AMD: time for the kill

AMD's tech analyst meeting yesterday was huge. The most impressive one was AMD's rapidly expanding capacity and its quadcore roadmap with next gen core. AMD is planning to add another 5K wspm to FAB36 and convert FAB30 to 20K wspm at 300mm. The conversion of FAB30 is a brilliant move, as it cuts the time. The total 45000wspm capacity will be enough for 80% of the world's x86 processor consumption. Also, AMD stated that Chartered FAB7 will be a major supplier.

In the near term, by the end of July, with FAB36 ramping and Chartered FAB7 producing one million units, AMD should have enough capacity for 30% of the CPU market. This represents a 50% increase of AMD's capacity over 1Q06. What will AMD do with this extra capacity?

Sell them.

As I previously pointed out, Intel will be most vulnerable from 3Q06 to 4Q06, during the transition period from Netburst to Conroe. Right now, Intel is stuck with $7 billion worth of legacy chips ($3.5 billion inventory, 49% gross margin), those are Netburst based chips that are hot and slow. Moreover, Intel has squandered most of its cash. Intel is financially very weak. It can't continue to feed 103K employees and will start mass layoffs as I predicted last year. Intel is hoping to convert this huge pile of legacy Pentium 4 chips into cash to fund the production of its Conroe CPUs. Due to Intel's manufacturing inefficiency, it won't be able to switch to Conroe production quickly. Its projected units for Conroe is 35% in 1Q07. Intel must dump its legacy inventory, or it will have to write it off later. Intel's price for the Pentium D 945 3.4GHZ dual core CPU is expected to be $163.

If AMD can succeed in preventing Intel from clearing the legacy inventory at a good price (above $120), Intel will find itself unable to fund Conroe capacity expansion. To achieve this strategic goal, AMD should prepare to sacrifice margins on desktop CPUs. While maintaining prices on Opterons and Turion X2s, AMD should introduce a low priced dual core and single core processors to stop the Pentium and Celeron dumping. There should be a 2.2GHZ, dual core, 2x 256KB cache AMD64 chip, priced at $150. There should also be a 1.8GHZ, 128KB L2 Sempron at $50 to halt Intel's Celeron flood in the 3rd world. In 1Q06, AMD's ASP was about $95, with rapidly increasing capacity, AMD can lower the ASP while maintaining revenue growth for 3Q06 and 4Q06.

By the end of 4Q06, AMD will have 50% of FAB36 converted to 65nm and 65nm parts shipping in volume. AMD quad-cores should be out by then too. AMD can then start a trade-in program for Sempron AM2 users to upgrade to 65nm dual core chips.

Even with massive cuts on desktop CPUs, AMD can still profit nicely from Opteron, Turion 64 X2 and 4x4.

AMD MUST make 4x4 available for the general market by dropping the requirement for FX CPUs. Why not give this awesome 4x4 power to the non-gamers? Personally, I like to have a 4x4 equiped with two energy efficient AM2 X2 3800+ at 35 watts each. Others may just start with one CPU, reserving the right to add another one. AMD 4x4 gives AMD64 systems much longer life. In this sense, 4x4 for desktop is a disruptive technology. Those CPUs with 4x4 stickers can sell with a 20% premium.

AMD's added capacity comes online with almost perfect timing. As long as AMD can sell all of its capacity, Intel is bound to collapse due to is weak financials and huge overhead cost.