What about contingency?
Whenever I read stories about undersea cables I am reminded that the only institutions equipped to sever them are navies, all of which are operated by governments, none of which can be trusted in the slightest, especially in times of warfare. And 'warfare' might cover more than missiles, these days, what with the rise of asymmetric ops and cyber-tactics. (One might also wonder about the security of landing zones and dry infrastructure: perhaps a radicalised nutjob could perform prodigies of economic harm with a crowbar and a can of petrol ...?)
Perhaps this seems like a paranoid viewpoint. But I would suggest that at the very least, major businesses should think hard about figuring out what happens when they lose, say, their transatlantic cables, or those to the Far East. If you were to map high-bandwidth seafloor cabling you might be surprised at how so few wires carry such vast torrents of information—I'll stick my neck out and make a wild guess that if you charted the routes, the nodes, the traffic and the operational criticality of the latter*¹, you would find some horrifying dependencies. This spectacle, visualised, could be extremely sobering.
It is extremely difficult to maintain ultra-high bandwidths without cable—and of course, satellites can be disabled or even shot down—but I wonder if we are devoting enough truly "innovative"*² thinking to other means of securely and reliably moving data wirelessly. Satellite TV and various web wheezes would suggest we are not exactly asleep on this, but who is actually looking at practical wirwless fallbacks if oceanic cables are lost? Some very clever work has been done on degradation-resistant encoding-and-encryption systems for potentially unreliable pipes, but more can be done. What could be achieved with state of the art laser tech, and satellites stationed in high defensive orbits, or even at Lagrange points? Latency might be high, but that is easier to cater for than crippling loss of overall bandwidth.
I'm not expert on this, though I've worked on the degradation-resistant stuff, so I'd be interested to know what readers think.
*¹ 'operational criticality':: clumsy phrase, and not an easy thing to estimate, I think: but basically I mean "Assess how much direct economic and other consequential damage (e.g. reputational; data breach; fractured tactical/strategic decision-making loops; loss of competitive advantage; etc) might accrue from sustained interruption of the data pipe"
*² 'innovative':: as compared with today's use of "innovative" to mean "incremental features no one asked for or needs"