As long as shrinking feature sizes and increased wafer diameters are the primary means of chip cost reduction and performance improvement, Moore's Law remains a good approximation, but we are now entering an era where fixed costs of design, packaging, embedded software, etc. are an appreciable percentage of cost per unit. Using the learning curve, we can predict future directions in technology that will deviate from Moore's Law. We can also predict where non-traditional innovation is likely to have its greatest impact. The speaker explores these projections for future technology and suggests where the discontinuities are most likely to occur.