SAN JOSE, Calif. - Is Intel Corp. preparing for the 450-mm era in its new R&D fab in Oregon?
One analyst thinks so, but he also questions if 450-mm is ready for prime time. As reported, Intel recently confirmed speculation that it will build a new R&D wafer fab in Hillsboro, Ore., and upgrade other existing U.S. facilities for 22-nm production at a total investment of between $6 billion and $8 billion.
The investment will create 800 to 1,000 permanent high-tech jobs and 6,000 to 8,000 construction jobs, Intel (Santa Clara, Calif.) said. The new development fab in Oregon, to be known as D1X, is slated for R&D startup in 2013.
''So what do make of this? We offer two key takeaways and then walk through some additional thoughts on 450-mm, as D1X is likely to be 450-mm capable,''said C.J. Muse, an analyst with Barclays Capital, in a report.
''With respect to Intel’s D1X, our checks reveal that it would likely be 450-mm ready (again, 450-mm READY), but not yet 450-mm capable,'' Muse said. ''What does that mean? It means that it would be facilitized with taller ceilings, enhanced clean room capability (capable of pumping out a larger volume and filling it with particle free air), and with pedestals able to support and contain any vibration from bigger and heavier tools. The point being - this does not mean 450-mm is ready for prime time, rather Intel is maximizing optionality.''
Still, is 450-mm ready? No, said Muse. ''We believe 450-mm is going to happen, but is likely to involve funding from chip companies. Equipment companies will not foot the bill alone this time. And it is not ready for prime time, in our view. Not at all, our checks suggest unequivocally,'' he said.
According to Muse-and Sematech-there is much work to be done in 450-mm. Chip-consorium Sematech is leading the charge in 450-mm, it was noted.
According to Muse, here's what ready for 450-mm: 1. requirement guidelines; 2. early design; 3. early prototypes; 4. interoperability test bed; 5. mechanical wafer bank; 6. technology intercept node defined; 7. single crystal wafer bank; and 8. equipment performance metrics.
Here's what is not done: 9. metrology and process equipment development; 10. equipment prototypes; 11. equipment demos; and 12. actual equipment readiness.
Plus, many equipment vendors are reluctant to devise 450-mm gear, because it is too expensive and there is no return. Only a handful of chip makers will build 450-mm fabs. Intel, TSMC and Samsung are the few that are pushing for 450-mm now.
Cost is the big issue. ''In addition, work at ISMI Sematech suggests, for beam tools like litho, implant and metrology, the throughput is a function of the area the beam can scan in an hour, the multiplier from 300-mm to 450-mm is not 2.25+ (2.25 + some additional gain due to edge gains, i.e., more dies can be squeezed in) but just 1.24 on average and actually just 1.18 for litho,'' Muse said.
''Also, during the same time, the litho industry faces the challenge of making EUV lithography work and overall the semiconductor faces obstacles like a new gate structure, new materials. So, just yet, we believe it is difficult to see 450-mm happening,'' he added.
It is a bold initiative for Intel to invest in America. That is how the future of the nation will be secured. It is necessary and should be commended by all. One hopes they have the right roadmap for the litho business.
Long term 450 mm makes sense but only for a limited set of semiconductor products with the volumes to achieve the cost benefits. The key question regarding 450 mm is how to get it done. Is there a smarter way (than 300 mm project) to fund and develop the technology?
Let me revise your question like that, Did 200mm or 150mm make any sense?
Answer is clear, yes they did, changed the global semiconductor market. Made more chips with less costs. Yeah its true 450mm is kind of premature, needs lots of attention and spendings however bigger is the better. More space to produce less worry about yields. ASML and Nikon has good projects for future of 450mm. i hope it will be accepted by all semi fellows rapidly.
With Intel pushing 450-mm the way they are, it seems like a no brainer that this facility will be built to accommodate 450-mm. Considering how soon Intel would like 450-mm to be ready, it would seem short sighted not the build the development fab to accommodate it. Of course 450 won't be ready when Intel wants it. But Intel needs to walk the walk. I wouldn't be surprised to see them make some public statements about this facility being 450-ready as a way to draw a line in the sand, so to speak. I was actually kind of surprised that they didn't state this in the announcement.
Intel does not fully invest in America. None of it's Etch tools come from the US. This is in spite of the fact that LAM, based 5 miles or so from Intel's HQ is the market leader in Etch. It's really kind of sad for them because they miss out on the great collaborative nature that Intel's think films groups have with the US based tool manufacturers. To truly invest in America, Intel needs to invest in it's American partnerships and that means seriously considering dumping some money on American made and supported Etch tools. It will create more jobs in the US than just building another high tech sweat shop for Ph.D.s.
Well, Intel is an international business and business do make decisions based on what makes sense for the stock holders and not purley on the political climate in one region of it's business. Aguing that Intel should do what's right for the US because the headquarters is in the US is like asking Google to pay their fair share of taxes. Think that's going to happen?
Intel's 22 nm chips are coming out next year. It is almost certainly too late to do anything about 15 nm now. In two-three years, 11 nm node process needs to be frozen by Intel. I don't think any wafer size change or EUV lithography will happen during that time. It is just too soon. They would rather focus on changing the transistor from planar to 3D.
I think that 15nm and 11nm will be quite interesting nodes as the cost will certainly be an issue with double patterning or even quadrapatterning. I do not see any other technology replacing conventional lithography in 4/5 yrs.
Like Reagan's Star Wars program,
Intel could be feigning the 450mm to suck Samsung
into an over-spending trap. Protectionist Samsung would try to do the 450mm utilizing their Korean supply chain. It would fail, sapping Samsung's R&D resources.
The cost of 450mm wafer fab is certainly an issue but as the cost at 15nm or 11nm node will be an huge issue in itself then it makes more sense to cancel the high cost of the nodes with going for larger wafer. I expect that most of the foundries will not be able to bear the huge initial investment of 450mm fab so there will be partnerships and collaborations.
@dnenni: Daniel, you may be right on this one, it may be another 5 years before we see production volume in 450mm wafers. It is more of a scenario where Intel may have a design that can exploit 450mm and they will fit the process to that design, it is not the other way around. If leading IDM's take this approach, the 450nm will remain accessible to a select few.
One often disregards the supporting cast of the ecosystem when it comes to 450mm wafer discussion; for example, the test solution providers for wafer probing; it was hard enough to get the 300mm wafer probing solution developed for Intel's flash memory products that some businesses went belly up doing just that. Few remain operating by a threadbare margin.
My point is that the discussion of migrating to 450mmm wafer needs to be put in a holistic perspective, especially when Intel is so selective in where it wants to participate in that ecosystem.
Dr. MP Divakar
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.