While the timing of the event comes as a surprise, the fact that IBM has decided to unload its technically excellent but unprofitable semiconductor manufacturing operation does not, nor does its choice of Globalfoundries, with whom it has had a longstanding relationship.
Strategically, this deal should have little if any impact on IBM's remaining Power and Mainframe customers. Supporting this thesis is the fact that this is not a sale – IBM is paying Globalfoundries (GF) to take it off their hands. A substantial part of that payment is almost certainly intended to be applied to subsidize the continued production of IBM's proprietary semiconductor designs. These designs use their own variant on industry-standard CMOS process to generally produce processors that are biased toward very high clock speeds. The payment is also to compensate for the fact that these products will have very low volumes for GF. The agreement specifies that GF will be the exclusive supplier to IBM for the next decade, which will take them through 22nm, 14 nm and to 10 nm and possibly miniarturization beyond 10 nm. Additional support for the thesis that this will not impair IBM's competitive positions comes from IBM's continued investments in semiconductor and systems R&D. At the same time that IBM is divesting itself of the manufacturing facilities, it has gone to great lengths to reinforce its multi-billion dollar commitment to ongoing semiconductor and system design efforts, ranging from chip and SOC design to basic semiconductor materials research.
For Globalfoundries, especially considering the included subsidies, the deal is a winner. They get in excess of $2B incremental revenues, an excellent array of CMOS process technologies and other IP, valuable manufacturing capacity, and a strong position as the dominant supplier to the OpenPower ecosystem if it takes off.
This move is completely consistent with IBM's ongoing strategic re-positioning of itself as a supplier of high-value software, cloud-based services and integration, and its operational impact on customers, and competitors should be essentially zero, assuming a modicum of good faith and competent execution.
For current or prospective Power or Mainframe customers, or customers who rely on other IBM custom silicon, this announcement can be viewed as more of a strategic cost-shifting arrangement, and should have no impact on short or long-term commitments to IBM. Competitors will likely think about using this as an additional wedge to convince prospects that IBM is exiting the systems business, but I think this event is essentially neutral across the board except for GF, who seems to have gotten a pretty good deal as IBM's partner.