There are claims that by 2020 we may have tabletop optical supercomputers capable to perform exaflops of operations per second. Are these statement grounded?

https://www.extremetech.com/extreme/187746-by-2020-you-could-have-an-exascale-speed-of-light-optical-computer-on-your-desk

Lab or commercially available? And whose table? (I've seen some big optical tables)

 

In the lab, maybe. It only has to achieve the goal for an instant. No way for something that can be bought and delivered somewhere.

Did you notice that he did not mention how the refraction patterns would be controlled? Or how they could be "combined and decoded"? I can see specific uses but have grave doubts that these things could be used for anything other than quite specific problems requiring individual construction. This is an analog device and not digital though it would be digitally controlled.

 

Using light or superconductors all is limited by the SIZE of the components with which you can work.

 

What he is talking about is a complete system on a chip and the very fact that he quotes Moore's Law should be a clue. If you cannot make transistors much smaller than they are presently getting you certainly are not going to improve on them a whole lot by using entire blocks of liquid crystals through which diffraction patterns can be formed only by turning a transistor on and off.

 

Certainly an interesting idea but the the world is full of good ideas that can't be made to work though the limitations of physics.