Well I have worked in companies making physical goods (shipping millions every year), and you just don't buy a team from one day to another. It's not just about production, it's about supply chain, it's about quality management and quality assurance, regulations for sale in different countries/states, market research, product qualification to improve the design, actual customer support, vendors, distribution, etc...
It's a whole new area of skills/functions you don't have, and that does not mix very well with a software-only company mentality. I think it's going to be very tough for them to make something solid with Facebook.
EDIT: They would have been better prepared for scaling with a company like Apple (whose main business is already making hardware), Microsoft or Google for that matter. Many people say Valve could have been a good choice, but Valve has said several times they don't want to go into the hardware business (and won't make Steam boxes themselves).
The point is that, while Facebook may be able to do hardware, servers is not a good example.
Disclosure: my team at Sanmina has been responsible for delivering the test automation & integration software for the rack assembly & test operation for Facebook.
For that matter, Google itself doesn't build anything at scale, either. Why would they? They outsource all their high-vol manufacturing to the same EMS companies everyone else uses (Quanta, Compal, Foxconn, Flextronics, Sanmina, Celestica, Jabil, Pegatron, Benchmark, Plexus and a handful of others).