Koub wrote:
Over short distances, bots are the thing. Over base-ish distances, when high throughput is needed, bots are a very inefficient solution. You won't replace a 6 or 8 blue belt main bus with bots in a big base.
I'm not really finding this to be the case, they seem fine over base-wide distances even (obviously not mine->base). For instance, I've got on average 3000 bots active at a time. With about 20 roboports, 500 solar panels and accumulators to support them. So at a rough guess, 80k iron/copper and 20k petroleum.
Now with that I've made around 500 level 3 modules (I think over a thousand), 10k laser turrets, 1000s of accumulators and solar panels to support everything having these modules. I'd estimate the amount of used materials being roughly 3M Iron/copper, 1M petroleum, probably more when accounting for miscellaneous garbage (edit: oh right, the whole tech tree too xD)
Having transport infrastructure cost less than 5% of production as opposed to less than 1% sounds a fair price to pay to completely eliminate a main bus and simply use the same very compact beacon/assembler blueprint for everything, anywhere. That particular blueprint feeds enough speed modules into the production assemblers, that they probably even make up for the extra resources in increased production efficiency.
I don't even feed copper into circuit assemblers or feed furnaces with belts. Thanks to the inserter bonus mentioned in the OP, a high speed circuit assembler can even work at full capacity in a tight space, using only two input fast inserters and one fast output.
I guess it ruins the fun a bit, but just pointing out that even though it's like 10 times less efficienct than belts, the transport network itself is cheap, so it may even be more efficient if you're heavy on modules and beacons. To say absolutely nothing about the massive time savings
Also, it looks super cool when you have a train come in and hundreds of bots pour out of a roboport halfway across your base xD