If you plan rendering with Cycles Render engine then yes, it is capable of using most NVidia based cards that support CUDA processing for acceleration.
The 1080 being a new top of the line graphics card is mostly expected to perform well under normal circumstances. Your real world mileage may vary however, depending on a number of factors like operating system and hardware.
You may even benefit from a multiple card setup in a non SLI configuration, if you have the budget for it.
Have in mind however that the 1080 card being a very new addition to the product portfolio is not currently well supported. According to this benchmark it's performance is still sub par and unoptimized.
Work is currently being done to improve the situation though, so expect future versions to improve in that area.
If you plan on rendering exclusively using GPU acceleration you won't need to spend so much on CPU and system memory, for rendering at least, since it will be the graphics card doing the heavy lifting, those will have a lesser effect if you are using only GPU based acceleration only.
They will however benefit other areas without GPU acceleration like opening large files, compute animations and simulations, and general Blender responsiveness, so they may be important if you also plan to use the computer as a workstation.
Before you buy make sure you won't need any of the unsupported feature of GPU rendering before investing in it, or you may find yourself unable to take full advantage of the hardware you bought.
SLI (short for Scalable Link Interface) is the NVidia proprietary system for using multiple graphics cards in one system. It basically creates a direct link between the two graphics cards making the operating system seem them as one, and thus seamlessly use them for games and applications potentially doubling the performance gains.
That is the theory at least, in practice there are some limitations, and system is quite inefficient so real world benefits are more far bellow double the performance, in some corner cases are even worse than a single card.
For Blender however there is absolutely no benefit in an SLI configuration, the cards should be installed as two discrete independent cards in the system without an SLI dongle. This will not give any performance benefits for regular operating system and applications, but Blender can use the two discrete cards as two independent processing units virtually doubling the precision, (it is my current setup in my home computer, I own a and older GTX580 and a new GTX980Ti)
As for memory constraints I work in Archviz and my GTX580 only has 1.5Gb memory and it has served me well over the years. Obviously some more complex scenes or when using HDR image based lighting that memory will be insufficient, but my 980ti as 6Gb and to date I have not found any project that could not fit in it's memory. The 1080 having 8Gb (I think) will be more than enough for most projects, I'd say, and probably faster than even the most expensive CPUs, so I'd say stick with GPU rendering if you can.
Obviously I don't know what type of work you do and how complex your scenes may be so I'd recommend you to make thorough testing with your current work. Tre rendering a few scenes and monitor how much memory is currently required for your most complex jobs. From there you should be able to make conclusions if GPU rendering will be enough for you or not.