Intel Corp. reportedly sent home Irish staffers early from production training in the U.S. after deciding to delay the introduction of 14-nm process in Leixlip, Ireland. What are the implications?
LONDON -- The good news for Intel's Irish contingent is that they get to go home early. The bad news is that they get to go home early due to a reported project delay.
The Irish Times recently reported that Intel was sending up to 600 Irish staffers home early from production training in the U.S. after deciding to delay the introduction of the P1272 14-nm process at Fab 24 in Leixlip, near Dublin
The workers had been training on 14-nm tools in preparation for bringing that process online at Leixlip. Intel confirmed earlier this year that Fab 24 would be the overseas fab to join D1X in Oregon and Fab 42 in Arizona as sites that would run the 14-nm process. Intel is thought to be spending more than $1 billion to refurbish Fab 24, and this was seen as guaranteeing the future of Intel manufacturing in Ireland for at least another iteration of Moore's Law.
The Times reported that the ramp of 14-nm production in Ireland has been pushed out six months to late 2013, attributing the delay to a slowdown in demand rather than technical problems. However, the reported delay is driving concerns that Intel may be reconsidering the ramp of the 14-nm process in Ireland.
That would cause disquiet across Europe, where Intel is the biggest private-sector employer with 4,500 skilled workers.
The presence of leading-edge manufacturer in Ireland has been encouraged by the European Union. The EU sees nanoelectronics as a key enabling technology, and is reluctant to see these skills leave the continent even where they are owned by offshore companies like Intel and Globalfoundries.
This brings us to one of Intel's competitors. It is interesting to note that Intel's 14-nm technology will be its second process node based on FinFET technology while Globalfoundries' 14XM process is essentially a 20-nm manufacturing process with FinFETs dropped in as replacements for planar transistors.
That means Globalfoundries' 14XM due in 2014 will be among the first to transition between manufacturing process nodes without a die footprint shrink as a cost-saving driver. Along with reduced die area comes reduced power consumption, something prized by mobile chip companies.
Intel's 1272, due in 2013, should represent a shrink from the previous P1270 22-nm process as well as a reduction in power consumption – maybe. Nothing is easy or automatic about miniaturization at this atomic scale. Even though Intel has said it doesn't require extreme ultraviolet lithography to make it work, the use of multipass patterning with 193-nm optical lithography is likely to have an impact on both the cost of manufacturing and yield.
If Intel can deliver the P1272 in 2013 – whether from two or three production wafer fabs – it will preserve its reputation as a chip maker that, whatever its architecture and design decisions, understands semiconductor manufacturing better than others.
On the other hand, if it turns out that P1272 14-nm has technical problems that require a fix – such as the use of silicon-on-insulator wafers as the starting point – a delay in delivery into 2014 will constitute a serious blow to Intel's reputation and to its position as the world's leading chip company.
The 14 nm ramp should proceed at Ireland, but with less training.
In the complementary approach, EUV is still one of two exposure passes because mask defects are not manageable yet for 10 nm. So it's still more expensive than current double patterning.
Intel Ramp engineer claimed moral at Intel was at a all time low.
Many forced plant shut down over December holidays. 14nm delay relates to "significant process problems"
Even worse his manager said there is talk about a work force reduction in 2013 that will be announced in Q1.
Intel has announced that they are going to build yet another Fab building right next to D1X in Hillsboro, Oregon. It will be slightly bigger than D1X. (Total area in the buildings will be larger than the Pentagon.) My assumption was that they were going to use it to bring up non-X86 processes as per various rumors. A part of this would be to clone their World's-Best Fab Bringup team. That team is concentrating on bringing up D1X real soon now.
If/when they succeed, they would normally do the copy-exactly to Arizona and apparently Ireland.
OTOH, they can bring the new building up while training a second team. The decision to use the new building for other device types or X86 can be made based on market conditions a year or two out.