Wednesday, June 2, 2010

Moore's Law vs. Learning Curves

Moore's Law is not alone in the world of manufacturing thumbrules. Haitz' Law is a corollary manufacturing thumbrule for Light Emitting Diodes. Admittedly, these Laws are exceptions to the general rule. The general case is better described by learning curves.

Ct = Co(qt/qo)^-b
PR = 2^-b
LR = (1-PR)

Co/Ct = initial/final cost
qo/qt = initial/final production
PR = progress ratio
LR = learning rate
b = learning coefficient

If you have a decent price/production data set you can solve for b, PR, and LR. You can then resubstitute the numbers to "predict" what the future costs of production might be. Two basic caveats: 1. These equations should only be applied to young industries. 2. The results are all SWAGs.

Example a) Photoelectric cells have historically had a learning rate of about 20%. Cell production was about 10 GW/year in 2009 and if you assume a constant 25% growth rate it will reach 100 GW/year by 2020. Since we know production costs were about $1.50/watt in 2009 we can "predict" that production costs in 2020 will be about $.57/Watt.
Example b) If you hold all the variables above constant but change your assumed growth rate to 35% you get production costs of about $.47/Watt in 2020.

As pointed out above, these examples are educated guesses. Thing is, when you get down to it, Moore's Law and Haitz' Law are also educated guesses. The surprising thing about Moore's Law and learning curves in general is that, when all is said and done they work pretty damn well. This is why governments and manufacturers continue to use them to guide policy and inform strategy.

The IF game part I... If Haitz' Law holds up for another 10 years we can imagine we'll be seeing a lot of LEDs. Given the validity of this risk we can imagine that Phillips and GE have transition plans and roadmaps for their lighting divisions.

The IF game part II. If photoelectric cells continue on their path for another 10 years we can imagine we'll be seeing a lot of them. But wait... something funny happens when people suggest that the learning curve for solar cells might hold up for another 10 years. Educated guesswork morphs into techno-optimistic faith. In general, you can put it down to territorial emotionalism and ignorance. Given the validity of this risk we should be thinking about transition plans and roadmaps.

No comments:

Post a Comment