We use both types of long-distance connectivity techniques, and there are good reasons for doing so.
I recently heard yet another smug yet semi-ignorant journalist glibly talking about how ubiquitous long-distance links have become. He clearly had no idea of the practical differences between undersea optical cables versus satellite-based links, as he mixed and confused the key attributes of each. His "ignorance is bliss" mindset, however, got me to thinking about these two links and their dominant features.
Let's put aside costs for now. Obviously that's important, but it's a very different topic than technical attributes, since it involves hard-to-assess initial costs and operational costs, as well as capacity all of which affect how you figure cost/capacity.
Fiber optical links are best known for their consistent performance. They are immune to EMI/RFI, including solar- and space-sourced interference, which can be a real problem over long distances. (Buried or undersea copper links are somewhat immune, but fiber is really immune.)
Also, they have relatively low propagation delay and latency, since they are essentially point-to-point links with data traveling at roughly 90% of the speed of light, so representstive trans-ocean path of 5000 km (3000 miles) takes about 20 msec (not counting any delays through amplifiers and repeaters). This is not only important for overall system performance, throughput, and link acknowledgment protocols, but vital in voice conversations.
But, they take a lot of work to install and maintain, especially if a cable is snagged, although techniques for doing this have improved greatly. And the fiber end-points are fixed, so you have to implement additional links to get beyond those points.
In contrast, satellite links have easier, albeit limited, maintenance possibilities. If the problem is at the ground station, it can be found and fixed readily; if the problem is at the satellite, it generally can't (although there have been some amazing cases of ground-initiated directives which re-routed signals around defective subsystems within a satellite).
Once a satellite is in orbit, access can be easier and even mobile. Given the right satellite and ground station, you can even set up a portable access point from almost anywhere on the planet, though it can be a problem or even impossible as you move towards the Earth's poles.
If you are sending a large packet of data, the satellite-link uplink/downlink propagation delay of about a quarter of a second for geostationary orbits can be a non-issue at first glance, but it certainly affects the back-and-forth acknowledgement that packets were received with low enough BER for successful recovery, and resultant requests for retransmission.
And finally, a satellite link is subject to both deliberate interference as well as unavoidable galactic EMI, whch can curtail or severely limit its throughput. Extreme EMI can cause internal failures (there are documented cases), and there is also the possibility of collision with some of the other stuff floating through spce, man-made or otherwise.
So, which long-distance link is better? As with most situations, the answer is simple: "it depends". What are your priorities? What are your concerns? It's not a "one solution fits all" world, that's for sure. ♦