He might add a footnote to this statement "28nm and 14/10nm likely to be long-lived nodes" such as "40nm with GaAs competes with current silicon at lower nodes and further scaling with GaAs will outperform any silicon that is currently available that is affordable".