SAN JOSE, Calif. - Who can afford extreme ultraviolet (EUV) lithography?
Answer: Very few companies. And the tool price tag is going up. ''EUVL: and the price is? The answer is...$125 million'' per tool, said G. Dan Hutcheson, CEO of VLSI Research Inc., in a report. ''Don’t choke. Years ago, (Risto Puhakka, president of VLSI Research) predicted this with a cost-per-pixel analysis. Everyone thought he was insane at the time, but many are now signing off on production tool orders.''
In 2003, Intel Corp., the proponent of EUV, wanted a per-scanner target cost of $20 million. EUV is supposed to replace today's 193-nm immersion lithography-if or when that runs out of gas.
But EUV has been delayed and R&D costs have exploded, causing tool costs to soar out of control. An ''alpha'' EUV tool from ASML Holding NV went for a reported $60 million per unit. IMEC and Sematech bought the first ''alpha'' tools from ASML.
Still, EUV is gathering momentum. As previously reported, ASML has six orders—and is essentially sold out--of its initial production-like EUV tool, dubbed the NXE:3100, which is due out by year’s end. That tool reportedly sells for about $90 million.
The Dutch company also has 10 orders for its next-generation EUV tool, called the NXE-3300, according to Barclays Capital. Hynix, IMEC, Intel, Samsung, Toshiba and TSMC are the initial customers for ASML Holding NV's ''pre-production'' EUV lithography tool, according to an analyst.
Plus the inspection tools are not ready and the COO of EUV tools is lot higher because they have much longer down time. More than 50% of all health care cost is spent on the last 6 months of a patient, same is true for Moore's law.
Today the "production like" tool costs about $90 million, with presumably a 10 watt light source. The estimates are that 100 or 200 watts or more will be needed for volume production. No testing has been done at this high power level, where there will certainly be more problems to solve, like heating and lifetime.
The one million question is what are Canon and Nikon gonna do now+ The have not chance to compete with ASML in EUVL-project.
So, they have to find another technology to help then to survive and my guess is nanoimprint technology (NIL).
What obout Mark LaPedus ?
Indeed companies have moved to 90 nm on KrF (248 nm), keeping their process window and avoiding new resists. They still seek to move to smaller nodes but at slower pace. When they decide to go to 65 nm, their choice would be interesting. Will they try to extend 248 nm virtually forever, as it now appears the big companies have shown how to extend 193 nm practically forever.
It's debatable if skipping to 193 nm directly to EUV is a good move, even for something like 90 nm or 65 nm. First, the DOF is much less. Second, there are fewer photons used for the same dose, so the image is noisier. Third, while there is no need for resolution enhancement techniques which tend to be expensive, the EUV mask itself is pretty expensive due to the wider variety of defect constraints.
Well the billion spent on just the EUV tools alone is a waste of good money that could be spent elsewhere. The only people who don't know that are those who don't know enough about EUV and are therefore not qualfied to judge.
There will not be a cheap solution for 15nm and below. Chip makers that have not adopted 193 are not producing high enough volumes to benefit from making a change. Industry leaders have to be on the bleeding edge or their will lose their competitiveness.
EUV is extremely complicated technology which is why it is so expensive. But if it does get adopted all other parts of Fab will have to develop applications that work for sub 20nm. The R&D capital that will go towards this is certainly good for the industry.
Citing that Intel spent a paultry 77Mil on venture as an argument against EUV is pretty moronic. How about the fact that Intel spends 13 Million a year on their CEO? How many more engineers could they hire with that money so that their R&D process engineers are not treated like slaves, and as a result, far less productive than they could be.
This is ~$100 million expenditure for one EUV tool, not even hitting 15 nm target. This week just read Intel spent $77 million to fund 18 startups. The EUV tool alone is not enough for EUV to work, the other pieces are not in place either. Comparing to imprint or maskless, those alternatives seem more cost-effective.
David Patterson, known for his pioneering research that led to RAID, clusters and more, is part of a team at UC Berkeley that recently made its RISC-V processor architecture an open source hardware offering. We talk with Patterson and one of his colleagues behind the effort about the opportunities they see, what new kinds of designs they hope to enable and what it means for today’s commercial processor giants such as Intel, ARM and Imagination Technologies.