Advanced Micro Devices
Price War?

Format for Printing

Format for printing

Request Reprints


By duttermohl
July 26, 2004

Posts selected for this feature rarely stand alone. They are usually a part of an ongoing thread, and are out of context when presented here. The material should be read in that light. How are these posts selected? Click here to find out and nominate a post yourself!

Catching up on my reading on the board has left me with kind of an unsettled feeling. AMD's stock has taken a big hit in the last few weeks based on an expectation that Intel is going to crank up the pressure on prices in both processors and flash. The Street sees AMD at Intel's mercy; a perception enhanced when AMD did not provide specific guidance for Q3-Q4 and Intel guided for reduced margins.

So what will prevent Intel from crushing AMD?

Can Intel hurt AMD? Of course. During the downturn, Intel added 50% more in manufacturing facilities with two new 300mm fabs at 90nm (like adding 5 - 200mm fabs) and are aggressively converting a couple of more 200mm to 300mm. With the die shrinks and capacity increases, Intel can control a much bigger chunk of the semiconductor market than ever before. Even though the Prescott has had a few teething pains in performance, the yields are good enough to eliminate 130nm production for processors. (And 90nm production is good enough to start 90nm flash!)

So Intel has a lot of production capacity to use up. How much can they cut prices to stimulate demand? Well, Intel has about a 60% gross margin with a 30% gross (before tax) profit, so, simplistically, Intel could cut prices in the 30% range and break even. (Note: Cutting prices 30% but filling fabs would increase margins so at -30%, Intel would still make a small profit.)

** The gross margin percentage for 2004 is now expected to be 60 percent, plus or minus a couple of points, as compared to the previous expectation of 62 percent, plus or minus a few points. The company expects faster growth in products such as flash memories, chipsets and motherboards that have lower margins. In addition, Intel expects microprocessor margins to increase at a rate slower than previously expected due to a slight reduction in microprocessor average selling prices and a slower than expected reduction in microprocessor unit costs.

Given the above statement, Intel does not seem to be interested in a price war. Are there any areas where Intel is planning a price war? I've looked at posts and links for desktop, mobile, server, and value processors and flash to see if I can get rid of my un-ease.


In high-end desktop processors, Intel commands a premium price based on brand, and AMD will almost always trade at a discount to Intel. For example when comparing Intel's best available with AMD's comparable chips:

$422 - Pentium 4 550 3.4GHz LGA775

$288 - Athlon 64 3400


The almost 30% discount for AMD is depressing.

However, when chips rated higher than Intel's best are added in:

$346 - Athlon 64 3500 939pin
$505 - Athlon 64 3700
$642 - Athlon 64 3800

Price per chip parity for AMD happens at somewhat less than 10% performance superiority. At price parity, AMD DOUBLES their revenue. Please note that these ultra high-end chips are low volume and don't affect the overall ASP all that much, but assume they represent a normal binning ratio (that is, the average chips AMD produces is somewhere in the order of 10% better than the average Intel chip). So, to avoid price pressure, all AMD has to do is keep their average chip speed somewhere on the order of 5% faster than Intel and they will prosper. (10% would make AMD wildly profitable).

Can AMD get a 5-10% better binning? That depends on Intel's Prescott ramp, and Intel doesn't seem all that optimistic given that they have accelerated the dual core chips timetable and canceled Tejas. Still, a lot depends on AMD's 90nm ramp. A good ramp with some decent speed increases and high yields bodes well for AMD. So far, AMD has done a great job keeping the lid on 90nm conversion issues, which usually means no real bad things have happened (bad news leaks fast!) The fact that AMD has consistently touted 90nm as on track and rumors that the 90nm chips out in the wild are faster and cooler than those at 130nm, and the fact that AMD looks to cut prices shortly after the 90nm launch, all suggest that the ramp is doing good.

So in the Desktop market, it looks like Intel has nothing with which to start a price war.


In the mobile space, AMD mentioned in the Q1CC that 90nm would yield a 25-watt part that would allow AMD to compete in the T&L mobile space. Given the slowish ramp of Dothan, AMD could gain share in mobile. As evidenced by Freyj's retail survey, AMD seems poised to increase market share in that space. Intel will, of course defend the mobile space with price cuts. So it will look like a "price war", but, in reality, any revenue gain by AMD in this space is a net gain, and a big price war would hurt Intel a lot more than it would slow AMD.


In the server space, AMD is doing well with the Opterons. Sun has released their 4-way Opteron and Opteron based workstations, and the interview with Symon Chang (the CEO of Tyan) sounds like 8 ways will become more common. If AMD can get the 25W Opterons in quantity, then the blade market looks like a winner as well. Still, if Nocona is accepted by the corporate world as the safer route to 64-bits... well, brand loyalty to Intel may slow AMD's market share gains. Brand loyalty to Intel is strongest in this market, so, as in mobile, Intel will trim prices to keep customers, but they will hurt themselves more than AMD if they start a war. As Eachus has said on several occasions, AMD and Intel really don't compete in this market as the superiority of one processor over another for a given task precludes selection based on price differences. AMD will continue to gain 2-way and above sales regardless of price, and Xeons will continue to dominate the low end.

The persistent rumor about Dell developing an Opteron based product line is the kicker. If Dell doesn't adopt Opterons, then Intel has an incentive to cut Xeon prices to keep Dell in the fold. AMD OTOH will have sufficient spare capacity to keep pricing pressure on. If Dell does adopt AMD, they will suck up a significant chunk of AMD's capacity. If Intel tries to cut prices then, AMD won't have to respond with cuts of their own. So, ironically, Dell adopting the Opteron may be the best way for AMD/Intel to avoid a price war!


At the low end, Intel does not have much of a brand name value with the cheapest Athlon-XP 2800 going for just $5 less than a Celeron 2.8. Of course, the Athlon-XP is a much better performer than the Celeron, so, again the performance superiority/price parity rule applies. AMD has an interesting wrinkle with the Sempron. As noted in pfosse's post, Sempron gets about 400 rating points higher than XP at the same clock speed. These are 1.6-2.0 GHz chips, which AMD should be able to crank out all day long at near 100% yields. Even at Celeron prices, AMD should be able to make a profit on them. Prices on these chips are low to begin with, and AMD already has a pretty good market share, so the chances of a new price war here is low, it will just be more of the same price pressure that has existed for years.


Flash is going to get interesting. Intel does not enjoy any brand premium for flash, so the ASP parity that has eluded AMD for years in processors exists in flash. When two companies compete for the same business, the winner is the one with the lowest cost/best quality.

When AMD re-equipped FAB25 for flash (180nm), they claimed the lowest cost per bit for (then floating gate) flash in the world. Modern equipment and really lean labor costs got them that claim. Now that FAB25 is running MirrorBit at 110nm, the cost per bit would be a fraction of that previous world's best. Remember, MirrorBit not only packs two bits per cell, but also is cheaper (by 40%) per cell to make.

Like AMD, Intel has a two bit per cell flash called Strata-Flash, but it is only offered on the 250 and 180nm processes. So it appears that the 90nm Flash Intel is starting to produce is floating gate only. Perhaps Alan81 could tell us if Strata-Flash can be produced at geometries smaller than 180nm?

Given that MirrorBit at 110nm had the bit density of floating gate at 55nm, AMD should still have a pretty big cost advantage over Intel's NOR flash, even with Intel's NOR flash on a more expensive 90nm process.

Like in servers, the superiority of one flash over another limits a price war between AMD and Intel. MirrorBit is most cost effective at big bit counts. As noted by Dr. Bertrand Cambou, Executive Vice President and President and CEO of Spansion, packaging costs eat up die size cost savings for smaller chips. As chips get bigger, the die size savings become more significant and the total cost of the finished product drops. To the extent that AMD markets bigger parts, AMD and Intel won't compete on price. On the smaller, chips, which start with lower ASP and margins, there isn't much price to cut. Although Intel will gain revenue here with aggressive pricing, it is not likely to be from AMD. Samsung, STMicro, and other vendors will be hurt a lot more than AMD.


So in summary, AMD is under pressure to make the 90nm ramp a success (not big news). From what I can see, mobile CPUs will see prices drop as Intel defends market share and small density flash will drop in price as AMD defends market share, but, there aren't any markets where a price war between these two seems like a sure thing.

So I feel better and feel that Wall Street's doom and gloom is misplaced, but the continued drop in share price (ANOTHER 3.5%!!!!) makes me uneasy again!


Become a Complete Fool
Join the best community on the web! Becoming a full member of the Fool Community is easy, takes just a minute, and is very inexpensive.