My blog about EDA being a second class citizen generated a number of responses that show that EDA has much public relations work yet to be done, in addition to its quest to enlarge its total available market and its profitability.
To begin, it was pointed out that it was wrong of me to state that IBM does not consider EDA companies as partners, since it has obviously collaborated with a number of EDA companies at various times. Of course I did not write anything of that kind - I merely pointed out that Adelio Sanchez of IBM did not mention the 35 EDA companies exhibiting at the Common Platform Technical Forum while describing the need for partnerships in order to succeed both technically and financially at the 65 process node and beyond. Since the title of the Forum was "Optimizing Silicon Through Collaboration", and since it is well known that all semiconductors foundries are certifying design flows at both 65 and 45 nanometers in collaboration with EDA companies, the failure to mention the EDA companies participating in the event caught my attention. And the original piece does state that such failure is the responsibility of the EDA industry, not anybody else. EDAC has for years tried to change the perception of EDA with its: "Where Electronics Begins" campaign, with little direct measurable effect. This is not due to the quality of the campaign as much as to the fact that the financial model for the industry has remained the same. The EDA industry behaves as a tools supplier, not a direct participant in enabling the growth of the semiconductor industry.
IBM and EDA
In order to dispel any misunderstandings about the role of IBM in EDA, let me state that IBM's own research efforts have produced insights and tools that have been fundamental in the development of products by EDA companies.
But the fact that the EDA industry's margin of profit is small, and the opportunity to increase market size quite limited, can best be illustrated by IBM's own business behavior in the sector. The company has more than once entered the EDA market as a vendor, the last attempt in 2005, and has relatively quickly abandoned all of those efforts. Not because its products were technologically uncompetitive, in fact they always were either the best or on par with the best. IBM abandoned the efforts because the profit margins in the EDA market were significantly below what the company could achieve by redirecting its resources to other endeavors.
As an example of the excellence of its technology that did not result in a market success, let me point out that in 2005 IBM won the EDN innovation award with its EinsTimer Statistical Timing tool that it exhibited at the 2005 DAC as a commercial product. The technology was so advanced that industry analysts speculated that it would replace Synopsys' PrimeTime as the tool of choice of IC designers. One year later, at the 2006 DAC, IBM did not exhibit any EDA product for sale. A significant portion of the static timing analysis technology was later incorporated in a Cadence product as a result of a collaboration between the two companies. IBM allowed some intellectual property from the research that led to EinsTimer to become public domain and was freely used by both Synopsys and Magma in products they now sell. In fact it is that intellectual property IBM gave to the industry that was at the center of the latest, now settled, litigation between Synopsys and Magma.
Low barrier to entry
One reader sent me email about my statement that the EDA industry must find a way to generate more profits. It says in part:
"Lets face it, EDA is a industry with $499 entry barrier. Anyone who
has a good idea can buy a laptop and start a company. With such
abundant supply of solutions, why would the IC makers pay more? The
customer can choose between innovation and better engineering from
small startups or "total solution" from big guys. It suits the user
community just fine."
What the reader points out was definitely true for most of the existence of the EDA industry, but, as systems houses have moved to the 130 nm process node or smaller yet geometries, the investment to successfully gain an entry in the EDA market has become significant. Technology alone is no longer sufficient, because point tools must be integrated into a flow or they will not be considered by any large customer. And without a large customer, people will just not pay attention to the tool. In addition, even if a startup succeeds to convince a significant potential customer to evaluate the tool, it must have enough funding to be able to exist during the evaluation which in most cases will take months, and provide suitable support to the customer. So the barrier to entry has gotten much higher.
Not only is difficult to bootstrap a new EDA company, but it is also difficult to find venture capital for a startup. Since the dot-com bust of 2000, venture capitalists have been more discriminating in their choices as to whom to invest with. And mergers and acquisitions are becoming scarcer and less profitable as well. No EDA company has gone public this century, even those, like Apache Design Solutions, who can show a remarkable string of profitable quarters.
All this says that although in theory the technological barrier to entry is low, the organizational and financial barrier has gotten much higher. The number of successful EDA startups is diminishing, not growing.
Is it really low quality?
Another email shows how some segments of the EDA customers base still prefer to maintain an adversarial relationship with their vendors. A reader who asks for anonymity writes:
"Financially speaking, EDA industry has managed to escape every downturn that the semi industry had to endure. The business models that they have created seem explicitly designed to avoid any resonance with the semiconductor cycles, and thus keep their
profits increasing even as the rest of the design industry goes into turmoil. Not only do EDA companies not share in the downturn, the industry psyche is to even avoid accountability. How many times have you seen EDA companies take the blame and
pay-up when an undetected logic error causes a million-dollar mask to go down the
I am sure that in the entire history of EDA there have been some first silicon failures due to a problem in an EDA tool. The development of an IC is a complex matter, and the opportunities for errors abound. I am not aware of any settlement between a design house and an EDA vendor involving legal proceedings for negligence on the part of the vendor. It is true that the license to an EDA tool comes with a disclaimer regarding the existence of bugs, yet such disclaimers would not be an insurmountable obstacle to a legal team that could prove gross negligence on the part of the vendor.
EDA tools are just that: tools. It is up to designers to chose them and use them wisely and appropriately. Some tools are difficult to use, some require knowledge of the semiconductors manufacturing process, or with specific requirements of the end-users market.
It is true that the EDA industry can be defined as being anti-cyclical with respect to the electronics industry. Statements to this effects have been made more than once by high profile representatives of the EDA industry. Walden Rhines, CEO and Chairman of the Board of Mentor Graphics, and Aart DeGeus, CEO and Chairman of the Board of Synopsys have pointed such financial behavior themselves. The speeches were directed toward proving that investments in EDA were sound and desirable when speaking to an audience that contained members of the financial community. But one can make numbers say almost anything one wants, so a closer look at those numbers is warranted.
Making sense of numbers
Since Dataquest exited the EDA industry analysis field, the only continuing sources of financial analysis left are the EDAC Market Statistics Survey (MSS) and the reports that Gary Smith issues, similar to what he did for Dataquest for many years. Both documents report revenue, not profits. Using the data from both reports it is clear that during period of recession in the electronics industry EDA revenue have not dipped as much as those of semiconductors or electronics products suppliers, but they have also not grown significantly. The strength during down cycles is due to two factors: it is cheaper for customers to continue to renew EDA tools licenses than to discontinue them and pay later to purchase them again at a potentially higher price, and the need to develop new products during the recession in order to be ready when the market turns positive.
The growth in EDA revenue does not directly translate in similar growth in EDA profits. In fact, no publicly traded EDA company has ever paid a cash dividend to its investors. EDA companies must reinvest profits in developing new tools, upgrading existing tools, and strengthening marketing and sales organizations. A close look at the reports of publicly traded EDA companies shows that, in many reporting periods, when results from operations are cash positive they are still book negative. This is mostly due to the fact that accounting rules require to report as expenses forms of compensations that do not use corporate cash like stock options grants, depreciation of assets, or some forms of mergers and acquisitions. Of course the list above is not meant to be exhaustive or complete.
Since the numbers of Independent Device Manufacturers (IDMs), independent semiconductors foundries, and large systems companies is shrinking, EDA revenue can only be grown by one of three methods: increasing unit prices, creating new market segments, or changing the revenue model. The first methods is the easiest to try of them all, and it has been tried almost yearly since the inception of the industry. It has had very little impact on revenue growth, mostly due to price competition among the largest three (now four) EDA vendors who control anywhere between 75 and 80% of the market.
Electronics System Level (ESL) and Design For Manufacturing (DFM) are two of the latest market segments created by EDA vendors. Of the two, DFM has produced the best financial results so far, mostly due to the fact that EDA vendors understand the requirements for the market better than those for ESL.
That leaves only one option: changing the revenue model. My blog last week suggested that if we could properly measure the financial impact of specific EDA tools on the revenue of products they help develop, we might be able to establish a revenue sharing method that would allow EDA vendors a more direct participation in the electronics market which is orders of magnitudes larger than the EDA market. Of course losses would also be shared by the industry, but the larger potential for profits would justify the risk.