This is an interesting post from the Sprint PCS thread (of all places). What he says about power/heating problems with increasing die shrinks makes some sense, but I'm not sure that has prevented Intel from continuing to shrink geometries so far. Any of you semiconductor engineers know if this makes any sense?
Message 18133134
Telecom/networking industry appears to have turned a corner: PCS FON NXTL Q VZ SBC BLS DCEL LVLT T AWE DT FTE WWCA USM TPC LWIN APS all have bottomed out and some are even on the upside. This is due to capital markets and a technical wall—so read on this gets interesting.
Capitalism is a always a catch 22, if your stock is too low and the business not going well, you can not get money from the capital markets, and if the stock is high and you don't need the money then you can get money from the capital markets. The surviving telecom companies named above are going to issue more common shares(finally--not preferred convertible, not bonds and not lone-shark bank credits) to survive and reduce debt. The fact that their stock is going up it means the capital markets are looking at this impending event more favorably. Even the Foreign telecom companies such as DT and FTE which reside in anti-capital anti-corporate semi-socialist Europe are going to do their dilution to the government and become semi-national again. There are many others like Worldcom, Williams, etc. which are in debtor-in-possession (DIP) financing situation and in bankruptcy are emerging--Williams came out already and there will be a hoard of others like Worldcom that will come out early next year and all of their DIP financing will turn into common shares so those companies will also be debt free. Being debt free means you can spend money to buy equipment and invest in your infrastructure.
The state of of networking in general is also improving—why? There is an incredible wall that the semiconductor industry has to jump over and little or no progress has been made. Drum roles---that is the problem of subthreshold current of FETs at .01um and below. This problem is a physics problem and can not be solved by improving etching techniques and fancy footwork. NO ONE HAS THE SOLUTION and it appears no one will for several years. The consequence of this wall is that power increases exponentially as you go to smaller geometries than .01um. Finally the laws of physics overrode Moore’s non-scientific-Law. You can make the damn thing faster but you will pay the price in power and life of the chip. But don’t sell your semiconductor shares yet—this actually bodes well for the industry because they now will have to sell more chips, how? Though the problem of subthreshold current means individual PCs have hit a speed wall, everyone already in consensus that no one needs faster PCs. The good news is that servers have hit a speed wall and now we have to throw the big computational problems to distributed computing, or network computing. The business proposition of faster semiconductors as a marketing ploy is also dead because everyone who knows anything about semiconductor reliability knows for every 10C increment in junction temp you reduce the life by a factor of 2--just to make the point: 20C increment factor of 4, 30C factor of 8, 40C factor of 16, you get the point. The smart CIOs know this and the dumb ones will be fired real fast. The die intensive parts (parts that require large amount of silicon to make and inherently have low yields in manufacturing) such as switch fabrics will have to bite the bullet and go to smaller geometries and take the hit in semiconductor life but the network will have to compensate with redundant modules to avoid outage, again they are going to sell more chips. But going back to the original theme--distributed computing means more data passing through the networks, and networks and computing are now inseparable propositions.
So, a final point: it does not matter who buys who, who has more cash, who has more whatever, what matters is that the patient has turned the corner. This roller coaster has bottomed and as fast as it took us down is going to take us up. |
