Half of that - $7 billion - has been spent by ASML of which $1.9 billion was contributed by Intel, Samsung and TSMC and the rest by other industry sources.
The other $7 billion has been spent over the last 30 years at other companies, government research institutes labs and universities.
But, despite the $14 billion spend, the problem with EUV remains the strength of the source light.
10W is being achieved, 80W is needed. Even 80W only delivers 58wph.
ASML has 1,000 engineers working on the source light which is more than the collective number of engineers working on EUV alternatives multibeam and nanoimprint.
All three technologies will will miss the 10nm node, which leaves immersion as the workhorse tool. Immersion will go to 3nm but may need octuple patterning.
Multiple e-beams are an approach taken by KLA-Tencor, Mapper and Multibeam which are said to have raised about $700 million but will need a billion to have a machine ready by 2018.
CEA-Leti's multibeam consortium, Imagine, is using a Mapper tool which delivers 32nm resolution on 1,300 beams. Next year it is hoped the tool will have 13,000 beams.
Canon Nanotechnologies (CNT) is developing a nanoimprint tool capable of 80 wph at 20nm resolution expected next year. The technology scales to 9nm.
Other options are directed self-assembly (DSA), photoresist dry shrink, self-aligned vias and an invisible hard mask.