This printer would work extremely well for MEMS devices since the complex structures such sensors can now just be printed rather than deposited and etched over and over again in a microchip fab.
I'm not sure how printing MEMs devices serially is going to be faster than parallel mass production on 12" or 18" silicon wafers. Printing them is analogous to laboriously machining a part in a CNC mill compared to stamping in a forge. Photolithography and etching are pretty fast processes. Well, etching can be slow but it can be done very well in parallel to multiple large wafers at once so per-device it's fast. Doing the printing as a prototype for a standard MEMs process production run won't work well since the material properties would be different.
And you still need to connect your MEMs devices to a circuit, so now you have to do a tricky hybrid integration process to pick up your tiny polymer MEMs devices and connect them to a chip and package your now non-planar device. Plus you need to be able to selectively metallize some of your surface for many MEMs applications - not sure how you do that given that stereolithography "printing" works on photohardening polymers not metals.
Right now it can take weeks to make complete microchip with the current fabrication methods. The fabrication size of this printer isn't that great however since most of what is seen in the TFA looks to be around 100 nanometers compared to the 28 nanometers a modern fab can make. However, it would be great to have for rapid prototypes of processors or be used to make devices that fabricate well at large sizes like flash memory.
It's a big leap going from hardening a polymer to printing full complex semiconductor circuits with dielectrics and metal interconnect. Unless you're just thinking of using this stereolithography process to replace the standard mask-based planar photolithography in the foundry, which might be a valid point if the stereolithography is faster or cheaper than electron-beam lithography or ordering a mask of the dimensions that this machine is actually capable of. Right now e-beam lithography can do this but it's slow and expensive.
For something like this to be applied to semiconductor processing another thought would be construction of stamps for nano-imprint lithography. Printing them might be cheaper or faster than the standard techniques of e-beam or optical lithography and etching at least for short runs.