I really doubt Vega delay* is related to HBM2. It's been in production for more than a year (P100). So for sure there are more than ~30.000 stacks available. It is just improvement over HBM1, not completely new technology, so it's not so harder or more expensive to make. HBM has been in consumer products for 2 years now. It is not new technology any more, from manufacturing viewpoint
I also don't think it's a bad decesion to go with HBM2 instead of GDRR5X or GDDR6. If we go back to this comparation
http://electroiq.com/insights-from-leading-edge/wp-content/uploads/sites/4/2016/03/Sys-plus-1.jpg
it is only ~40$ more expensive to have it instead of GDDR5. Maybe even less at this moment. Which is totaly ok for $400+ cards. So yes, earnings will be little bit lower, but you can have better performance with smaller amount of memory (with HBCC), lower board TDP (10-20W), smaller cards (SFF), new technology (useful for marketing). And of course if/when HBM becomes as cheap as GDDR (and there will be some low-cost stacks soon), you could use it in mid-range products.
*In August 2016 AMD anounnced Vega will be launched in H1/2017. Which means Q2/2017, otherwise, they would say Q1