Although his details are likely correct, this is not an analysis from physical laws (and other hard requirements, eg usability requirements) which is what you need to show optimality, it's an exploration of the known options.

He makes the point that labs wouldn't adopt a cheaper machine - yes, much cheaper processes are often adopted first by outsiders, who couldn't afford the current ones. Not clear if there's a huge market for home PCR or similar (many DNA-active chemicals need to be used under controlled conditions because mucking with your DNA causes cancer -not sure about PCR specifically)

From a physical point of view, I wonder if energy- transfer thermal cyclers could be replaced by adiabatic compression, which is likely much faster. Depends how well these enzymes work at high pressure. Could be problematic at a process level though.