Yesterday I spent an hour with John Bruggeman, who has now been CMO at Cadence for just over a couple of months. As I’ve said before, I knew John before he joined Cadence since he was the CMO for Wind River during the time I was VP marketing at Virtutech, both of us in the embedded space. One thing I didn’t know was that John was at Oracle when, nearly 20 years ago, they decided to move from being focused on technology (our relational technology is better than yours) to tying their technology into the business processes of their customers.
John’s view is that EDA has been driven entirely by the need to make designers more productive. My own view is that everything comes back down to semiconductor economics, but the arguments may basically be the same. If we got back several process generations where the issues where clearer, we find that the move from, say, 0.5um to 0.25um was driven by the fact that a 0.25um chip was no only faster and lower power than the 0.5um one, it was about 50% of the cost per transistor. So even if you were quite happy with 0.5um you had to move anyway. But each process generation brings new challenges, not least that chips get bigger all the time, so the EDA tools and methodologies need to improve designers’ productivity enough to keep up. This, in turn, drove EDA companies’ revenues.
Looking a little deeper, the EDA companies would work with the leading edge early adopters of each new process generation to produce the tools required for success. There wasn’t that much money in the new process generation at this point, too few designs were being done in the new node. Eventually the leading edge would move to the next node, and the mainstream would come through as a sort of cash-cow. Engineering had already moved on, but the previous node tools could be sold in high volume for significant revenue.
This model is somewhat broken now. Semiconductor economics means that only the highest volume chips can justify doing a design at 45nm or 32nm. Semiconductor is a mass-production technology and the mass required for economic viability of a chip goes up at each process node. In turn this means that the mass market that made the EDA business grow is not coming through and so not driving revenue in the same way. Of course there are still designs going on in 90nm and 130nm but, by and large, everyone has all the tools they’ll need already.
John told me that obviously Cadence will continue to push on the productivity strategy. After all, Cadence’s bread and butter is ensuring that the leading edge semiconductor companies can design leading edge chips successfully.
But John wants to have a two-dimensional strategy, where the other dimension is profitability. Design tools potentially generate an incredible amount of data. After all, every button click, every file write, every verification run generates information about how the design is progressing. But there is currently no way to tie this data into the overall design process, let alone the company’s overall product development process. Since so much of a chip design these days is software, by some accounts 60% of the cost, attacking this problem will require pulling embedded software into the mix in a way that adds value.
If this type of strategy is successful, it has the potential to make Cadence (and maybe all of EDA) into a more strategic asset of their customers, rather than simply being an expense to be managed by corporate CAD and purchasing. Unfortunately for EDA, its customers don’t think “if I gave 20% more to Cadence how much additional money would we make” any more than they think that about expanding their IT budget for Microsoft. It has always been an indictment of EDA that despite their strategic importance to their customers, they didn’t really have access to the boardroom in the same way as SAP or Oracle. Oracle really did escape the IT department whereas the EDA companies are largely stuck inside the CAD department.
Of course Cadence has tried before to become more strategic, primarily back in the day with a strong push for services. The consultants brought in from companies like Anderson Consulting (now Accenture) or Coopers & Lybrand were already experts at dealing with the highest levels of their customers. And Cadence knew lots about design and electronics in general. However, there was never sufficient cross-pollination to get the consulting knowledge into the EDA silo or to get the consultants to understand design deeply. The end result was a lot of consultants with Cadence branding who knew little about the domain where they were supposed to be giving strategic advice. Electronics Infusion was a catchy name but it never really came to much.
It will be interesting to see how this works out this time around. John has one big advantage and one big disadvantage over more obvious choices for CMO: he is not an EDA insider, and he is not an EDA insider. He has not got jaded about the ability of EDA to escape the CAD group. Instead of having spent the last twenty years inside EDA, he spent it in Oracle, Mercury Interactive and Wind River. Making Cadence more like Oracle, and tying its design process more into the area where Wind River was strong, certainly have the potential to make it a different kind of company. But EDA has tried many times before to crack this problem, so we’ll have to wait and see.