How Intel's geese lay golden eggs
And why AMD is in a pickle
Analysis Just after last Christmas, adverts started appearing in the UK press for Celeron processors, with an egg-related theme. The ads may have run elsewhere than in the UK, and showed eggs in an eggbox, each branded with the Intel Inside logo. At the time, we dismissed this as just another bit of Intel tomfoolery but it has occurred to us since Easter that we were deeply mistaken. Chocolate company Cadburys now sells its Easter eggs all year round, although at one time sales were confined to a period just a few weeks before. The thin end of the Cadburys wedge started maybe 20 years ago, when a week or so after Christmas, its cream eggs started appearing in shops, much to everyone's shock. Intel sells eggs too, but as its Q1 figures show, they are golden eggs which hatch into gross profit margins of nearly 60 per cent. Just like chocolate eggs or real eggs, its chips have a limited shelf life and can go very smelly and pooey if they're left hanging around too long. Is this true or false? The answer is yes and no. For Intel, chips cannot be allowed to hang around too long. The geese in its fabrication plants have to keep laying new, bigger eggs. The end user must keep buying these eggs if margins are to stay high. That is not good for consumers, who do not necessarily need a 666MHz or a 1.2GHz Intel chip to run their copies of Encarta 2000. If you compare prices of the Pentium III now to the prices they will be in September, you can see how this Intel egg machine works. The graph here shows how Intel is practically halving the cost of Pentium III 500/400 chips over the next six months. Overegging the cake A couple of years ago, Craig Barrett, recently promoted to take over day-to-day duties from the poorly Andy Grove, outlined Intel's plans for the segmentation of the microprocessor market. The sub-$1,000 market, which he originally called the Basic PC market, would be served by the Celeron. This particular egg would be packaged in a Slot One configuration but without the cache associated with the Pentium II. At the time, many people pointed out that this was Intel sleight-of-hand. The Celeron is a Pentium II but with bits taken out, they said. But Intel did not want to be caught with too few eggs in its basket. It foresaw the growth of the sub-$1,000 segment a little too late and re-engineered its marketing and its packaging to sell the 14-carat version of its golden egg. To this end, it poured millions of marketing dollars into the Celeron platform, introduced products at cutthroat prices and continued selling the Pentium II at much higher prices. This pattern continues in the latest set of prices we have obtained from one of Intel's OEMs. Fee, fi, fo fum, I smell the blood of an AMD man When Advanced Micro Devices (AMD), filed its S-10 form with the US Securities and Equity Commission (SEC) a week before bad financial results came out, it was plain, as eggs is eggs, that Intel's pricing on Celerons and Pentium IIIs was hurting it badly. Unfortunately for AMD, it does not have as many geese laying golden eggs as Intel, and its prospects, geese factory wise, are also very limited. When Jack came back with the market with a few beans rather than money from the cash-cow, he got sent to bed without any supper and the useless beans got thrown into the garden only to sprout into a huge beanstalk next morning. Plucky little Jack eventually made off with the Giant's golden goose, and as he returned to the surface of the planet, undercut the beanstalk, pitching Intel to its doom. And everyone lived happy ever after. This market, however, is not a fairy story. The likelihood of Giant Intel falling off the beanstalk to its doom is slim. These are the market realities. AMD's K7 may be more golden than anything Intel has currently to offer, but Jack cannot produce enough of them to compete with the Giant. Worse, the chip giant has noticed AMD undercutting its share in the retail market and will continue to slash its Celeron prices and it will continue to erode the average selling price (ASP) of K6-IIIs and K6-2s. It will do the same with the K7 when limited quantities start appearing in July. Is AMD's very eggsistence threatened? When we were being roasted by the desert sun at the Intel Developer Forum last February, we were sitting behind one guy from Chipzilla who was chatting to a customer about AMD yields. If you're attempting to compete in the marketplace with a giant like Intel, the number of eggs you have in your basket is important. Intel has a huge basket and lots of eggs. They're not just desktop eggs but they're mobile eggs, server eggs and enterprise eggs. All of them make a profit and while some (like the original Celeron) are addled, most ain't. AMD hasn't got many eggs and its basket is tiny, though 300 or so people are wickerworking like mad. We couldn't help overhearing some of the Intel conversation about AMD yields and hastily scribbled down some figures. These, it must be stressed, are Intel projections and not Register speculations. K7 .18 micron technology will come on line in the first half of next year, produced at its Dresden Fab 30. This year, it will start producing .25 micron technology at its US Fab 25. Intel's thinking is that in the year 2000, it will manage over 5,000 wafer starts a week at Fab 25 and 1,500 at Fab 30. That is likely to move to over 5,000 at Fab 25 and over 5,000 at Fab 30 in 2001, meaning half of the wafer starts will be .18 micron. AMD may ship as many as 28 million K6-IIIs next year, but only four million K7s. The year after, the K6-III will virtually disappear with only eight million being sold, while the K7 will rise to 50 million. Full capacity at Fab 30, which will cost AMD nearly $2 billion when it is fully staffed up, is likely to be only 5,000 200mm wafer starts a week. Logically, it must therefore somehow find (borrow) the money for a new fab in the relatively near future. The Intel estimate, according to the overheard conversation, is that a .25 micron K7 will cost AMD $60 to make this year (including the cartridge), a .18 micron K7 will cost around $150 at the beginning of next year, but in the second half of the year, that will drop to around $60 again. These estimates, if they are true, mean that AMD needs a white knight to ride to its rescue. This is given that Intel has said it will continue to compete and compete highly aggressively in all sectors of the market. It is also a given that Intel will move early to .18 micron technology and its fabrication capacity is enormous, compared to AMD's. How much margin can AMD make on the limited quantities of K7s it can ship, while its shareholders bay at its door? While AMD has a deal with IBM Micro to fab up its chips, this also costs money. Perhaps, the only white knight that could conceivably save AMD's bacon is Compaq. This could be less likely now that Eckhard Pfeiffer has resigned as CEO. Another firm that has the means is Motorola. Eggs is eggs The Register does not take sides in the battle between Intel and AMD. We know from our internal sources at both companies that US executives view us as hostile. We are not football supporters and we do not consider that this is a battle between Jack the Giant Killer and the Giant itself, where we have to cheer on the underdog. Intel has 65,000 employees, vast financial liquidity, huge manufacturing capacities and excellent process technologies. Its revenues to the end of 1998 amounted to $26.273 billion, its net profit $6 billion, and its assets $31.47 billion. Its long term debt was $702 millions. AMD has 13,000 employees, limited manufacturing capacity and has been caught on the hop by its yields this year. AMD has a turnover of around $2 billion a year, and some apparently excellent technology in the shape of the K7. AMD's revenues to the end of 1998 amounted to $2.54 billion, its loss to $104 millions, its assets to $4.25 millions and its long term debts $1.37 billions. Quite obviously, AMD is no underdog. From the outside and from the outset, we have consistently reported instances where we feel Intel has taken unfair advantage of its size and position in the market. The US Federal Trade Commission (FTC) looks like it will endorse a settlement with Intel, which will not be good for the industry. But AMD has its own house to put in order, and the financial and fabrication facts speak for themselves. ®
Sponsored: RAID: End of an era?