I think I can safely say that the parts won't be flat to 1000 of an inch (.03mm) - or whatever your tolerance is. However this is something that any existing process can achieve with arbitrary shapes either.
I think this process replaces sand casting, a process that also has many slight deformations. You just make the parts a little big and then machine the important sides to the exact side you need. The downside of sand casting is it often leaves a little bit of sand embedded in the metal which will destroy your tools. If the internal metal quality is as good as sand casting, the ease of doing arbitrary shapes and lack of sand in the metal make it a winner. That you have to have a lathe/milling machine to finish the part is not a change.
If the above is right (it seems reasonable, but I don't know if it is), then beams and and flat stock will continue to be made with existing process. However odd shaped things like an engine blocks it could be a winner.
Engine blocks will almost certainly remain as a cast part, simply due to the production capacity of casting is orders of magnitude higher than additive manufacturing, as well as being relatively cheap. Metal additive manufacturing is currently used in aerospace to reduce complex assemblies of parts into a single piece, or to make small complex shapes that can't be made by a traditional machining or casting process. Additive manufacturing, for the foreseeable future will not replace load bearing parts as well, mainly due to the different processing steps load bearing parts undergo. It will also most likely stay with aerospace for the time being, since the price per part is prohibitive for automotive use currently.
> Parts also shrink up to 15 percent during the debinding and sintering process – but again, that's all automatically managed by the system.
Given this change in volume (which I assume would be highly nonlinear depending on design density and geometry?) I wonder what repeatability is like.
Perhaps for the manufacturing (non prototyping) printer they have a closed loop feedback mechanism to build a few outputs and feed any dimensions out of tolerance back to the input stage.
It might also be possible to model the shrinkage in software and compensate for it automatically, assuming you know what materials you are going to use.
Could be a minor surprise when you check that you have a 200x200mm work area, design a 100x180mm part, and then get told that it won't fit on the build area.
It's not just materials, but the geometry of the part that plays a role in the shrinkage. Though I wonder if some form of finite element analysis could overcome this (that software won't be cheap).
The sintering process is the hard part of printing metal and where a lot of Desktop Metal's core technology is. They basically use traditional heating combined with microwave heating controlled via infrared temperature sensors:
