research on IPF has to be in the billions by now, isn't it? Or is that another overblown number that gets tossed around?

You are trying to reconcile the money that has gone into IPF in total with the revenue generated in one year. I have seen plenty of estimates of total IPF $ spent, but if we are talking profitability, I would be more inclined to compare the costs for one year with the revenue for one year.

In terms of costs, there is R&D (fixed costs) and manufacturing (variable costs). I think Chipguy's estimate of R&D is a good place to start - $300-400M. I would estimate the manufacturing costs to be $200-300 per chip.

In terms of revenue, I think you are being conservative about ASP. I think it's closer to $1750, given that a large amount of CPUs sell into Superdomes or Altix 3000 machines for $4,600 a piece. It also seems that Intel exceeded their goal of 100k Itanium 2 CPUs and instead sold about 110k. Given about $250 in manufacturing costs, that's 110k CPUs for $1500 of profit each, or $165M.

IPF may have a lot of expenses going into it, in terms of compilers, software porting, marketing, sales, etc, but from a purely CPU R&D perspective, I think $165M is already covering that. If Intel reaches their goal of 200k units next year, it will come close to covering the other expenses, too.