You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!

  • SEMICON Day 3: Leti, Intel Keynote, 450mm (Pic!) and Innovation Keynote

    Intel say fabless model collapsing... really?-eiffel.jpgLet's start with yesterday evening, so technically yesterday. It was July 14th, which is the equivalent of Independence Day in France. So the perfect day for Leti, based in Grenoble, to present a lot of the work that they are doing on 3D "more than Moore" type technologies, including photonics. Also, wafer-scale LED lighting. And some need micro tomography for looking at things like a single transistor or a single copper pillar.

    I've lived in France and worked on European projects so I've always been a bit skeptical of these sort of funded research programs where the results are PhD theses and a lot of the funding seems to be vacuumed up by large companies. But Leti seems to aggressively move stuff out into startups when the technology looks viable. For example, the wafer scale LED lighting is going into production in 2016. Some of the 3D technology looks promising too, such as the CoolCube which is looking at some of the right issues namely how to build up vertically in a way that keeps the temperature low. One of the issues with semiconductor manufacture is that you need diffusion ovens in the FEOL and they are too hot for the metal stack in the BEOL.

    Intel say fabless model collapsing... really?-intel.jpg

    Anyway, onto today. Doug Davis of Intel gave the keynote. He is the GM for IoT (which I was interested to hear includes Wind River which, in turn, includes Virtutech where I used to work. I guess if I'd stayed I'd be a "thing" too). His thesis is that the Internet of Things is a disruptive force. I've no doubt it is, and he quoted the 50B things by 2020 number. But as I discussed on Day 1, it is a high volume market where it is not clear the component suppliers make the money. When stuff gets up into the cloud, then obviously Intel has a play since they make a lot of money in the datacenter.

    Doug started talking about a Swedish company Yanzi creating smart offices. Sensors on wastebins to know the are fill, washroom towels and soap, copier paper and toner. Cute, but not exactly that disruptive. He then moved onto four areas where he though there was lots of scope:
    • aging population: turning the home into a connected care center so that old people can age in place (by 2050 then 40% of us will be over 60)
    • environment: I think he cheated here since the example he used was improving an Intel factory in Malaysia by detecting imminent issues, reducing maintenance by 50% and spare parts by 20%
    • urban boom: which mostly means China and Asia since 46 of the worlds biggest cities in 2015 will be in China, and half in Asia. They have pilot programs going in London.
    • feed the planet: we waste a lot of food so we have enough, in fact agricultural productivity is rising faster than the population. His example was a program (using the much-mentioed "Intel IoT Gateway") to automate water feed to Malaysian rice paddies which ended up saving 10% of water but, more significantly, increased production by 50% from 2 harvests per year to 3

    Interestingly two of those, medical and agriculture, were ones that ARM called out two days ago as areas ripe for disruption by technology (and computer vision will probably be involved in some of these other areas too).

    I can see how the service supplier and the system integrators make money. I still don't see how a company like Intel makes serious money on low-margin products. Especially Intel, who have to keep their margins high because that's what they do. They are not Atmel. I know they have smaller processors and stuff these days, like Quark, and Doug even showed us a module called Curie the size of a dime (incidentally, as he pointed out, the cost a car would be if they had obeyed Moore's Law pricing for the last 50 years. They'd go fast too. Insert your own "blue screen of death" joke here).

    Next was the annual G450C update. Three people presented and I may have their names wrong since they were not on any slides and the presentation was billed to be by someone else. Chris Borst, who is the fab operations director of the 450mm fab at SUNY, CC Chen, from TSMC, and Erin Fria of Intel.

    Current status:
    • Excellent test results with 51 tools installed in Albany
    • The Nikon scanner arrived (previous scanning was done in Japan at Nikon)
    • Notchless wafers with 1.5mm edge is the SEMI standard
    • All partners are supporting equipment in NY through Q1 2017
    • 40nm and 28nm process
    • 73 wafers patterned to date

    So bottom line:
    • >30% die cost advantage
    • Everything is still viable
    • HVM 1.5 to 2 years from the "go" signal, ready for intercept with 300mm at end of decade, currently targeting 10nm insertion

    Proof: me holding a fully patterned copper BEOL flow (just one layer of metal for sure) 450mm notchless wafer!

    Intel say fabless model collapsing... really?-img_2325_fotor.jpg

    So now the question is whether it truly makes sense from a business perspective. I don't know how the 30% die cost advantage above was derived and what it assumes about the cost of all the equipment.

    Next meeting was with Gary Patton and Subi Kengeri of GlobalFoundries about 22nm FD-SOI, including the schedule, PDK availability. I went to GlobalFoundries panel session coming on the 22nm FD-SOI ecosystem so I'll blog about both of these together later, in a separate blog. Look for that Friday morning.

    Intel say fabless model collapsing... really?-inno1.jpgSo last thing of the day was the end-of-event keynote for the Innovation Forum. It was by Steven Forrest of University of Michingan on Moving Innovation from the Lab to the Marketplace: the Critital Role of Academia in the Innovation Chain. He said the first part was 1000 things academics could do wrong and about 3 they could do right. He was at Bell Labs, where he invented the photodetector now found at the end of every fiber optic line in the world. Then he and his students invented the AMOLED (which is apparently 100% efficient due to some quantum effect so perfect for lighting once they get costs down) and more. Factoid he knows as a result: Samsung Galaxy phones are the most-owned electronic gadgets ever at 750M.

    Issue #1 he said universities need to deal with is that they all have offices of technology transfer who think they are judged by how many royalties they bring in. Companies hate that attitude and run away. He told them that the only thing he cared about was how much technology got transferred. The royalties would take care of themselves.

    In the US we have it lucky. The VC's attitude to failure is "someone else paid for your lesson." In the rest of the world it is (largely) not like that. But we have a huge political immigration problem (and I don't mean illegal immigrants) because it turns out that most startups are started by immigrants. I remember my own experience a few years ago where I worked for a startup (that the board eventually gave me to run but that is another story) with an Israeli CEO and an Iranian CTO. That doesn't happen in most of the world.

    So what are his recommendations?
    • Building a company is done brick by brick, no shorcuts
    • Be prepared to "slay dragons", they multiply fast
    • Use every resource, "doing it yourself" usually leads to disaster
    • Don't worry about an exit strategy
    • Nobody's technology is good enough to be "ahead of its time"
    • Inventors (professors) make lousy CEOs. Pick a single career
    • Hard lesson: companies and their founders' interests eventually diverge so exit as gracefully as possible

    He also studied Israel (also famous for innovation) and the US and looked at the challenges of those two ecosystems of innovation. I leave you with that slide. It's been a long day.

    Intel say fabless model collapsing... really?-inno.jpg

    Final conclusion, in the large post-industrial lab (think Bell Labs or Xerox PARC) post-Moore's Law era, universities and governments have a shared destiny in the "globalized world of science." We can't succeed without each other.

    This has been an experiment today. I have blogged each event immediately after it happened. Did anyone read any of it? Was it a good idea? Answers in the comments.