One of the biggest problems is that every 150km or so you have to convert light into electrons and back into light. Latency over long distances isn't limited by the speed of light, but rather this inefficient decision making via election.

The actual story of the article really is just: https://en.wikipedia.org/wiki/Photonic_integrated_circuit

The ability to just keep everything as light and be able to process the 'headers' and make routing decisions is really the magic we need.

Alas, this is bleeding edge stuff by scientists and we're far from commercialization.

And now we can start being even more inefficient and make all our electron apps gigabytes in size because we can transfer them so fast! Yaih!