On April 6 the EDA Consortium (EDAC) announced the results for 2005 of its Market Statistics Service (MSS). Superficially the results provide more fuel to the pundits who have, especially of late, found that all was wrong in EDA. A closer look at the data results yields a more positive judgment of the industry.
The results at the macro level show that for the full year 2005 revenue totaled $4,575 million, a 3% increase over 2004. The total revenue for the semiconductor industry for the same year, according to Gartner Dataquest, was $235 billions. Sales of consumer electronic products in 2005 were approximately $125 billion according to RNCOS research. This figure includes segments that do not use EDA tools. The two largest in revenue generation are distributors and resellers as well as manufacturers that buy components and subassemblies from other companies and do not need EDA tools. It is thus clear that the total available market to EDA companies is much smaller than the total of the two segments. It is highly likely that users of EDA tools generated revenue of approximately $250 billion in 2005.
Thus EDA revenue is about 2% of that generated by its customers. A superficial analysis of these results would lead one to think that revenue barely kept up with inflation, that the EDA companies do not know or are not willing to follow a disciplined approach to pricing, and that drastic changes are needed to re-invent the EDA industry to get a larger piece of the pie. After all without EDA tools either there would not be any electronics industry. In fact revenue of the EDA industry has very little to do with revenue of the semiconductor companies or consumer electronics vendors.
The EDA industry is a service industry. It provides tools that are used in the design and development of products, whether those are standard integrated circuits, ASIC and FPGA devices, or printed circuit boards. EDA customers generally obtain the tools by purchasing a term license that spans between one and three years. The funds for the purchase come from the engineering budget that almost all companies keep well below 20% of the corporate budget for the year. When EDA companies sold perpetual licenses for their products, some customers also used the capital equipment budget to finance the acquisitions since they could depreciate the cost of the tool. But this can no longer be done, since the tools are now rented and thus no longer an asset. The result is that the amount of money available to purchase EDA tools and services is directly related to only two factors: the number of new engineering projects planned for the year, and the nature of each project. Only some of the projects will need more tools or new tools, while the majority will continue to use the existing tools. A license renewal or the purchase of additional licenses for a tool that has been in use by the customer typically generates less revenue than issuing a new license for a new tool. This pricing structure is not a perversion of the capitalist system by EDA companies, it is just common demand/supply behavior in a market that is driven by innovation in the manufacturing process every two years or so.
Since the tools are used in design and development, their price cannot be correlated to a ratio between the total revenue of the customer and the perceived value of the tools. Revenue is the results of many factors and engineering efficiency is only one of the components that impact total revenue. Marketing and sales have a much more visible and direct impact on revenue, thus arguing that EDA companies should be able to charge more for their products in view of the potential revenue they might enable is wishful thinking and not supported by corporate accounting practices.
Many critics of EDA companies point to volume discounting as the culprit for lowering total revenue in EDA. Specifically the four largest, publicly traded EDA companies that represent over 75% of total EDA revenue, are vilified as the perpetrators of unjustifiable discounting. But letâ€™s again look at what the capitalistic market requires. Prices are a function of demand, competition, and profit requirements. The customer must need the product, and it will always try to pay the least amount for it when a choice is available. Any EDA company, especially a public company, must show a reasonable return to investors or they will demand major changes in executive management at the minimum. Thus, it is evident that market forces are establishing license prices, not ignorance on the part of the EDA companies.
Sang Wang, former CEO of NASSDA and a director of Golden Gate Technology, wrote in EE Times that EDA companies have generally been behind schedule in delivering needed tools. This has resulted, according to Dr. Wang, in semiconductor companies developing their own tools internally, thus limiting the revenue available to EDA vendors. Unfortunately this section of the article is not very informative since it states that: â€œDataquest estimates that such in-house development investment increased to 27 percentâ€; 27% of what? The International Technology Roadmap for Semiconductors (ITRS) has consistently held that EDA tools lag the introduction of a new process in a research fab by approximately two years. Given the predictability of Mooreâ€™s law, it may in fact seem inefficient for EDA companies to wait two years before tools to support the latest process are widely available. Yet, once again, economic reality has the upper hand. It is useless for an EDA company to offer a tool before there is a reasonable market for it. No one would buy it. The two years time frame generally represent the amount of time between the validation of the process node and the volume use of the node. In the intervening time, EDA companies provide beta tools to early developers, and both parties benefit. The fundamental problem facing the EDA industry today is not that they are tardy, but that the processing nodes are too early for the average electronics company. In the next issue we will discuss specific EDA market segments in order to understand how they affected the results for 2005.