By their own actions in the last few weeks, Mentor, Magma, and Cadence, have reopened the debate on whether or not IP belongs in EDA. Their answers respectively are: no, yes, and "on my own terms".
It was about a year ago that Michael Santarini and I had the last of a number of discussions on the topic: Michael was against the idea and I was defending EDAC's decision to include IP revenues in their Market Statistics Summary Report. My argument was based on historical facts. Long before the concept of Fabless Semiconductor Company was invented, Synopsys began to sell IP blocks under the DesignWare brand. If I remember correctly it was 1990, but certainly no later than 1991. The blocks were simple compared to what is available now, and the primary reason for selling them was to give a hand to designers still struggling with logic synthesis. Design Compiler was a tool at times difficult to use, both because it was new and because designers had never created synthesizable logic before. Will EDAC continue to include the IP market in the report? I think it depends on Cadence: read on.
It Costs Too Much
About a month ago Mentor announced that it was closing its IP division and selling the assets in a cost cutting move. Hal Barbour, president of CAST, remarked that to be profitable in the IP business a company needs to keep its costs as low as possible, using contractors and part time support personnel whenever possible. Mentor, with its full time staff of sales, design, and support personnel, saw its margins shrink. Mentor's principal IP customers were FPGA designers, and this market segment is very price sensitive, to put it politely. Or has someone has famously said before: "You cannot get blood from a turnip". Mentor was not only competing with the FPGA vendors like Xilinx and Altera that have a substantial portfolio of IP blocks in their inventory, but also with consultants and very small business entities that used IP as a leverage to win consulting contracts. As every one knows, or ought to know, IP and development tools divisions are not profit centers within FPGA companies.
The Analog Quest
On the other end of the spectrum, Magma purchased Sabio Labs, a provider of analog IP, at the end of February. This goes hand in hand with Magma entering the analog/mixed-signal market segment just a few weeks prior.
(see Magma acquires analog IP specialist Sabio Labs
and Magma Releases a Titan)
A few analysts have reacted negatively to this move, pointing out that almost all the startups targeting the analog IP market have failed.
I am not one of them. What is not widely known is the fact that Magma has also purchased all of the intellectual property of Barcelona Design, probably the best known analog IP company not to survive its startup phase.
If Synopsys' relative success in the IP market has anything to teach the rest of EDA, it is the fact that IP, by itself, is not a very profitable business for a large company, but can be a good business when it is synergistic to another corporate business unit. In the case of Synopsys it is logic synthesis, and in its own case, Magma believes it is as a differentiator in the analog market. And besides, who is to say that by gaining knowledge in the analog IP market, Magma may not finally find the key to success in the analog synthesis market?
Other People Secrets
This week comes the news that Cadence has acquired Chip Estimate, the company that had finally found a way to be profitable as a IP broker when everyone else had previously failed.
(see Cadence buys IP reuse specialist Chip Estimate)
When asked about their approach to IP in the past, Cadence has always maintained that it did not want to be in the IP business because doing so would mean competing with its own customers. And they have been very careful this time in positioning their acquisition by stating that the purpose for the move was to provide a better way for its customers to get access to IP providers, not to be in the IP business.
I certainly cannot fault Chip Estimate's stockholders for wanting a lucrative and quick exit strategy. But I am a bit sadden by the fact that the company passed up the real opportunity to become another Denali. A profitable independent company that can prosper from its own cash flow. What has happened to Chip Estimate is probably due to the difference between a bootstrapped company and a venture funded one. Venture capitalists, ITU and a small number of unspecified angels in this case, do not invest in order to get dividends.
Chip Estimate's business model underscored independence and equal treatment to all. It certainly had not shown any discrimination toward Cadence's customers! So, at first blush, it is difficult to justify Cadence's decision. Of course, those of you more versed than I in accounting mysteries may be able to point to some financial benefits accrued by this use of funds for acquisition versus some other possibilities. Others, more involved than I in the venture capital world, may favor the theory that Cadence owed a favor to Chip Estimate's backers. I prefer a third conspiracy theory: by having a complete picture of the traffic through Chip Estimate, Cadence now can know what some of its competitors' customers are up to.
The Chip Estimate portal can become a valuable tool to build corporate relationships with companies who would, otherwise, be more difficult to approach. Knowing the type of IP they purchase will also tell Cadence what designs they are undertaking and thus what EDA tools they might need: it' s market intelligence, baby! And I bet it did not cost an arm and a leg.