Satellite communication won't "drastically cut down delays". On the contrary - it will increase it, as the signal will have to physically go through a much longer geostationary orbit detour between two points on the earth. Satellites offer you additional bandwidth for somewhat lower costs (compared to establishing fiber optic land line), bigger delay, and penalizing partial dependence on atmospheric conditions.
Why stop at 2 gigabits/sec? Why not go all the way up to 10g, 40g, or even 100g? It's not like rain or fog or other atmospheric effects will cause much problem for free space optics or tighter band RF (e.g. 60-90ghz millimeter wave) communication in space, so was it a physical or budget limitation? At these higher throughputs, the same relay network might be usable for other things in the future, such as streaming terapixel video...
Atmospheric conditions actually have quite a bit of effect on satellite communications.
As far as the link speed between nodes in space well 1.8 gigabits is fast even for fiber optics, when you add the distance between the 2 nodes and the fact that at least one of them is moving at very high speed when counting for beam dispersion and phase shift due to movement 1.8 gigabit seems very impressive (this is considerably faster than pretty much any point to point IR link here on the ground[1]).
That said satellite internet doesn't makes much sense at least not how the internet infrastructure is setup today going to space means that you add at the least 80-70,000KM to your round trip no matter what kind of bandwidth you are getting in return (which in every case will be very asymmetrical) it's not worth adding sometimes seconds worth of latency to the round trip of your packets, the internet today simply can't really work like this.
I would've said that in the future you might get a super smart home hub that can negotiate a single internet address (probably not going to be standard TCP/IP at this point) over multiple links (yes I'm aware that multi-link PTP has existed for decades) and automatically route the traffic based on it's nature; uplink can go over fiber/DSL; downlink for something like say netflix can go through your sat link; web browsing including streaming can start over fiber/dsl for a more responsive experience and if you are streaming say YouTube just offload it to your sat link transparently; applications that require low latency such as games and video/voice comms will always go over your fiber/dsl.
The only problems with this is that it seems to be too expensive and complicated and by the time you'll have good enough coverage of satellite internet the adoption of fiber will increase and since you can push upwards of 1gig these days over copper DSLlike technologies might actually catch up also.
In the '80s we built circuit boards for a company called Laser Communications in Lancaster Pennsylvania. They could transmit fast (at the time) data over 2-3 miles pretty easily, but water particles in the air (rain, fog, etc) caused enough refraction that the links would go down. I'm curious to know how they solved that problem when going through so much atmosphere (I'm not buying the 50% to 100% without an explanation).
The laser links are between different satellites not between the base station and the sattlelite.
Terrestrial FSO is still pretty much limited to about 2500-3000m, no one is going to be using laser to talk to satellites from the ground not as long as the earth has an atmosphere.
It communicates with other satellites by laser. Ground communication is over Ka-band radio. The 50-100% figure is simple geometry. A satellite with an orbit that is always in view will have 100% coverage. Other satellites will have at least 50%.
