I'm going to start with an incredibly cheap $15 an hour. That's 31k. Then add in whatever overhead per employee such as HR and management costs. Add in any benefits. Let's cap it at a cheap 50k. Keep in mind there will come a point where human infrastructure such as bathrooms will become optional. Which means certain plumbing will become optional for certain buildings. There are huge swaths of expense that can be removed once people are out of the picture in a manufacturing environment. If it gets more expensive per employee then the robots become more attractive. If the operation is capable of 24x7 operation then that means a robot can replace three employees. But let's knock that down to two and give the robot a shift to be maintained.

At this point the value of the robot is 100k per year. The robot will not bitch at the boss or spread anti-company gossip or attempt to unionize, or sue the company, etc etc etc. There is a whole lot of benefit to the robot, far beyond the money savings, but let's start there.

What is the reasonable expected payback time for this type of factory floor automation? I see between 6 months and 5 years so I'm just going to make up 2 years. That would be 200k.

The moment the robot is cheaper than 200k, it becomes reasonable to replace the person with it from a simple cost benefit analysis based on these assumptions. Since you see I went cheap if the people are any more expensive then the robots can be more expensive or will have a quicker payback.

A factory floor worker in a auto manufacturing plant costs a hell of a lot more than 50k per year. People digging in mines make a hell of a lot more than 50k a year. Either way the robots I see capable of factory automation don't look like they cost more than 200k to build. When they go full factory level automation I say they cost 50k and are sold for 100K for a healthy profit margin while still being very attractive to the final consumer companies.

Also, the robots don't have to be general purpose. They just have to be cost-effective enough replace people and the production environment and the company will be willing to modify the environment to accommodate the robots if that meant they could get cheaper robots. Legs and feet and a highly specialized balancing system is a lot more expensive than treads. So if we don't need the thing to walk steps then it can have the much cheaper treads. If the job is a person sitting at a single location and reaching forward with their hands and then swiveling, then there doesn't need to be any mobility at all. No mobility requirement means no battery packs and full-time wall power. Much cheaper.

The reason all of this is possible now is because we didn't have a vision driven manually dexterous system. The AI brings it all together. Now let's see what the bare minimum of human-like parts are required to replace a factory worker on an assembly line. Eyes arms hands fingers attached on a swiveling pole. There will be lots of those type of balancing decisions to keep the costs down.

Iron may be expensive relative to silicon, but it's far cheaper than people.