NeverMoore?

Arik Hesseldahl believes Moore's Law is about due for retirement. Moore's law states that the number of components that can be built on a chip can be expected to double roughly every year. This rule has held up for four decades, but now...

The complexities and expense involved with keeping it on track are enormous. At the same time, the most advanced chips may not be needed to handle computing tasks for which there is the most demand.

Well and good. However, I read an article some time ago describing the extrapolated limits on computer size as calculated back in the days of vacuum tubes. Given the mean time before failure (MTBF) for vacuum tubes, you could calculate the maximum size of a computer that would be up and running for long enough to do any useful calculations.

Using completely made-up numbers, if your vacuum tubes have a MTBF of 1000 hours, then a computer using 1000 vacuum tubes will run for about an hour before one of its tubes burns out. If your computer uses 60,000 tubes, then one will burn out after a minute of run time. If you need 360,000 tubes to handle your calculation, it had better take less than a second, 'cause that's all you're getting.

Today's computers have considerably more than 360,000 vacuum tubes worth of elements. They run more than a second or two (despite the flakiness of certain operating systems) because transistors don't burn out nearly as fast as vacuum tubes do.

The moral of the story: extrapolations based on the limits imposed by any given form of technology are going to go badly astray sooner or later. More likely, sooner than you expect.