There's no direct labour. But there's still labour. People have to program the machines, and then keep the machines stocked and running, and service the machines.

At the moment that's skilled work and a small number of US workers will gain at the cost of a bunch of low-paid workers in poor countries. But it won't be long before that work becomes semi-skilled (see also sub-contract electronic engineering) and is done in Poland / China / where ever, for less than it can be done in the US.

Importation from China is not free, but I can still buy a human-made t-shirt imported from China for a couple of dollars.

Intel is a perfectly good example of how robotic manufacturing works very well and scales.

If we can use robotics to mass assemble something as complex as a couple nanometer billion transistor chip, we can most definitely bring that to garments with some effort and scale it.

Intel can pump out a billion processors with probably a fraction of the employee count they have now (they have 100k total, but they're not all needed to just produce processors). That's intense efficiency.

Obviously the Intel example is a fairly ideal robotic manufacturing scenario. But the point is, that kind of ability will expand outward to all aspects of manufacturing as robotics leaps forward.

Actually, last I heard Intel had outsourced the actual assembly of their chips to places like Malaysia with low labour costs because that's still not entirely automated. The only thing that they do in high-labour-cost countries is patterning and doping the wafers to create the silicon die inside the chips. That's both a lot easier to automate than sewing (no moving parts) and a lot more important to automate because, due to the small scale, they need to keep out contamination by dust which is a lot easier if few humans are involved.