The process will be in production in late 2018, delivering gate pitches as small as 30nm initially using only today’s optical lithography.
Separately, the company will support a new embedded MRAM in sub-Gbit densities starting in 2018 for chips made in its 22nm fully depleted silicon-on-insulator (FD-SOI) process. The memory technology, licensed from Everspin Technologies, will provide faster write speeds as well as lower power consumption and die size than current variants of embedded flash.
The news boosts the foundry's competitive position against larger rivals such as TSMC. Both the 7nm and FD-SOI technologies are largely the heritage of IBM’s semiconductor group that officially merged with Globalfoundries in July 2015.
“For the first time we have a differentiated road map…we’re taking an independent view of where the market is going and developing technology for it,” said chief executive Sanjay Jha, a former Qualcomm and Motorola executive who took the reins at the foundry less than three years ago.
“I was TSMC’s biggest customer for a long time, we are not exactly where they are but the progress we have made is remarkable and I think customers are beginning to notice,” Jha said in a meeting Wednesday with press and analyst.
He noted the foundry is still only about half the way to bringing uniformity across its fabs that came from AMD, IBM and Chartered Semiconductor. For example, some things as simple as file-naming conventions still vary among factories.
Among Globalfoundries’ other rivals are much larger companies with leading-edge technology but smaller foundry divisions.