You can do phylogenetic analyses on almost anything.  The bigger question is whether the analyses will help you understand anything about the organism you are studying.  If you are looking at SNPs from one species (maybe a plant or animal or fungus) with many chromosomes, the types of information you get are completely different than if you are looking at 180 different species of bacteria with just one chromosome each.

Of course there is a whole range of conditions between these extremes.  But even within one narrow group such as vascular plants, the answer depends on whether the plants you are studying are sampled all over the earth and have evolved in place, or if you are looking at 180 samples from one local population of inter-breeding plants.

You can concatenate all SNPs and treat as a single large locus, and run using standard analyses (e.g. RAxML for ML, MrBayes for Bayesian...), although concatenation methods have been criticized for creating inflated support values, and ignoring potentially multiple independent gene histories in your data... Thus it is possible to create apparently strong support for an "incorrect" topology, or misrepresent the support for a "correct" topology (insofar as there exists an ascertainable "correct" topology). Some also advocate "binning" SNPs/loci and treating them as "suer-loci", but I have no experience with this.

Many methods (such as *BEAST, BEST) for species tree estimation from multiple loci will not be useful for you, as they require the estimation of gene trees, and I assume from your question that you only have SNPs... There are however some new methods, such as SNAPP (implemented in BEAST2) which simultaneously integrate over gene trees while inferring the species tree (see attached Bryant et al. 2012). I have also been using SVDQuartets (implemented in PAUP*- currently in beta), which also uses SNPs using quartet inference... Which does something along the lines of 1) For possible 4-taxon groupings, evaluate which of 3 possible topologies is best supported and 2) Use some agglomerative algorithm to assemble supported quartets into a tree; 3) PAUP* does bootstrapping to assess support (see attached Chifman and Kubatko 2014, and associated publications)...

In my experience (with my particular organisms and my particular data) I am recovering the same topology with SNAPP and SVDQ as with the concatenation method, but with lower supports. YMMV, and of course the appropriateness of these analyses will need to be assessed (as explained above by Brian Foley), but again- that is for you to decide... But yes you can do phylogenetics from genome-wide SNPs, and here I have mentioned a few methods (some more "appropriate" than others) for you to do that. Good luck!

https://mbe.oxfordjournals.org/content/early/2012/04/18/molbev.mss086.full

http://www.ncbi.nlm.nih.gov/pubmed/25104814

many thanks