My impression is that they whole world believes there is a mad rush to 0.25u so naturally the Taiwan foundries need to put that technology in place for the customers that will need it. typically a foundry will provide services for the multitudes of fabless semiconductor companies and those companies with fabs that need additional or start up capacity for the new technologies. Being ready to service these needs is where the foundries have been able to garner real good profits.
A fabless company only needs the design libraries, modelling, synthesis, and device simulations prior to committing first silicon to prototypes. They do not need to spend 10s of millions of dollars or more (much more) to bring up the manufacturing processes required for the production of these devices. They use foundry services for this. Once the prototypes have been debugged and samples sent to the end user of the device, these end users test and debug their electronic unit in which this device goes. From there you have the production ramp up of both the device and end product. Based on market conditions and Real End user needs and desires, the production flow of the IC manufacturing occurs.
The same holds true for the companies that have internal fabs. The design function in many places is restrained by the manufacturing capability of the internal fabs. Much more time is needed to develop and new manufacturing process and implement the new equipment required to run these processes than it is to design and simulate new libraries and device functions in advanced technology or smaller feature sizes. You also need to factor in the Capital Appropriations cycle, the Purchasing Department Equipment Negotiations, Lead time for the equipment, delivery and installation of said equipment, and the implementation, testing and qualification of both the equipment and the process at the wafer fab and at the end customer site.
Device qualification can take months from the end user. Couple this with the lead times of equipment and the implementation of the equipment in the wafer fab and it runs into months, if not quarters to a year.
While a limited capacity might even exist at these places to do prototypes, they may not be sufficiently capable of ramping production as fast as they would like. Hence, they use a foundry partner to accomplish this initial ramp or prototyping until they can get their own operation in place to move forward.
Now a little history (rather crude) of Foundries. they first came on the scene as low cost "me too" players to supplement existing or older technologies. Usually, with the exception of the fabless companies, they ran the older technologies for companies as fabs were switched to higher technologies. One can see that this put the fabless guys at a slight disadvantage. slowly but surely, these foundries began to reach parity with the "Fabbed" companies to benefit all customers, fabbed or fabless. This was done to meet the over-riding need of "getting you product to market" first. So much so that certain companies have even funded these foundries with advanced equipment to bring them up to state of the art and leading edge. Therefore, we have entered into a new phase of the foundry business, leading edge and bleeding edge. In this case 0.35/0.50 micron can be considered leading edge and 0.25u is bleeding edge, becoming leading edge.
Why do all this??? Pure and simple. When capacity was tight (more business but not enough capacity to make devices) these foundries would make decent profits. When business was slow or at parity with the capacity requirements, the foundries would be beat up on price and profits would sink. You garner less margins when you are a "me too" provider since everyone shops around and drives your prices down. When you are state of the art or bleeding edge, you are able to maintain higher margins for your services and limit your downside in pricing wars. These same foundries are starting to desert the older technologies in favor of the higher margin advanced technologies just like their customers. The expansions and retrofits at these foundries were to put into place more advanced processing capabilities while slowly exiting the less profitable businesses. Sort of like INTC deserting the 133 MHz CPU business in favor of the 350MHz Pentium II business, allowing the
Therefore, Foundries COULD be considered a bellwhether as to what is going on in the industry, in general. Keep in mind that technology is moving ahead at a faster pace. Shifts from 3u to 2u to 1u were measured in years on the technology roadmap. Using Moore's law, over time this progresses at even faster rates. The shift from 1.0u to the numerous submicron technologies are now occurring in 18 months or less, on the average. These technology shifts are starting to occur at such a rapid rate that they are now shorter than the production implementation, reliability testing, process development, and customer qualification cycles. With the push towards ever shrinking technologies for cost effectiveness and device performance, no sooner do you qualify something for mass production than you have another emerging technology barely out of R&D.
NOW TO ANSWER YOUR QUESTION AS TO WHAT THIS MEANS.
Thr foundries geared up for producing 0.35u devices in full production expecting to see the shift from the 0.50u technologies and above. they also were furiously moving forward to supply the necessary capacity or capability to provide the next generation of products for their customers. Somewhere in the scheme of things, the high volume production of the 0.35u device technology has stalled. Either it is because there may be a shift to skip 0.35u and go onto 0.25u more aggressively, the market for 0.35u devices has not developed fast enough to require the need for foundries, the design to customer acceptance and ordering cycle for this technology has stretched out, or companies are concerned about the overall market conditions and whether they really need to pay the price for the improved performance. I have seen companies decide to design in older technologies and create some real cost effective products that meet customer requirements. This is done without having to spend huge sums on new equipment or foundry services. We have seen this in the form of 0.25u lithography. There are discussions as to whether i-line can do a majority of this instead of DUV.
IN all of this chaos, the new 0.18u technology is ready to come out of R&D while very few companies are in high volume production at 0.25u.
Technology may be moving faster than the manufacturing implementation cycle or the customers are taking a more conservative view of how fast to move forward. Who knows, but what is being mentioned is that the ramp up of the 0.35u is not happening as expected. The 0.50u and above are still strong and there is no real rush at the 0.35u. where these foundries could produce more devices. Bad news here since the pricing on 0.50u product is lower than the 0.35u product. This will affect bottom lines. IF there is a slow down at 0.35u, it is questionable as to how much the 0.25u process could be delayed.
Keep in mind that this is not gloom and doom. This is just a Foundry manufacturing perspective. If the in house capability to produce advanced technologies reduces the need for external foundry services, you have a very healthy situation for the US companies. Their margins start looking real good with the internal profits generated from advanced technologies and they are not turning over part of the profits to the foundries. Foundries cost money in the form of profits, when you use them. If you can maintain in house control of your advanced technologies you can maximize profits. Foundries for the most part are necessary evils. these guys rarely, if ever, have any intellectual property or designs of their own.
As foundries fail to ramp advanced technologies this can be seen as softening in the business and the first sign of a retracement in the homeland. A well managed company will have about a 20% foundry requirement such that in lean times, they can ramp down the foundries and jeep their fabs full.
Sorry I went on for so long and for any type of babbling. I started this reply early this morning and have been interrupted all day. I had to keep coming back to this so there are probably breaks in the stream of thought or redundant comments. It's free so deal with it<GGGG>.
Andrew |
