It appears that IBM and its alliance will adopt the bulk 32/28nm, 22/20nm and beyond, abandoning FDSOI technology. IBM was successful in development of partially depleted (PD) SOI for 165nm, 95nm 65nm technology nodes comparable or superior to Bulk in some area. Such success has helped attract a number of international semiconductor and design companies to form alliances and SOI industry consortium. However, such success has not materialized in development of fully depleted (FD) SOI technology for 45nm, 32nm and 22nm. IBM has been a strong advocate of FDSOI technology for Bulk versus FDSOI battle for a long time and now finally it comes to end. I often wander why it took so long for IBM to make the final decision to adopt Bulk.
Intel is a big winner in Bulk technology development race because 22/20nm Bulk with highK/metal gate and gat-last approach are all Intel invention. Intel is far ahead of its rivals in high volume manufacturing and high yield, and reliability as learned from 45nm, 32nm and 22nm manufacturing. However, as others pointed out, power will be the major issue for Intel’s 32nm and 22nm. Based on the published data for 45nm and 32nm, the 32nm device shows significantly higher DIBL(drain induced barrier lowering) effect. It means lower Vt and higher leakage current. The low Vt increase transistor performance while higher power consumption. Also, transconductance degradation not seen in 45nm is observed in 32nm. Strangely, Intel has not published 22nm transistor transfer characteristics even though its 22nm will be high volume manufacturing this year (2nd oater). Intel may not want to show transconductance degradation and higher DIBL effects seen in 32nm.