Why TSMC should buy Cadence: Part 2 letters for and against ...
Before you read the letters, please note that I had a chance to speak briefly in person this week to Rick Cassidy, President of TSMC North America. He gave one of the keynotes at ARS 2005 (The Advanced Reticle Symposium) at the San Jose Convention Center on June 21st. I posed my question to Rick after his speech: Rick gave a big hearty laugh and said, "No! TSMC is about deep collaboration with our partners very deep collaboration!" I told Rick that I had received a number of letters in response to an article I had written and that many people sensed there was something in what I had to say that the foundries are poised to make a move into the EDA space. Rick gave another hearty laugh, then adamantly reiterated his statement: "Well, thanks for writing about us, but TSMC is only involved in deep collaboration with our partners. Very deep collaboration!"
Letters to the Editor ********************************* Letter No. 1 - The cafeteria factor Peggy, I have been [thinking along these same lines since last year], but I have other reasons that you did not cover. At DAC 2004, [there was] a meeting addressing "RTL handoff" in which people talked about handing off designs to the foundries at the RTL level. Two conclusions were obvious at that time:
Faced with that [situation], the definitions of front-end and back-end change. If RTL handoff becomes real, back-end now includes everything from RTL down, which includes synthesis, verification, place & route etc. And this becomes a service of the foundry especially as designers wish less and less to deal with these issues. Couple this with the economic warnings about the EDA industry (less ASIC starts, more complex tools required meaning more expensive tools that less people can afford) and the foundries have to make sure that the tools are available for their customers. And a good tool flow will become a competitive advantage for the foundry. Couple this with the need for the EDA companies to know more intimate details about the processes, which the foundries are less willing to give out, and you have only 1 conclusion the foundries will need to buy parts of the EDA companies. I see TSMC and UMC buying parts of Cadence, Synopsys, Mentor and Magma. [So], in fairness to the other side of the question, I don't know if TSMC will buy all of Cadence. They need parts of all the EDA companies (i.e., they need synthesis, so they need Synopsys). If they buy Cadence, will they lose Synopsys customers? Bottom line I'm not sure how it will play out, who will buy which parts of who and how that will be done, but I believe foundry purchase of EDA capabilities to be inevitable. To your question of when I suspect they are waiting to the last minute. As the EDA companies fall on harder economic times, it will become cheaper to buy parts of them. Regards, Anon ********
Peggy, My vote would be for TSMC to buy Synopsys. They've brought in a lot of execs as well their new Senior VP of Marketing Jay Greenberg, and other people behind the scenes they aren't talking about. And Synopsys has the tools, as well. I think they are a stronger company [than Cadence] from a technology standpoint Regards, Anon ********
Peggy, TSMC buys Cadence? Good for creating buzz but not realistic. You obviously know Cadence very well, but have you ever been to Taiwan? Reminds me of the time Avanti bought Meta Software (HSPICE). They gathered each company's engineers into a large room with a divider. The divider was removed and on one side there was China and on the other [side] was Berkeley, California! The silence was deafening and I have never seen Birkenstocks move so fast as when the Meta people left Avanti. Cross-cultural psychology should be taught at DAC each and every year. Anon ********
Peggy, I don' t buy an acquisition by TSMC. TSMC does good business and they've got a clear focus. They play the various EDA vendors off against each other and their loyalties are clear. If they bought Cadence, they'd immediately [be in] a cold war with Mentor and Synopsys who are the two real leaders in design-for-manufacturing. At the very point when they'd need EDA skills the most to drive toward 65 nanometers, they'd hit black ice. On one hand they'd expect to partner with Mentor and Synopsys, on the other hand they'd be selling against them. Not a good thing. These kind of combinations are extremely dangerous because of the risk of destroying the web of supplier relationships that keep them on top. I just don't see anyone buying into the EDA space. If it were, I'd look for one of the big software companies, probably in the CAD/CAM space, but the problem is simply that you can't get synergies (read, I can fire lots of people when I combine the companies and still have the same amount of business) with these type of buyers. The sales forces don't overlap. The development teams don't overlap. You can probably squeeze a little in economies of scale out of corporate center functions (Legal, HR, IT, Marketing, etc.), but not enough to really make a good business case for the deal. The only other place I could potentially see it happening is [through] a semiconductor equipment company. They sell to the same general customers (though very different people) and might think they could squeeze some synergies out of sales. I don't think they could, but someone might talk themselves into it. Regards, Anon
Why TSMC should buy Cadence *************************************** [First published June 21, 2005.]
TSMC should buy Cadence. What kind of evidence did I find at DAC to support my conjecture? Lots.
Means ... First of all, let's look at the numbers (per Yahoo Financials): Taiwan Semiconductor Manufacturing Co. Ltd.
Cadence Design Systems Inc.
Conclusion
Motives ... It's easy to site numerous motives that TSMC might have for buying Cadence: * No. 1 On June 10th prior to DAC, I had a long phone call with Ed Wan, TSMC's Senior Director of Design Service Marketing. He was explaining TSMC's new DFM Toolkits Yield Plus & Yield Pro. "Yield Plus, developed by TSMC and its EDA partners and implemented at the designers discretion, includes Action-Required rules, Recommended Advisories and Guidelines for semiconductor design. Also included are DFM utilities to implement the rules and advisories." "Yield Pro includes several unique services that are implemented by TSMC at the manufacturing stage to quickly move a design into volume production. Yield Pro Services include a Lithography Process Check (LPC) Service, a Yield Sensitivity Analysis (YSA) Service; a Package Modeling Service; and sophisticated Scan Diagnostic services." The more I listened to Ed brag on the TSMC announcement, the more TSMC sounded like an EDA company. I told Ed that if it walks like a duck and talks like a duck, it sure seems like a duck Ed rigorously denied my allegations. He said TSMC is not an EDA company. But clearly TSMC is concerned about the flow; they're concerned about getting design for manufacturing considerations/data into the designers' hands; they're concerned that the job of designing chips be more thoroughly integrated into the job of making chips. TSMC's Toolkits "improve product yield, increase device performance and enhance semiconductor companies return on design investment Together, these services create a powerful methodology that increases yield at the design stage, and in early semiconductor manufacturing stages. This accelerates production ramps, enabling greater returns on dollars spent on design and manufacturing." What EDA Company today isn't saying exactly the same thing? Motive # 1 TSMC wants to be an EDA Company.
What EDA Company today isn't, for all intents and purposes, saying exactly the same thing? Motive # 2 TSMC really wants to be an EDA company.
Last year's panel included Mentor Graphic's Wally Rhines, Synopsys' Aart de Geus, and Cadence's Mike Fister. This year Wally, Aart and Mike were there again, but ARM CEO Warren East and UMC (United Microelectronics Corp.) CEO Jackson Hu were there as well along with Jay Vleeschhouwer from Merrill Lynch doing Podium duty. Hundreds turned out to witness the event and although many saw a high-profile DAC panel with the "3-minute" opening statements, per usual, running 10 and 15 minutes long I saw something more akin to a 90-minute Fellini movie, such were the conflicting realities between the left side and the right side of the stage. Aart, Warren, and Wally had a love fest seated to the left of the podium praising each other's tools, design flows, IP, interoperability, and mutual commitment to World Peace. They demonstrated their congeniality with on-stage smiles and hugs. It was heartwarming. Alternatively on the right side of the podium, Jackson and Mike documented grim realities instead. Jackson said it's a nightmare tools vendors, designers and foundries all pointing fingers at each other today and demanding that the other guy be responsible for getting things right. As a result, he said the foundries need to start thinking and acting like Virtual IDMs, knitting the entire semiconductor design-to-manufacturing value chain into one cohesive whole. He said it was the foundries who will drive the process who will insist on a greater linkage between design and manufacturing. Jackson said that's tomorrow's reality and he made it sound like a do-or-die proposition. Mike also seemed troubled by nightmares. He was not smiling when he said Cadence needed to pay closer attention to customer needs, to provide kits of tools aimed at specific end-user markets, and to pursue opportunities for technical adjacencies. As opposed to the left side of the stage, the world of design and manufacturing on the right side of the stage was not a happy place. It was a place of gritty, burdensome woes. Meanwhile, in the middle of the stage at the podium, Jay made it abundantly clear that things just aren't working in EDA on either the left or the right side of the stage. He begged the industry to improve its breadth, its growth, and its cash flow. Motive # 3 Foundries need to think and act like IDMs. EDA companies need to look to the larger world. Those agendas are both addressed if TSMC buys Cadence.
In other words, you're looking at something like the European Economic Block joining forces to take on The North American Economic Block. My theory, however, says that Cadence will leave the European Block to join the New World Block, with the caveat that the "New World" is no longer North America it's Asia, home to TSMC. Motive # 4 TSMC needs to counter the emerging "IBM/Chartered/etc. Economic Block" by:
Opportunity ... Having said all this, let's look at the opportunities that TSMC would be capitalizing on if it actually made the move to purchase Cadence.
Opportunity # 1 Cadence is supremely confident and what they don't know about M&A (probably) isn't worth knowing.
With numbers that large, for a company as small as Cadence, the only way Fister's mega-package can be explained is to presume that he's the catalyst needed to make a mega-deal happen. Clearly Intel's not going to buy Cadence or else why would so many Intel exec's have bailed from the chip maker to join Fister's EDA enterprise? On the other hand however, TSMC buying Cadence might be a great, logical, and do-able mega-deal. Perhaps Fister has what it takes to close such a deal, to create the ultimate "adjacency" for his company, and to solve the need for TSMC the foundry to re-invent itself as a "virtual IDM." Opportunity # 2 Cadence needs to prove to the stockholders of Cadence that the compensation package bestowed by the Cadence Board on the Cadence CEO is fully warranted. Being acquired by TSMC would provide that proof.
(Programming Jobs Losing Luster in U.S. TSMC is part of the Asian Equation. They play from an obvious position of strength in Taiwan, and a growing position of strength in the PRC and Singapore. They're well situated to take excellent advantage of the burgeoning intellectual capital represented by the legions of engineers coming on-line across Asia. Opportunity # 3 Cadence, like every EDA Company currently in existence, needs access to the Asian channels. TSMC would provide that channel.
Chuck and I chatted about TSMC's fabs in Taiwan, China, and Singapore. We talked about shrinking process geometries, capital equipment, libraries, IP, software, and international trade. I really enjoyed our chat because Chuck's a great conversationalist and a long-term and very loyal TSMC guy. He's also absolutely positive that TSMC is not looking to get into the EDA industry. Unfortunately, he didn't convince me. What I saw in talking to Chuck was a company that has grown in 15 years from start-up to its current position as one of the largest semiconductor chip manufacturers in the world. I saw a company that has been aggressively nurtured by a visionary management team that never said never, and still never says never. With Chuck Byers and TSMC exuding that kind of sophistication and poise, it's hard to believe that their next step won't be yet another daring one. Opportunity # 4 TSMC is supremely confident. There would be true synergy between two companies who claim confidence, breadth, presence, and poise as their watchwords.
Okay there you've got it: * Means. Ladies and Gentlemen of the Jury, TSMC has the means, motives, and opportunity to buy Cadence. Everything I saw at DAC 2005 helped to support that conjecture, spirited disclaimers to the contrary notwithstanding. Therefore, in my mind the question is not: * If TSMC will buy Cadence. The question is: * When will TSMC buy Cadence.
Peggy Aycinena owns and operates EDA Confidential. She can be reached at peggy@aycinena.com
|