HSINCHU, Taiwan (via press teleconference) -- United Microelectronics Corp. believes it will be able to accelerate the availability of 0.10-micron foundry processes by a full year and begin pilot production in 2002 after reaching a technology alliance with IBM Corp. and Infineon Technologies AG, said a UMC executive during a teleconference with U.S. journalists today.
"As the semiconductor industry moves deeper into submicron processes, many fundamental materials and reliability issues will surface, and this will make it impractical, if not impossible, for any one company to independently bring these technologies to market," said Jim Kupec, president of worldwide marketing and sales for UMC.
To overcome those hurdles, IBM, Infineon and silicon foundry UMC announced a four-year agreement this week to jointly develop common logic processes for 0.13- and 0.10-micron ICs, which will have the ability to integrate mixed-signal circuitry and embedded DRAM on a single chip (see Jan. 27 story).These technologies will be the foundation of silicon foundry services offered by Hsinchu-based UMC under the brand name "Worldlogic."
The two-generation technology pact covers processes for copper interconnect and low-k dielectrics, and it will help stabilize design rules early for development of reliable, advanced ICs, said Kupec in the conference call with U.S. journalists. He disputed comments from rival Taiwan Semiconductor Manufacturing Co. Ltd., which released critical statements about the alliance from its chairman. During an investor conference call on Thursday, TSMC chairman Morris Chang said his foundry company had turned down an offer from IBM to join the alliance because there was little value to the partnership (see Jan. 27 story).
When asked for a response, Kupec suggested that TSMC did not reject the proposal and it was probably unable to close on an agreement with IBM and Infineon.
Kupec estimated that UMC will gain seven to eight months in making available mature specifications for design platforms using 0.13-micron technology as a result of the R&D alliance. IC designers will be given more details about the jointly developed 0.13-micron technology "in early Q2 or late Q1" this year, according to Kupec.
UMC believes the alliance will enable it to offer 0.10-micron foundry processes one year earlier than its previous plans. Kupec said this would be about a year-and-a-half sooner than 0.10-micron logic node in the 1999 technology roadmap from the Semiconductor Industry Association (SIA). He added that UMC is not yet able to say when 0.10-micron design information will be made available to foundry customers, but the company now expects to place that technology generation into pilot production in 2002.
Below is a transcript from portions of the press teleconference held early this morning.
(Q--Will this alliance accelerate the availability of 0.13- and 0.10-micron technologies from the partners?)
Kupec--Yes, I think it is safe to say that. I also believe that the alliance will in fact work together to really pull forward the 0.1-micron process for all three companies. Having the combined effort on the longer-out technology--meaning the 0.1--moves it forward for all three companies.
(Q--For 0.13, how much time is UMC gaining? And how much for 0.10 micron? What timeframes are these being made available?)
Kupec--For the 0.13, there are two aspects of pulling ahead. One is at the point in time we publish design rules to allow customers to initiate designs. That is pulled ahead about seven to eight months. In fact, it also provides a much more stable target in that it uses combined resources to build a process to a spec.
The actual production time of 0.13 is only pulled up several months because that is equipment-bound issue. We will have beginning production late this year--2000--on the 0.13-micron process...
We haven't set a design availability time for 0.10 micron but pilot production will be in 2002. It is about a year ahead of UMC's previous plans -- a year-and-a-half ahead of the SIA roadmap. We were planning to be ahead a little bi of the SIA roadmap prior to the alliance...
(Q--What fab facility this new process technology is targeted for?)
Kupec--We have two fabs we are currently building that are 0.13-micron capable. They are 8-inch fabs. Those will be the first two fabs to implement 0.13. They are fabs 8D and 8F. We will have the 0.13-micron process in both of those fabs since that requires the same equipment set. Which one we will put it into first has not been determined.
(Q--Do you see this partnership evolving into the possibly of capacity joint ventures?)
Kupec--We are not making any announcements on that, and in fact there are no joint venture deals in the works.
(Q--Do you know if IBM plans to offer these jointly-developed technologies to their foundry customers?)
Kupec--I don't know the specific details of what IBM's exact plans are in the foundry world. Although I do know in general IBM targets into the foundry world more specialty processes than generic processes. But you would have to ask them for the specifics.
(Q--Can you disclose how much investment is being made by each of the partners in this?)
Kupec--No we are not disclosing investment levels but I can tell you the planned for co-development is balanced among the three partners.
(Q--Are you basically licensing IBM's copper process under this agreement?)
Kupec--No. Let me help clarify. This arrangement is not a licensing type arrangement at all. This is a co-development. We are co-developing. Each party brings in their own intellectual property into the agreement and then uses it as a platform to move forward. We have our own copper process that we are running in 0.18- and 0.15-micron technologies. IBM has their own copper process that they are running.
We will meet and define the future 0.13 and 0.10 with a mutual agreement among all three partners of what the new process is and how to proceed.
(Q--Will this also include low-k dielectrics?)
Kupec--Yes it will. In general, the deeper sub-micron processes with copper are going to migrate along the line of having low-k dielectrics.
(Q--UMC has struck a deal with Hitachi for a 300-mm joint venture. Will these technologies go to other places than your own fabs, and will these process technologies eventually make their way to the joint venture with Hitachi?)
Kupec--That is an open issue in terms of where we will place the process.
(Q--So will all of your 0.13-micron technologies will be based on this jointly-developed common process?)
Kupec--All of our standard foundry fabs will be using the 0.13-micron process. The purpose is to help develop a stable platform standard so that at the 0.13-micron node we can give people a nice secure design platform to target.
(Q--Will designs move around from UMC's foundry to Infineon and to IBM? Will customers have multiple sources for production? That has been implied.)
Kupec--It is theoretical in some cases and in some cases it may be practical in terms of moving back and forth. However, I suspect that is not going to be an often-used business practice. It is very expensive to move back and forth.
(Q--Now that you have heard that TSMC had rejected this alliance, is UMC having second thoughts?)
Kupec--That's a very good question. We see it as a deal which was a very favorable deal to the three partners. TSMC's comments, I don't believe, necessarily mean the deal was turn down as a result of their own rejection. It may be an issue that they could not close on the deal as we have closed on it.
(Q--When do you really see the market for 0.13-micron technology becoming mainstream. We are barely getting to the crossover between 0.25 and 0.18.)
Kupec--The design availability is set for early Q2, late Q1. And those would be design layout rules and electrical device targets. The early adopters will likely be the only ones who initiate designs starting that early--the people who have long lead times.
Our production capability comes on line late in this current year--2000. We will build production capacity through the following year of 2001. And, I believe, late 2001 is when we will start seeing what I would call mainstream adopters and continuing rising through 2002.
So if you are talking in terms of a mainstream market, it will be well into 2002 before I would say there is large swings in volume to 0.13.
(Q--Will the three companies standardize on a common tool set for 0.13 and 0.10 for such things as copper processing? Will UMC's equipment be the same as IBM's?)
Kupec--Each company will choose their own tool sets for the processes. There is nothing to automatically tie tool sets together. However, you are correct in your assumption that often technology for major important tools tends to limit the choices, but it is not 'religious'--if you know what I mean.
(Q--What is the prospect that this agreement might be extended beyond 2003 and beyond the 0.10-micron node?)
Kupec--The agreement cover through the year 2003 formally and all parties believe that we would like to engage yet for an even longer period.