Will EDA growth be “incremental” or “explosive” over the next five years? It could be explosive, says analyst Gary Smith, if EDA vendors figure out how to tap the embedded software development market.
Gary’s forecast presentation has become a tradition at the EDA Consortium’s annual pre- Design Automation Conference (DAC) networking gathering. Perhaps reflecting a “less is more” theme, there was only one short analyst presentation this year, but it was strongly worded. Gary argued that this is a “make or break” year that will determine whether EDA follows a rather anemic growth curve, or an accelerated one. And the answer to that question has largely to do with embedded software development.
Gary first offered an “incremental” EDA revenue forecast. This forecast predicts an 8.6% compound annual growth rate (CAGR) through 2013, “sort of usual, not explosive,” he said. CAE (front-end design) is expected to have a 10.7% CAGR, partially because ESL design will add to the number of FPGA seats. IC CAD will have a 7.6% CAGR, experiencing slower growth as RTL signoff grows. And PCB CAD will have a 4% CAGR.
How to do better than this? The way to grow “explosively” is by adding seats, Gary said. He showed a graph that depicts system-on-chip and board design seats as relatively flat, while embedded software development seats are growing. If we can grow ASPs in that area to $30K/seat, the result will be an $8 billion market, he said.
The problem is that the embedded software development tool (ESDT) market is fragmenting and is being consumed by acquisitions, most recently Intel’s purchase of Wind River. EDA vendors have not, by and large, focused on this market. It’s mostly a real-time operating system (RTOS) market anyway; tools are generally a low-cost add-on. EDA is a more natural fit with tools, especially multicore development tools.
The question now, Gary said, is whether semiconductor vendors will offer proprietary software development tools for multicore platforms, or whether commercial alternatives will emerge. Designers don’t want proprietary tools, he said.
I agree with many of Gary’s arguments. EDA vendors have not had much involvement in software development previously because tools are free or cheap. A good parallel compiler for multicore platforms could be a high-value, high-ASP type of item. But will people pay tens of thousands of dollars for device-independent compilers if device vendors are giving proprietary compilers away so they can sell more silicon?
The answer, I think, lies in changing a corporate culture that is willing to spend money on hardware design tools but not on software development tools. Recent data shows that software development costs exceed hardware SoC development costs as process nodes shrink. That realization is driving a market opening for virtual platforms and hardware/software co-verification. If dollars start flowing to software development teams, an “explosive” opportunity could indeed follow.
Richard Goering