If the numbers are correct. At 14/16nm we've reached the point where reducing transistor size doesn't reduce cost of device. (Hence development of planar full depleted SOI (FD-SOI) based solutions like GloFo 22X) /SemiWiki (not related to AMD or GPUs specifically) Total addressable market (TAM) for GPUs is ~$3Bn (ref: AMD's own estimate) but AMD's current market share is only ~18% and in Q315 it made ~$424M (-181M). At the 20% marketshare stated it would bring in $600M of $3Bn TAM, which is not enough for AMD given that in Q314 it made $781M and still posted a (small) loss. It needs to make around ~$800M to break even (ignoring that it laid off more staff recently/restructured GPU division) = ~27% marketshare. Therefore if AMD can't claw back at least ~9% marketshare and almost double its revenue (that's just break even, not including investment in next gen) its Polaris core could be its last, as cost of developing on leading node (10nm will follow 'quick') will only increase into ~$100Ms for each GPU core, yet, it cannot avoid using a leading node otherwise it will become non-competitive overnight. AMD has to knock it out the park in 2016. EDIT: If 18% gives a $2.355Bn market then AMD's uphills struggle is even more challenging: 34% marketshare required to break even = AMD needs to achieve an 89% growth in GPU market this year.