Especially given the first one laid in 1850 failed within hours http://atlantic-cable.com/stamps/Cableships/indexstc.htm

Me too but then if you think about it you can lay a cable around the biggest part of the earth the equator that's about 40,000km. If you have a geostationary satellite that's about 35,000km. Pretty close but latency is a tiny bit worse for the satellite T=D/V

Equator at 40,075km / 209,855km/s=0.19s (saw a Stack Exchange question about speed of light in optical fibre slows light 30%).

And for satellites 35,000km / 299,792km/s=0.12s but I'm sure air, humidity, big balls of floating water aka clouds, solar radiation just weather in general must make it far worse I've seen mentions of typical latency of 0.25s.

But most cables are not 40,000km long I don't know what the longest cable is of the typical length probably 20,000km max that would give 0.09s latency and for 10,000km 0.05s.

The cables themselves fail surprisingly often (anchors, fauna, earthquakes, material faults...) but most connections have redundant paths (almost every undersea network is bidirectional ring) and can be readily repaired (which is the thing that still surprises me)