If you frequently use rapid prototyping services, such as Protolabs or Quickparts for 3D printed parts, you may at times consider whether you should buy your own machine and bring the work in-house. Some things you may consider when making a decision are:
Saving Time. You could save some time if you had your own machine. The primary benefit of rapid prototyping is the huge improvement in lead times compared to traditional manufacturing, allowing quick multiple iterations that improve the final design of a part or product. While rapid CNC machined parts can still be gotten overnight from services like Protolabs, the lead times for 3D printed parts are typically at least three days plus shipping. Three days is fast, but if you are working on a project where you know there will be several iterations before arriving at a final design, three days x the number of iterations can seriously slow down the project, so it would be beneficial to shorten the lead time with your own in-house machine.
Material, quality, and build volume. The machines that the big service providers use are state of the art, high speed, expensive, and produce great parts consistently. Many of the parts made can be fully functional for the intended application because of the material options available. The service providers have many different types of machines, so many materials and build volumes are available. Even metal 3D printed parts are available.
Printer cost. If you need an industrial grade and large build volume or high speed 3D printer, they are expensive. They typically can cost from tens of thousands, to hundreds of thousands of dollars and up. However, for some there could still be cost savings to owning an industrial grade printer depending on which machine you buy, and whether you need to make sufficient quantities of parts that justify the initial capital cost. Don't forget there will be ongoing material expenses and operational expenses. Someone will need to run and maintain the machine, offload parts, possibly post process the parts, clean up as well as keep raw materials in supply. If you need more than one type of machine, buying multiple industrial grade machines is even harder to justify.
For many companies, it would not make sense to buy an expensive industrial grade machine comparable to what the service providers have just to do in house prototypes. There is a middle alternative that works great for me and that I recommend for most companies. Buy a low cost (~$1000 to $4000) machine for quick, several hour or overnight iterations when you only need a few parts and can live with more limited material and size options. You won't be able to make large quantities of parts quickly and you won't have all the material and size options that a 3D print service provides, but it often allows you to get through that 1st or 2nd iteration more quickly. Dimensional accuracy can be as good as commercially printed parts when set up right. Later, if needed, use the external services for the final, fully functional parts with vastly more material and size options, and higher quantities of parts.
If this makes sense to you, then the question is which type of machine should you buy? There are two main types in the ~$4k and way under range, FDM and SLA.
FDM, or Fused Deposition Modeling, works by heating a plastic filament and extruding it through a nozzle. There are hundreds of these inexpensive machines available that can produce very good quality plastic parts from PLA, ABS, Nylon and many other materials and in many colors. There are too many FDM suppliers to mention, but the printers I have are from Sindoh, and Prusa. I'm very happy with both. MAKE magazine has a free comparison report of many low cost FDM printers here.
SLA, or Stereolithography printers work by projecting a laser or light onto a bath of photopolymer, solidifying one layer at a time. The part finish and dimensional quality from SLA machines is excellent. SLA can produce more highly detailed parts than FDM. SLA has fewer material and color options compared to FDM. Another drawback to the low cost SLA machines is the small build volumes. They are too small for many of the parts that I need to make. One of the best known suppliers of low cost SLA printers is Formlabs. I have part samples from them and they are very impressive. For a review of several low cost SLA printers, check out the All3dP.com website.
In summary, use FDM if
Building low-cost models
Surface finish and very fine details are not critical (but can still be very good)
You need a larger variety of colors and material properties
You need larger parts than the SLA printers provide
You don't mind learning about and tweaking some settings to get the best prints. It will be different with different materials.
Use SLA if
Extremely higher intricate details and a very smooth surface finish are important
You need more accurate prints than FDM can provide
You create molds for casting to facilitate mass-production of jewelry or toys
You only need small parts
You don't need much variety of material properties or colors
You don't mind a more expensive printer and material cost
A more detailed comparison of the two types can be found here and here. I've been using this two pronged approach to prototyping for almost two years now. I have two inexpensive but good quality FDM machines for making initial prototypes and iterations very quickly. Once I got over the learning curve, the operation is simple and not very time consuming. They produce accurate parts to tolerances as good as the laser sintered parts from the service providers, although of limited quantity and size. I still use the 3D print services when I need large quantities, a smoother finish, or stronger parts made by SLS for more aggressive use. This two-pronged approach gives me the best of both worlds. I can get quick 1st and 2nd iteration samples in hours, not several days, and still access all the benefits of the 3D print services when I need something more.