Get email delivery of the Cadence blog featured here
There is a famous 1981 movie called My Dinner with Andre. If you've not seen it, you might expect that it is like most movies and rather more happens than is conveyed in the title. But, no, the movie is basically a conversation over dinner for two hours, between two actors using their own names. Well, I had breakfast with Lucio Lanza a couple of weeks ago. Unlike the movie, which was so low budget they shot it in an abandoned hotel in winter without having enough money to pay for heat, we met in the very comfortable environment of Il Fornaio in Palo Alto. Even for breakfast, Lucio invites you to an Italian restaurant because...well, how else will you get good coffee in a country that is not Italy?
Yesterday, I recapped my discussion with Lucio when he had just won the Kaufman Award, and we discussed how he went from studying a lot of philosophy to switching to electrical engineering and has ended up, in Lanza TechVentures, as one of the only people who will go near the fabless semiconductor ecosystem.
One area that Lucio is interested in is machine learning, connecting technology to human beings on their own terms. He hates the term AI, artificial intelligence, since it is not artificial, just different.
Another related area is the internet of things, IoT, since he feels it will require a big sea change in the semiconductor ecosystem. This is a long-tail market, with a large number of designs in relatively small volume, as compared with the mobile market, which is a few chips in massive volume. IoT devices need to have power that is orders of magnitude less than currently available, and they will not be designed in the most advanced processes. Chips in those nodes are too expensive to bring to market with design costs and the non-recurring engineering costs such as mask costs, overwhelming any savings from volume (because generally there isn't much by traditional semiconductor metrics).
But Lucio thinks bigger than this. He also thinks that the corporation as we know it is over. The distributed corporation is the future. Everyone going to work in a building away from where they live is going to fade away. People can work wherever they are. Having worked both from my home and in a normal office, I think that the view that everyone will telecommute all the time underestimates the social aspect of work. I may just be old fashioned, but I like to meet people face to face and not just communicate by email and phone. It seems to be an important part of being human.
Even if big corporations don't vanish, they will change. When Lucio started his career at Olivetti, he would go into companies and they would have rows and rows of accountants since bookkeeping was not yet computerized. He feels IT departments will go the same way. But he is more interested in the changes he sees in design departments, especially in the analog area.
One company he has been working with is eFabless (when a VC says "working with" he means "invested in"). This embodies both of the above points, that costs are too high and that people don't all want to show up in the office every day. There are lots of really good analog designers "living in small villages in Italy" who are under-utilized because they don't want to move to Silicon Valley. eFabless allows them to work and get paid. I have to admit that analog designers often do seem to be eccentric and the sort of people who do not want to work for a large company. Those people are all over the world, not just in Italy but also UK, Tanzania, academia, towns in the US mountains, and more.
The business model of eFabless is similar to 99Designs in the graphic design world (even the tagline is almost perfect "need custom design'). Lucio brought up his model as Jack Hughes' TopCoder, which has been in existence since 2001 and does over 7,000 projects per year, with $80M paid out to their community. People want work done, they specify it, people compete for the job and the winner gets paid. There is additional money if the design is used in a highly successful product.
Lucio had the idea to "crowd source" analog design about six years ago, before the term "crowd source" even existed. He talked about it to Mohammed Kassem, who was in charge of analog design at TI and he started to make it real with his contacts. He is the CTO. Lucio brought Michael Wishart on to run the company, wth a background from Cypress Semiconductor and Goldman Sachs. Mike is making it into a real corporation with a unique set of people. Lucio feels strongly that people should not have to leave their country to be successful. Or even on a smaller scale, have to leave where they grew up to go to the big city. Italy is full of small towns where there are only old people. I'm not sure everyone wants to remain in towns like that, there is clearly an aspect like the old song after the second world war: How ya gonna keep 'em down on the farm (after they've seen Paree)?
Lucio's view is that IoT is going to be far from the leading frontier of process technology, at 180nm. Interestingly, last week I wrote about Léman, whose chip is mostly analog and in, yes, 180nm. In Mike Wishart's own words:
The acute need for customized silicon for smart hardware products will be met by a community of unaffiliated designers on affordable, re-purposed 180nm nodes with libraries of proprietary and open-source processors and on-demand analog and mixed-signal IP.
The first design challenge is over, with hundreds of people interested, resulting in five designs, one of which they will use. The contest was sponsored by X-FAB (based far from Silicon Valley in Erfurt in the eastern part of Germany, the semiconductor city there that isn't Dresden).
It remains to be seen whether the model scales. Designs need a lot more ecosystem that software (EDA tools, sometimes basic building blocks like memories, perhaps test silicon, and so on). But if you predicted in the late 1990s the existence of cloud datacenters, you would probably have also predicted the operating system powering them would be Solaris, Windows/NT, IBM/CMS, HP-UX...certainly not Linux. And yet...
Take a look at the eFabless website to learn more.