I agree pundits that predict Intel will eventually stop developing IPF and transfer development to HP.
If you remove HPs servers from the total shipped, there were only 1,856 servers shipped by all other vendors combined. Further, if you remove IBM, there were only 743 servers shipped by all other vendors combined. Why is this important? Well,
1. Intel doesn't earn any profits (other than foundry profits) on chips destined for HP (70% of the market).
2. If IPF ever really started to threaten Power, IBM would likely pull the plug on it (why would they risk a multi-billion investment to "endorse" a rival that generates 1,100+ server sales per quarter).
3. Intel is currently earning profits on chips going into 1856+ servers per quarter. At 8 chips per server (chip guy estimate from a while ago) and an average chip price of $1,400 (pricewatch), Intel is generating all of $83+ million per year in total IPF sales. If they lost IBM, they would be down to $33 million in total IPF sales per year.
This is total revenue for IPF before cost of goods sold is taken into account and overhead for on going development. How long can they keep investing?
4. Dell doesn't want IPF to succeed as they will forever be at a competitive disadvantage to HP, their largest competitor.