Q. How do you see the evolution of the fabless industry now versus the traditional IDM model?
A. The Semiwiki.com folks have championed the fabless industry as the number of IDM’s has dropped due to high cost of capital equipment, and need for IC volume to utilize the larger wafers that give lower costs. And the bleeding-edge Memory and Processor products will be forced to follow the larger wafer and smaller images down that high cost path. Only a few (2 to 4 perhaps) IDMs will remain in that segment of the IC market.
But the Analog, MEMS, and specialty chip makers do not need largest wafers and highest cost lithography tools. So there is a recent resurgence in 200 mm fabs including new tools for that size wafer while the bleeding edge folks play liar’s poker with each other over WHEN they might go to 450 mm wafers and EUV lithography or both. The next few years should show the 200 mm resurgence, which later might become the 300 mm resurgence IF and WHEN the 450 mm tools start moving into mass production. These next few years may clarify that scenario as cost of capital perhaps goes UP if this prolonged non-inflationary period ends smoothly. Perhaps its that uncertainty that is delaying the 450 mm conversion.
Q. Do you see a need for Fabless companies to integrate ERP systems with Yield Systems?
A. long ago ERP (and earlier MRP) systems were office systems while Realtime Control Systems ran the factories and the IT skills were very different for Office vs Factory systems for decades. Word Processors and Green-sheet reports gave way in offices to the Macs and PC’s and personal laser printers. And Appletalk or HP SRM gave way to ethernet in various proprietary forms and finally the victory of TCP/IP protocol that led to the world wide web.
So as Office Automation got fully integrated, Factory Automation was benefitting by the move from 20ma current loop to RS232 to ethernet connectivity..slowly…over many years and now Office and Factory can be integrated AS NEEDED. So the question is simply which customers drive which companies to fully integrate Office and Factory systems.
And for non-factory fabless companies, the question is how closely integrated do they need to be with their foundries and their remote assembly and test partners? The Oil and Gas industry is finally integrating core drill analytics to remote controlled drilling platforms and crude oil transportation systems. This INTEGRATION effort is never-ending and cuts across all physical industries, but is slower than the “big data” news from the consumer communications industry which dominates the news. Its all about the leadership vision and the cost of capital for each industry segment. The technology is available if the leadership can envision optimal systems for where he/she wants to take their company. Luckily, the protocol wars are less of an impediment now days.
Q. How you see data management systems like yieldHUB integrating with analytics languages such as R?
The R language is a graphics and statistical toolset that requires another language to capture and format the data, but it is great tool to customize and supplement proprietary analytical systems as long as the customer has local programming talent and sees the need for customization over time. I have seen R used very well to augment SAS, JMP, Oracle, and even analytics companies that work with many clients using many different data sources, but only if they have long term programming and support capabilities locally or with cost-effective support services.
Its not Excel. But its worth the effort if the commercial systems need a bit of customizing. Yield Hub clients may or may not have R support talent, but MfgVision could add R to their internal bag of tools for prototyping and fast customer response. Its a very fast growing set of tools in open source, but now that Rendevous R was purchased by Microsoft and integrated with their Analytics Services that seems have accelerated the enterprise view of its value. ATT gave us C and S, then S+ became R in open source world in my view. Every language evolves, so there are newer open source clones for Matlab for example. Any technology company can now find open source tools for any engineering task they need but the key is long term support.