Last time I worked for Cadence in the early 2000s, Adriaan Ligtenberg ran methodology services and, in particular, something we called Virtual CAD. The idea of Virtual CAD was to allow companies to outsource their CAD group to Cadence. In effect, we would be the CAD group for the design company. In 2000, Synopsys had launched DesignSphere Access with Avant! (this was pre-acquisition so Synopsys didn't have physical design tools). There was a big press conference but it didn't get off the ground. I would say that this was an idea before its time. It is hard to recall now how primitive the internet was in that era, cable modems and ADSL were just starting to be introduced. It was still another 6 or 7 years before the iPhone would be introduced. Google was two years old. Mark Zuckerberg was a 16-year old teenager. Mark Benioff had just become CEO of salesforce.com but nobody had come up with the term SaaS, or software-as-a-service. The beginnings of cloud existed, but not under that name. The buzzword from that era was Application Service Provider (ASP). Venture capitalists founded literally hundreds of such companies with the idea that software companies could provide services available on a pay-by-use basis. They pretty much all failed in the 2001 downturn. Even Pandesic, which was backed by Intel and SAP, and others with gold-plated backers failed. Banks were not yet online. eCommerce was getting going, but it was just servers in company datacenters. But, more to the point, not a lot of people wanted what we provided. I think that there were 3 big reasons. The biggest one of all is that the Internet didn’t have a security mindset yet, and the companies doing the designs regarded them as their crown jewels, and not to be let off the companies’ own servers under any circumstance. This already sometimes caused tool debugging issues since customers were reluctant to release test data to Cadence. A second problem was that customers wanted to incrementally add Virtual CAD to their existing infrastructure, not get rid of an entire existing CAD department and all their in-house servers. Of course startups are different, but semiconductor startups were a rare bird after the 2001 technology recession. But that sort of hybrid didn’t really work well if only outsourcing, say, the physical verification and signoff of a design, that’s a lot of data to be copying back and forth the whole time. The third problem was that a pure Cadence flow wasn’t what customers and foundries wanted. It is no secret that back in that era, everyone signed off their timing on a non-Cadence tool, and mostly did their physical verification on a non-Cadence tool too. Ultimately, these early implementations just didn't offer enough value. You didn't need your own computer, and you could use a web-browser, but on its own that wasn't compelling. Over time, the Virtual CAD idea at Cadence morphed into what is now called Hosted Design Solutions (HDS). This works well for small companies, who can get away with never having to buy servers, hire IT people to manage them, and hire CAD engineers to maintain the flows. Their design never leaves Cadence until tapeout. You can read about a good example in my post Léman Micro Devices: Blood Pressure on Your Smartphone . However, it doesn’t really work for large companies. Cadence doesn’t have enough servers to allow a major semiconductor company to outsource their CAD organization. A large company often has tens or hundreds of thousands of servers. But as I said earlier, only outsourcing a little of it doesn’t really work well either. We simply can’t scale up and scale down the resources in our own datacenters. The earliest cloud company was probably salesforce.com, taking business processes and putting them in the cloud. Not only was the risk perceived lower, but ROI was high since you could avoid all the hordes of Siebel and SAP consultants. I remember Cadence in that era was bringing up a big Siebel CRM system in house, too. For EDA, one of the big value propositions is the potential for hyperscalability. We don't need to scale to millions of individual users, like Salesforce (that would be a nice problem to have!). But we do need to get very fast access to high number of CPUs at reasonable cost for jobs that require a lot of parallelization. At the recent ESD Alliance CEO panel, the cloud was almost the only thing discussed. See my posts The ESD Alliance CEO Panel: Forecast Very Cloudy and CEO Panel: Cloudy with No Chance of Meatballs . At DAC, there is a new infrastructure alley this year for the first time, which I talked about last week in my post The Design Infrastructure Alley. Sign up for Sunday Brunch, the weekly Breakfast Bytes email.![]()
↧