'According to IBM, the creation of the new cell was aided by factors including: "Band edge high-K metal gate stacks, transistors with less than 25nm gate lengths, thin spacers, novel co-implants, advanced activation techniques, extremely thin silicide, and damascene copper contacts".' (source - ZDNET)
A bit odd that they would say transistors less than 25nm gate lengths for a 22nm node (which by the way should have transistor gate lengths in the 19-20ish or smaller range)
I think what we have here is a typical IBM PR announcement which will be found to have several holes if the details are ever presented. Also of note, a working SRAM cell is not the same as a working SRAM chip, as Wolfgang somewhat implies in his commentary. From info on the web there is no IBM mention of an actual working SRAM test chip, just a 1 bit (byte?) cell.