Predicting an end to Moore’s law sometime in the next 20 years is probably foolish right now.

1. The next 3 manufacturing nodes are all down to engineering. So predicting an end before 2020 is just clearly wrong.
2. Research labs already have functioning CMOS photonics and graphene transistors. So the switching speeds of chips could skyrocket 10-20x even as transistor densities plateau. I’m pretty sure I would consider a 300GHz graphene processor or a 10THz photonic CPU a reasonable extension of the price/performance regime of Moore’s law even if the physical features on those chips grow in size during the initial production nodes for those respective technologies.
3. Here’s the status of the next 20 years worth of technology nodes on Moore’s law. This doesn’t even take photonics, quantum computing, or reversible computing into account:
   • 22nm chips = currently out
   • 14nm chips = demoed and will be shipping in a few months
   • 10nm chips = experts see “clear way” to do this
   • 7nm chips = Silicon-Germanium transistors will work
   • 4nm chips = prototypes 3 years ago for quantum computers
   • 2nm chips = prototypes available 2 years ago
   • 1nm chips = exotic graphene prototypes made 5 years ago
   • 0.2nm chips = single atom transistors made last year