We use orders of magnitude here, so powers of ten.
Elon Musk presented the Terafab idea yesterday, here.
The idea is to manufacture so many chips per year that they would use a terawatt of power, or 10^12 W. US electricity usage is 0.5 terawatts.
A Starship V4 is expected to launch 300 tons to orbit, or 10^5 kg.
We can assume a satellite to produce and use about 100 W/kg or 10^2 W/kg. (This is debatable.)
From these we can calculate that to launch a terawatt of satellites to space would be about 10^10 kg, and thus take about 10^5 Starship flights.
Launching
A Starship uses about 10^6 kg of methane.
Methane has about 50 MJ/kg energy content, we can round this to 10^7 J/kg.
This means a Starship uses about 10^13 J of energy.
A single Starship’s payload would produce 10^7 W of power.
3 years is 10^8 seconds so over that period, it would generate 10^15 J. So the solar power produced is about 100 times the amount of energy that went into the launch. Again, very rough order of magnitude calculations.
The Starships used to launch the terawatt of satellites would use 10^11 kg of methane.
World methane production is 10^12 kg per year. Thus about 10% of that would go to launching Starships. Global air travel uses about 10^11 kg of fuel per year so this would be similar to that. It would show up in global emissions, but it can be argued it would be offset by lesser data center power usage on earth.
More realistic nearer term scenarios
Starship has already done some test flights, it’s constantly improved and new units and models are constantly being built, multiple per year.
If we assume one Starship being built per month and one flying once per week, then after one year, the capacity will be 12 starships and 50 flights per year, meaning 600 flights per year. At 300 tons this will be 180 000 tons or about 10^8 kg. The satellites doing computation, at 100 W/kg this would be 10^10 W or 10 gigawatts of power.
It’s conceivable for example that these 12 starships could be built during 2027 and once-per-week flights could operate in 2028. Again, these are just example numbers to get a feel what it would mean.
This 10 gigawatts would be a significant player already in computation.
The methane usage would be roughly about a million tons or 10^9 kg, or a thousandth of world production.
Assuming constant building rate and flying rate per vehicle, after that it would be 20 gigawatts per year, then 30 gigawatts and so on.