Last week the 50th Design Automation Conference (DAC2013) was held in Austin Texas and, of course, fortylines was there.

First impressions

It is the first time I attended DAC. Veterans told me the showfloor was a lot smaller than it had been, maybe up to 1/5th the size. Personally, for an industry projected to reach $6B in 2014 and dominated by 3 major players (Synopsis, Cadence, Mentor), the showfloor looked of reasonable size.

When you compare to the Game Developer Conference or meetups around the thriving San Francisco startup ecosystem, the first surprise was the rather older crowd. Young faces were singularly missing - apart from the booth babes you typically find at major conferences that is to say. It looked odd there did not seem many people below 40 years old attending when you know DAC distributed free passes months in advance.

Second surprise, the French speaking contingent was there in full force. Outside Montreal, I have not been many places where you can indistinctly choose English or French to communicate with vendors. DAC2013 was one of them.

Last first-time impressions, for an industry that builds the tools that builds future technology, demos were very conservative; not much social platform integration (Few people knew about Asana), very little cloud integration, zero talk of leveraging big data analytics and no kinetic or other pushing-the-envelope human-computer interfaces. Hell, fortylines was the only company that could demo its solution on an iPad there!

The EDA Hunger Games

The discussion panel titled The EDA Hunger Games was very well animated and truly informative.

While most of software giants, following Twitter lead, have somewhat agree to see patents as deterrent defensive weapons, pretty much any of the big EDA and IC guys can and will bankrupt your company through a lawsuit if they feel threaten. Lawsuits have been talked as a real problem to foster a vibrant EDA and IC startup ecosystem.

Funding is also an issue for many EDA/IC startups. The minimum angel investment required to get off the ground is huge, in the tens of millions of dollars. Even at say $50M, that would not be so much a problem for Venture Capital Funds if the return-on-investment was not highly speculative. VCs are looking for homeruns (10x ROI) and so far the EDA/IC industry has only delivered a handful in the last thirty years. When you compare this record to social and e-commerce industries that can get off the ground with minimal capital and deliver a couple homeruns per season, you understand why VCs are not flocking to bet the house on a pure EDA/IC startup.

There was debate about financing the startup ecosystem through strategic investors. Either it is Intel, Samsung or Qualcom, they are all more than willing to make huge investments because they recognize innovation as a key component of their business. They are though public companies and as such have two major constraints. First, they cannot do equity investment otherwise the financial results of the startup will show up in their books. Second, they have to justify strategic investments to their shareholders. Right of first refusal to technology sale, life-long discounts, etc. are at the center of a negotiation with a strategic partner. It seems obvious yet many entrepreneurs make the mistake to present a VC-oriented slide deck to a strategic partner.

Last data point of the debate: there are about 100 super sales rep that know all the channels and accounts in the industry. That is a really small number. Should a startup then invest in its own sales force or go through one of the major sales middleman?

On the web and in the cloud

I had a fun discussion with an engineer at SkillCAD on using Microsoft kinetic to create wires in their visual custom IC layout tool, but for the most part I was interested in what is happening around EDA and cloud computing.

Most of the EDA vendors have made attempts to move into the cloud (see Olivier Coudert - FMCAD 2012). Talking with business developers, their IC clients are usually the ones resistant to the initiative. Protection of Intellectual Property is the number one concern on everyone's mind. Anecdotally companies in China are the most concerns about theft of IPs.

ChipPath deserves kudos for an imaginative and pragmatic use of the web. They store a regularly-updated database of data-sheet information for various IPs and provide a web front-end to explore and estimate back-of-the-envelope area, power and performance. ChipArchitect will also generate some scaffolding code from what I understood. This is a pragmatic approach to build an IC marketplace while alleviating most of the concerns with storing IPs in the cloud.

Onespin solutions is another interesting company that has taken a pragmatic approach to cloud computing. Onespin formal verification toolflow is simple to understand and gets around most of the security concerns. You compile your IC design with their client software on your IT infrastructure, upload the compiled version to their service in the cloud and get back the results which are correlated back to your design through the client software. Actually in the early days of fortylines we thought to provide cloud-based verilog simulation in a similar fashion.

On the DAC2013 showfloor many companies have integrated the fear of their customers for words like public and cloud. You would thus hear a lot of multi-site thrown around. The thruth is: When you want to integrate revision control and manage large data set across sites like IC Manage does, you need some kind of sophisticated cross-continent IT infrastructure.

With two sites connected through the global Internet, you implicitly trust many agents: the cable company, the encryption software, etc. The threat model is completely different from the air-gap network, perimeter security (think Airports) you had with a single site. So what does multi-site vs. cloud really mean? From a security threat perspective: nothing. From a liability perspective: it boils down to the rule of law and who you trust.

Designing ICs in the Cloud

Cloudification always comes in three steps across all industries:

• 1. Current tools are re-engineered into a cloud-based version.
• 2. Business models evolve to various Software-as-a-Service (SaaS) models.
• 3. New applications emerge around the elastic nature of cloud infrastructures.

The cloud provides leverage and thus directly benefit Small and Medium Businesses first and foremost. SMBs are always competence starved (We are in a knowledge economy after all) and thus are often a lot better off outsourcing IT security. Using an analogy, they are better off depositing their cash at the bank rather than keeping it under the mattress.

What a SMB needs the most is customer awareness, aka marketing. IC startups are no different. Any cloud initiative must focus on building a marketplace for IC designs centered around trust management. E-Bay and Credit Card Companies are great examples of trusted third party (i.e. A and B do not trust each other but both trust C). Nothing new and special here.

Now the advantage a trusted IC marketplace is that you can test drive IPs before buying them. You can build and test your whole IC without ever seeing the code of the IP you are designing into your system. That requires various design space exploration, simulation and verification software. Fortunately many third-tier EDA vendors are struggling against the big three and are eager to find alternative revenue channels.

Every IC/EDA Small and Medium Businesses benefit from such a design-in-the-cloud marketplace. At this point, the only constraint to grow a profitable marketplace business is the health of IC/EDA SMBs. Fostering a vibrant IC/EDA startup community is a key element to that end. That's why fortylines focuses first on education and tinkerers looking for fun stuff to do using open source tools.