Here is a dumb and expensive plan you want 3000 cans of soda. Your plan to buy 3000 cans of soda retail from an airport store at $5 per can. The airport store only has 20 cans. They do not sell in bulk. It would take doing to 150 stores to get all of the drinks you want. This would cost $15000. Paying the most expensive retail is stupid. This would be especially stupid if you were a soda bottling company. Not only do you sell 30 can packs to Walmart at $7 cost to you and Walmart sells it at $15 or so for 30 pack cans. You need about 100 packs at your cost of $7 or about $700. But you are a bottler, you will make gallon containers at $1 each. Each gallon is a 30 can equivalent. 100 of those would be $100. You can make make high pressure tanks and do the whole thing for $10.

Why did I go through this ? The moronic $180 billion cost estimates for putting a gigawatt of AI into space.

The Bank of America estimate can involve the $30 billion for chips for a gigawatt. Whatever that price estimate is does not matter because those chip GPU and memory are the same on earth or on in space.

The it is buildings, gas turbines and gas operating costs and cooling versus satellite bodies and solar panels. Starlink V2 mini cost about $300k to 500K each. Starlink V3 at 30 kilowatt are about $1 millon each in mass production. Starlink V4 or AI1 at 150-250 kilowatts will be about $2 million each in mass production but could eventually get back toward $1 million each. This is without the chips. 250 kilowatt satellite then we need 4000 to get to a 1 gigawatt. $800 million for 4000 satellite bodies for one gigawatt. Once you know how many satellite bodies and solar you need you can calculate in your head based upon your estimate for the satellite cost.

Falcon 9 costs $60 million to someone buying the ride from SpaceX. It costs Space $12 million. They reuse 80% of Falcon 9. Falcon 9s cost them $40 million new but they are using $32 million of it 30 times and only replacing about $8 million each launch.

SpaceX is flying on the order of 160 Falcon missions a year NOT 3000 launches. Most of those are internal Starlink plus existing customers. New commercial Falcon 9 bookings past 2028 have been getting turned away. About four-fifths of the 2026 Falcon manifest is internal. There are not 3,000 free Falcon 9s sitting in a lot.

Someone who came up with the 3000 can $15000 plan would be called a moron. Bank of America and others pricing out 3000 Falcon 9 launches for a gigawatt at $180 billion from retail Falcon 9 pricing are morons.