Until the last couple of years, the progress was metronomic; every 18 months, a new process node was adopted. But things have changed and it is becoming harder and harder to move to that next node, despite the industry's 'can do' approach.
Leading manufacturers, including Samsung and TSMC, are already talking about having 10nm processes available at the end of 2016 and there are plans for a 7nm node. But there is more than a suspicion that moving beyond 7nm might be extremely difficult and, in some people's opinion, not practicable. One reason is that there are simply not enough atoms to work with.
Gold Standard Simulations – the spin out from work undertaken at Glasgow University – has been developing software which uses a statistical approach to help semiconductor manufacturers understand the risks involved in process shrinks. CEO Prof Asen Asenov believes the most critical factor is how to manage risk. "And there are big risks in chooosing technology."
GSS' focus is beyond the 10nm node, where Prof Asenov believes the best candidate is nanowire transistors – also known as gate all round transistors – which are likely to scale to the 5nm node. "But 5nm is likely to be the end of scaling," he said. "Beyond that, it's likely to be physically impossible."
So the industry's focus will have move from 'More Moore' - continual process shrinks - to 'More than Moore', where the focus is more on applications. But the 'can do' attitude will still be a major factor.
