> "Hundreds of cities and dozens of countries, fine for living but previously entirely unsuitable for a place of business, will blink from red to green on the map once Starlink goes live." this reads like a marketing piece rather than actual analysis. Throughput of a single satellite is 20Gbps. The finished constellation if I recall correctly is supposed to be about 12k satellites. So about ~250.000Gbps. So assuming…
On the other hand, it's far worse than "20G per satellite." Bandwidth is limited by available RF spectrum, and that's not magically gonna increase with satellite count. Best thing they can try is limit the coverage area of each radio as much as possible, but there's limits to that and now you have an onslaught of handovers since the satellites are moving quite fast.
I honestly am too lazy to do the math, but my gut says this may work great for larger areas with lower population density (let's say the Amazon, U.S. Midwest, Australia, Sibiria, etc.) but it's not gonna replace high-density broadband access.
Also, just for comparison, current optical network technology (in the market, not research) puts 50G on a single wavelength on an optical fiber. There's 80 wavelengths in a medium-density DWDM setup, so that's 4T ... on a single strand. A low-density cable has 12 cores (that's what you deploy on the outskirts of your network), so, uh, 48 terabit/s with current technology. Infrastructure cables are commonly 144 cores, that's 576 terabit/s...
(These are obviously "maximum" numbers, you rarely see all 80 wavelengths or all 144 strands in use, and there's gonna be slower links like 10G occupying some wavelengths too.)