There's a very narrow band of sun-synchronous orbit with consistent enough sunlight to make orbital data centers viable. Even with a tight constellation of 500,000 to 1,000,000 satellites packed into that band, it still doesn't make economic sense compared to what terrestrial data centers can do. At 500,000 satellites you're looking at maybe 50 GW of solar power. Then factor in the lifespan of LEO at ~5 years you're looking at hundreds of satellite launches daily to maintain that infrastructure.

Now work is being done for increasing solar access (Starcatcher) and others are working on improving LEO refueling and repair capabilities, but I would say we're decade+ away from establishing any true compute infrastructure in orbit.