XbitLabs talks a bit more about the Smith Barney report on AMD’s ability to make Hammers.
There is also a post in Investor’s Hub with additional information from the report.
They both become a lot more interesting if you compare the numbers in them to the estimates from the Inquirer (which most likely came indirectly from AMD) just a bit over three months ago.
So far, those reporting on the story have focused on AMD actually only making 170K Athlon 64 when they had said to the mobo makers that they were going to make more than that (300K or 340K or 433K are the figures being floated around).
Interestingly, though, I also found one mobo maker refer to 100K chips at the time of Computex, too
There’s a claim by an AMD person out in a branch office that they actually sold 500,000. That’s nice, but we’ll note that “sold” and “shipped” can be two entirely different things, and why isn’t corporate headquarters making such announcements? After all, that’s where the numbers would come from?
Seems like AMD was telling different people different things.
In any event, the number of chips AMD actually sold should become pretty clear in the next few days when AMD releases their finanacial results. Last quarter, AMD sold $503 million worth of CPUs. Since it’s likely Athlon XP sales went up somewhat, too, if AMD’s CPU revenues are in the $600-650 million range, that means the Smith Barney estimates are roughly right. If the figure is more like $750-800 million, that means the AMD Taiwan estimate is roughly right.
How many CPUs got sold in the fourth quarter is relatively minor, though, compared to the numbers expected to made over the next six months.
By far the most important comment (from the Investor’s Hub post is this one):
According to Kirk Yang (see ‘The Asia Motherboard Beat’ of 12/24/2003), AMD’s
channel partners are apparently being told 700-800k Athlon 64 units are
likely in Q1 and 1Mu by Q2’04. (Kirk Yang, BTW, is a stock analyst for another brokerage firm.)
Let’s do a comparison between what mobo makers were led to believe back in September with what they’re saying now:
AMD Then: 1690K
AMD Now: 700-800K
Big dropoff, isn’t it? Now look at the next quarter:
AMD Then: 5120K
AMD Now: 1000K
Huge dropoff, isn’t it?
Mind you, the “AMD Now” estimates are supposed to be what AMD is telling suppliers. These are not the Smith Barney estimates. Smith Barney doesn’t think AMD is going to even make the “AMD Now” estimates and implied rather strongly that AMD’s estimates to the mobo folks were . . . uhhh . . . fibs.
What these figures mean if correct is that AMD is not going to seriously ramp up Hammer production until the second half of the year, and probably won’t have croosover (make more Hammers than XPs) until the last quarter of the year.
What’s The Problem?
The Smith Barney report assumes that since AMD is not going to make a lot of processors, the reason for that is they can’t make a lot of processors.
I don’t think that’s the only or entire explanation.
What is probably the major factor operating here is not that AMD can’t make the CPUs, but rather that they can’t sell them to a very important group of people, namely the big desktop OEMs.
They’ve not done a lot with desktop Hammers so far. None of the big guys are selling FXs. For Athlon 64s, Dell doesn’t (but that’s normal). Hewlett Packard is selling only one HP and one Compaq model with an Athlon 64 3200+ currently, and they consider this to be a pilot project.
eMachines does have two models out. IBM isn’t selling any. Gateway isn’t selling any.
AMD can’t sell five million CPUs a quarter through Newegg, Alienware, and Falcon, you know.
Why won’t the OEMs bite? The AMDroids will of course blame Intel and only Intel, and no doubt they’re trying, but if Intel really could keep everybody from buying AMD products, why would they “allow” Athlon XP sales, which still constitute the vast majority of AMD CPU sales?
No, the OEMs must have some problem with the desktop Hammers themselves.
It could be that Smith Barney is right and AMD really can’t guarantee that they can make enough Athlon 64s to satisfy the OEMs, though that isn’t necessarily a technical problem.
Athlon XP sales remain pretty robust. Trying to continue to make tons of those to keep the same OEMs happy, plus make a ton of desktop Hammers while converting the fab over to 90nm might be just too much for Dresden, and if the OEMs didn’t want to commit to buying enough A64s for AMD to convert XP production to A64 production for a short period of time, it makes sense to keep the bird in hand rather than go for the two in the bush.
A big reason could also be the lack of Windows for AMD64. It messes up the marketing to try to sell a 64-bit system that can’t run 64-bits out of the box.
Or it could be as simple as the processors have been/are just too expensive for a big OEM to build a big-selling model around. Before you bring up Intel 3.2s or EEs, those CPUs are just the most expensive option in a wide range of PIVs and Celerons. Until a week ago, you had just one (expensive) Athlon 64 processor to choose from (not counting the apparently half-broken Athlon 64 3000+).
Any one of these could be a factor. They could all be factors.
What does it mean? It means that for practical purposes, desktop Hammers won’t become AMD’s mainstream seller until the 90nm ones start coming out.
Is that bad for AMD? Well, it’s not good, but it’s more a missed opportunity to make more money than financially ruinous. Unless Athlon XP sales go down the tubes, the revival in PC sales ought to be enough to let AMD muddle through this, provided, of course 90nm transition isn’t a problem.
A Sense of Proportion
Many continue to grossly overestimate the number and importance of Opteron sales. The Smith Barney report also provides estimates of what AMD sold in 2003, what it got for them, and projections for 2004.
Average Selling Price
**Athlon 64 estimates do not take into account the drop in estimated sales for A64s in Q4 by Smith Barney. If they did, the
figure would be about half the stated, or about $60,000,000.
Please note the very low figure for Opterons and the minute proportion of revenues they provided to overall revenues.
You may say that Opterons need time to get established, and I’d agree, so let’s see what Smith Barney estimates for 2004
Average Selling Price
**Don’t read too much into that $146 figure. Smith Barney’s financial model assumes that the bulk of A64s would be sold at the end of 2004, and remember that the cheapy “Paris” chip is supposed to come out by then. So don’t assume that you can get a cheap A64 real soon.
Again, the Opteron figures aren’t very big, and their contribution to revenues is dwarfed by either Athlon series. Even if you doubled the Opteron numbers, they still would constitute only about 15% of AMD revenue.
This isn’t a knock on Opterons; if we had a breakdown of Intel sales, you’d see much the same for Xeon sales as compared to PIV sales. In both cases, these server lines are a very profitable sideline (a good deal more money for relatively little extra effort), but desktop sales are the main show.
To expect otherwise is like expecting the tail to wag the dog.
The Smith Barney report is politely scathing about how AMD does business.
“Recent information suggests that Athlon 64 shipments were roughly 170ku in
Q4. AMD has not given an official target for Q4, only saying that unit
shipments would be “in the hundreds of thousands.” Channel partners were
apparently told at the start of the December quarter to be prepared for 300k
units. . . . perhaps the 300k target was given to channel partners to create a sense
of critical mass.
“According to Kirk Yang (see ‘The Asia Motherboard Beat’ of 12/24/2003), AMD’s
channel partners are apparently being told 700-800k Athlon 64 units are
likely in Q1 and 1Mu by Q2’04. If the 170ku figure is accurate, it would
make the chances of 700k-800k units for Q1 seem improbable, in our view, and
again may be intended to create a sense of critical mass. We are currently
modeling 500k for Q1, but there is a wide range of error possible around that
Our investment thesis for AMD largely hinges on the company’s ability to make
Athlon 64 units in sufficient volumes, and our Sell rating is based upon the
belief that AMD will struggle with yields for at least the next six months. . . .
We would get more positive on the company if evidence of Athlon 64 manufacturability were to come to light. . . .
Bottom line, almost every quarter for AMD is difficult to forecast, and it
sounds like there are still enough pieces in motion right now to cause us to
avoid the stock.”
These are nice ways to say, “We think AMD has been BSing their production figures to mobo makers to try to get
more of them to make boards for them,” “AMD won’t answer our questions” and “AMD’s information is unreliable.”
Understand what’s going on here. This is not AMD giving the finger to some Internet journalist. This is AMD giving
the finger to a major investment firm that advises a ton of clients (Smith Barney is part of Citigroup, one of the largest
financial institutions in the world) on whether to buy or sell AMD stock. This is AMD preferring to have Smith Barney tell its clients
to sell their stock rather than answer Smith Barney’s legitimate questions. This is insane unless the story is basically true.
To be crude about it, it’s like a potential ladyfriend asking you if you’re . . . length-challenged. It would be insane to refuse comment
about that if you’re not. It only makes some sense if it is basically so, and telling the truth would leave you at least as badly off as a “no comment.”
Well, there’s one difference between you and an AMD spokesperson. You don’t have the SEC to worry about. You won’t end up in prison for fraud or be sued if you lie to the lady (at least not yet), which could happen if AMD flat-out lied to those stock analysts.
You may say, “Well, Intel hasn’t been very forthcoming about Prescott.” You’re right, and that proves the point. Companies try to hush up when the news is bad. Intel is now talking
quite a bit about Prescott at CES, which means they’re now ready to get it out the door.
When a company won’t answer legitimate questions asked by important institutions, you start looking for tea leaves. One indirect indicator of
how AMD is doing is what the company execs do with their stock options. AMD execs have historically been much more likely to get their options and sell right away than
Intel execs (Intel execs tend to get paid less, but get more options on better terms than AMD execs). Follow the links and you’ll see how much stock Intel execs hold compared to AMD execs.
A few years back, when AMD was nearing its stock price peak during the Thoroughbred era, there was a massive unloading of AMD stock by
the execs, including Jerry Sanders (former head of AMD). When AMD’s stock prices crawled around $5 a share, only Sanders bought back a hefty amount of AMD stock
at the time. There hasn’t been much stock activity by AMD execs until recently because the options issued the past few years were worthless until the company’s
stock price got up to around $14-$16 a share.
Well, now AMD stock is around that price level. What are the execs doing?
Up to now, with one exception, they’re pretty much staying pat. The exception is Robert R. Herb, who is the senior vice-president of sales and marketing at AMD. Over the last couple months,
he’s sold over a quarter-million shares of AMD stock. He doesn’t own any at the moment.
It’s a U.S. government requirement that people like executives of a company have to report their purchases and sales of stock in their company, which are a matter of public record. So this isn’t a rumor or speculation; it’s a fact.
His activity alone shouldn’t necessarily signal a red alert (he may have needed the money for something else), but it’s something worth watching. If AMD’s stock price stays the same or heads up somewhat higher (making the stock options out there worth good money), and the AMD execs start cashing almost all of their chips, that could well be a big warning sign.