Forecasting has gotten a bad reputation this year. Thoroughly discredited, along with the forecast they spawned, have been the assumptions that we were in a new world, that there was now a new business model completely different from those of the past and that the new supply-chain management tools had eliminated the business cycle. Keep in mind that a forecast is merely a mathematical calculation based on assumptions. You can never question the numbers; you should question the assumptions.
That brings us to the strange EDA growth forecast of 20 percent, strange at least in a year of a record downturn in the semiconductor industry, gloom and doom in telecom-in fact, electronics in general. The sound bite claims the EDA industry is countercyclical. The truth is, not quite.
All large cycles are made up of smaller cycles. The EDA industry is driven by design, not semiconductors. And although semiconductor design is the most important of the design styles, it is only one of three possible ways to design a system. What really gets sold is systems, not semiconductors. Now multiply that by the three skill levels: power user, mainstream and late adopter. Then add the three areas of CAE, CAD and CAM.
Finally, take into account the two parts of the design cycle, design and verification, and you have 54 actual cycles in the design world. (You can throw in the idiosyncrasies of the 61 types of design tools if you really want to get thorough.) After all that, what actually happens is that you look at the main market drivers and use assumptions based on their cycles to drive the forecast.
The thing you need is a good top-down, demand-driven look at the design industry. For the last three years the major purchasers of EDA tools have given Dataquest a projection of 20 percent spending increase. This year has been no different. We have checked out all reports of reduction in EDA spending and found only one major company that is reducing spending. Obviously, it is in real trouble. So why hasn't there been 20 percent growth in previous years?
First is the switch from perpetual licensing to subscription licensing. The only downside was that we had to walk through the bubble caused by front-loading the revenue-recognition hits that ranged around a negative 30 percent. The second reason was the tools weren't out on time.
In 1996 Sematech started exploring the connection between EDA and the semiconductor world, finding that the EDA tools were tied to two process cycles. That is, every two process cycles the designers needed to swap out a significant part of their design tools. What had happened was that the EDA industry had settled into a five-year R&D cycle. That worked well in the 1980s, but unfortunately the automation of semiconductor process development had shortened the EDA window to two or three years.
As the EDA industry raced to shorten its R&D cycle it ran out of R&D engineers; the shortage is estimated at 30 percent to 50 percent. No R&D engineers means no tools out in time.
The countercyclical sound bite is an attempt to describe the major design cycle. First of all, we should probably stop calling this the semiconductor cycle because, although semiconductors are by far the best indicators of a design cycle, they are not the cause.
Actually, supply-chain management tools have been doing their job fairly well. Except for some DRAM overcapacity problems, the minor inventory corrections have disappeared from the semiconductor world-that's why we had an unprecedented period of prosperity. What the cycle might be called is the idea cycle. We have run out of ideas, or killer apps as they are sometimes called. The Internet appliance boom that was to displace the PC hasn't happened.
Same problem with the third-generation, or 3G, cell phone: No one really knows what features the consumer will be willing to buy. That is what the design cycle is all about-new ideas. When things are going well, system companies tend to design incremental improvements to their product lines. They don't really have the time, or the inclination, to design the next thing.
When the downturn comes they have to design themselves out of the downturn; that is, they have to design fresh, breakthrough products or lose market position in the upcoming cycle. This means heavy investment in EDA tools for two design cycles. The two design cycles are somewhat staggered between the power users and the mainstream users, the two areas where the money is and that usually run two to three years.
There is another design cycle: the two-phase design cycle. That spending is actually exceeding spending for the idea design cycle. Those tools, primarily the IC implementation tool set, were two and a half to three years late to market. They are the main reason we've been seeing a 20 percent demand forecast. But those tools are now on the market and will be the major portion of the 20 percent growth forecast for the EDA industry in 2001.
Actually, there is a third cycle: Every 10 to 12 years the design community has taken a jump in methodology. This time the jump is to the electronic-system level (ESL), a jump that gives the designer an order-of-magnitude increase in design productivity. We also tend to see an explosion of new systems, just what we need to lead us out of the bottom of the idea cycle. What the EDA industry sees is that the designers will spend five times more for their new ESL tools than they did for their old (RTL) tools. That will drive the EDA industry to a $10 billion to $15 billion level. Again the tools are late; we won't see them till 2004.
Once you look at the assumptions, the forecast isn't as crazy as it first seems. EDA has become the candle that is lighting the darkness of a bad semiconductor downturn.
Gary Smith is Senior Analyst for EDA at Gartner Group/Dataquest, in San Jose, Calif.
Return to 2001 Midyear Forecast