>Do you remember Cyrix? Cyrix had some licensing agreement with IBM and IBM undercut Cyrix in terms of price.<
That would be why the agreement would be for IBM to produce Barton processors. The strategy would be to move the AMD segment of the market as quickly as possible from K7 to K8.
IBM's plans are to start shipping product from Fishkill in H2 2002 and have it at full ramp in early 2003. AMD could commit to producing only Clawhammer and value segment Hammer processors at Fishkill. This means that by the time that Fishkill is at full ramp in early 2003, AMD's half of this 300mm fab would be flooding the market with high end desktop through value segment Hammer processors.
At the same time, IBM would be flooding the midrange and value segment markets with inexpensive Barton and Appaloosa processors. At 0.13-micron, IBM could probably make a tidy profit on Appaloosa at an ASP of $60 and do even better with Barton at an ASP of $100, even with Barton using SOI.
AMD could agree to sell each value segment Hammer processor at the same price as IBM's same speed grade Barton processor, depending on the lower cost of the Hammer mobos to actually take the value segment Hammers into the value segment of the market, at least initially.
In addition to having a monopoly on Hammer (at least until the AMD/IBM joint venture Fab35 comes online with the migration to 0.10-micron design rules and introduction of Timberwolf/Greyhound/Dingo), AMD would have a monopoly on mobile and low-power server variants of Barton.
Of course, the IBM deal is purely hypothetical. AMD could strike similar deals with other potential players, such as Siemans, TI, or NEC. The point is that the 20% foundry cap can be sidestepped by AMD actually buying into a foundry, rather than farming out production of processors to foundry services, and could in the process create an alliance that could cripple Intel. |
