So far as I'm aware, the big distinction right now is whether you want dsDNA or oligos.
If you often want lots of oligos, then getting your own synthesizer likely makes sense. Run the numbers and see what the time for return on investment will be and you'll see if it works for you financially.
If you want gene-scale dsDNA, you're currently generally better off getting a specialist company involved (or a foundry, if you've got one that you can access). Not only do they have economies of scale, but they have a lot of experience figuring out how to handle difficult sequences and assemblies. If you take it all in house you might appear to save money but actually end up losing significant money and time when you count in the time and frustration that your lab ends up spending to go from oligos to genes.
If you've got particular expertise or highly specialized needs, it's a whole different story, but that doesn't sound like the case based on your question.
Note that there are a number of desktop synthesis companies that are working to make this answer obsolete right now, and it may well change quickly in the future.