SAN JOSE, Calif. We sat down recently with James Plummer, dean of Stanford's school of engineering, for a wide ranging interview on the outlook for CMOS, engineering, education and globalization.
In his role as dean, Plummer helped rally companies including Exxon Mobile, General Electric and others to create a ten-year, $225 million program for research into alternative energy. He is also a widely regarded researcher in semiconductor technology, and currently sits on the Intel Corp. board.
EE Times: What's the outlook for CMOS scaling?
James Plummer: I think we are reaching an inflection point because the costs of a new generation semiconductor factory are just huge numbers. So consolidation in the industry is inevitable.
We will see before much longer only a few players at state of the art who will control the technology. Some think there will be three or four companies at state of the art--Intel, Samsung and a foundry or two.
EET: What are the technology issues?
Plummer: I worry a lot about the lithography issue. It's unbelievable to me how far we've pushed 193nm lithography. The tricks we used to get to quarter wavelength and beyond are running out and increasingly expensive.
Dean, Stanford Engineering School
We can get maybe two or three process generations more with 193nm lithography. At 15 or certainly 11nm, I just can't see how those generations can be done without extreme ultraviolet lithography.
The EUV option still has a bunch of technical hurdles, but even if it works the machines cost $75 to $100 million each. Only a few players can afford them and—along with the cost of masks--this will mean significant changes in the industry I believe.
EET: What kinds of changes?
Plummer: You will have to have a market that’s huge, like microprocessors or memory chips, to recover those costs. I think unless something happens, that will create big issues for the foundry model.
Some think the foundries will get stuck at a tech generation behind what the big microprocessor and memory vendors use. The question is what the industry dynamics look like if that happens. It could change the way the survivors think about the industry because they will have a technology edge.
Intel now arguably has a generation lead. If they and Samsung have a two or three generation lead that changes the industry a lot in terms of what kinds of products might come to market.
EET: Are there any alternatives to EUV?
Plummer: People have looked at a lot of things. Electron beam lithography could work at small dimensions, but the throughout is so slow that you need hundreds or thousands of parallel beams. The result is a huge complex machine, and it's further away than EUV.
EET: What lies beyond CMOS?
Plummer: I don’t know of a nanotechnology replacement for CMOS. There are people looking at new switching devices to replace CMOS such as spintronics, but none of those look that promising as a replacement for the basic CMOS switch.
Nanotube transistors are basically CMOS using carbon versus silicon. But you still have to print patterns and make switching devices, and these are still basically MOSFET switches. Nanotubes still need gates and all the patterning of silicon circuits.
EET: So how would you characterize the outlook for a replacement for CMOS?
Plummer: My view is it’s a very long ways away. Something that will look very much like CMOS will be around for a long time.