I see three main failed predictions:
- 10 GHz
- long pipelines, I think we're currently at around 14 stages, which is more than P6 but less than Netburst (Pentium 4)
- EUV lithography is still not a thing at 10 nm