About four years ago, Multibeam was in trouble. Its focus on e-beam as an alternative approach looked weak against an industry doubling down on EUV.
Investors wanted out of Multibeam. As a board member, Lam saw potential for the technology and organized a buy-out of the startup's assets although he was still unsure just what to do with them.
"I knew I had to find a smaller, more focused application as the market entry point for e-beam, but nothing had solidified or crystallized yet."
"Shortly after that I heard Intel's Yan Borodovsky give a talk [on using e-beam as a complementary technology] and the light bulb came on," Lam said. "In September 2010 I tried out the idea in a talk at a mask conference and I got a standing ovation" from mask makers happy to hear about e-beam used in a way that did not eliminate them from the process, he said.
Lam's aha! moment was predicated on another insight. Chip makers had quietly moved from etching lines in chips in two dimensions to just one dimension, opening the door for e-beam as a new etching tool.
Evidence from third-party analysis of Intel chips showed the CPU maker shifted from 2D etching at 65nm (above) to 1D etching with its 45nm process in 2007 (below). Others followed later.