Re: $150M for a fab
I have discussed a similar article with Drew which reportedly could slash fab costs to 1/10th of what they currently are. He used ASYT as an example, but the answer would apply to AMAT or any other equipment maker. Here is Andrew's reply:
===================<<Below is a crosspost>>
*AV*--Certain things are now making sense relative to slashing costs to 1/10 of what they are currently.
First, would this impact the business of ASYT?? Answer is yes but it does not put them out of business. Cluster tools have been around for awhile and the concept discussed in this article is a further refinement of this type of process equipment technology.
If I could use a slightly poor exaggeration: The goal of the IC industry is to move towards and attain perfect yields(both device and manufacturing) in the shortest amount of cycle time and at the lowest possible cost to manufacture advanced device technology. With this said, the ultimate goal is to place the bare silicon wafer into a "black box chamber" that does all the depositions, diffusions, implant, etches, and patterning within that one chamber, driven by the computerized download instructions from the previously simulated comput aided design library. Sort of one stop processing where all manufacturing functions required are accomplished within one tool.
The above is the perfect world. Instead of isolated individual processes and process equipment, you are trying to cluster both the tools and the process into as few interrupted steps as possible using the least amount of human intervention and equipment transfers. Much like Moore's Law for device density, the equipment and processes have been moving towards reducing the amount of individual process steps. Unfortunately the more advanced devices, more highly integrated device functions, number of functions, and the desire to shrink the size of the device have actually caused the number of process steps to increase. It used to be 1-2 levels of metal with an interconnect level. now we are talking about 6-8 levels of metal to achieve densities and functionality. Suffice to say, the complexity of the devices made today are phenomenal compared to what was being produced in the same area or volume of silicon 15 years ago.
Dr. Ohmi is a very well respected individual and I take seriously everything he shares with the community. However, we must keep everything in context. There has always been a drive to simplify the process by integrating more and more of the process steps within the equipment base. A basic example is the resist strip process. Way back when, we used to have 2 sulfuric acid - hydrogen peroxide wafer bath tanks that we placed the wafer cassettes into. The first bath provided the initial crude strip of the resist while the second bath was a clean up strip using a less contaminated acid mixture. Once it came out of that bath, it was transferred to a water rinse tank to clean off the residual acid. From there it went into an overflow DI water bath where it was rinsed to resistivity and then placed into a spin dryer. Without going into all the subsequent iterations of this process equipment set, including some of the automation of this process, we moved to the present process where the wafers can go into an FSI integrated batch system where the cassettes are loaded into the chamber and all the aforementioned processes are done, one right after the other thereby decreasing human intervention, floorspace, and chemical consumption. There are other examples that can be given that have given rise to some of today's cluster tools.
As the number of tools decrease, the number of items that ASYT would deal with does decrease. However, floorspace in a wafer fab is always utilized whether it is for new technology or additional capacity. The existing fabs, if retrofitted, will undoubtedly put in as much equipment as can be fit in to the existing space to create the maximum utilization of the floorspace. Therefore, we should look at ASYT as being more dependent on the amount of square footage available. These cluster tools will undoubtedly utilize ASYT integrated indexers and you could make an argument that the number of ASYT products per wafer process might decline with the integration of more process steps within each piece of equipment. However, there will probably be more of these pieces of equipment because as the number of processes increase the wafer throughput of the equipment decreases. Therefore, to make the required number of wafer outs, you would need additional equipment. Finally, these tools will incorporate some ASYT "robotics, PODs, transfer units, integrated indexers, electronic tags, etc." as a function of the number of wafers processed, the size of the wafer diameter, and by virtue of each of these tools having both an input and output location for the wafers needing to be processed.
Now, just to go out into left field for a second. Even if less equipment is required, and even if less floorspace will be required, and even if less new Fabs will be built, and whatever what ifs we might add to this equation, there are a few items we must be reminded of.
1. The 300mm technology will required new forms of wafer transfer since these wafers are HEAVY and still must be transported from one cluster tool to another.
2. These 300mm wafers are going to be extremely expensive to process so you want to make sure there is a minimum amount of "threats" to these wafers in the form of misprocess, human intervention, breakage, etc.
3. The smaller the fab (size and capacity), the more the brick and mortar costs factor into wafer costs, especially for environmental controls.
4. As devices shrink in dimension and get more complex, ambient environmental contamination becomes more of a factor. It is getting to a point where the breakdown of the ULPA filter media in ballroom Fabs are causing contamination of their own. PODs whether they are vacuum or inert gas injected ares still one of the only ways to transfer wafers from one process tool to another to eliminate the ambient wafer fab environment that can add numerous types of defects.
ASYT addresses much of the above by not requiring as much environmental infrastructure for new or existing fabs since the controls are placed around the individual process tools and the wafers are protected within the self contained environmental sound PODs.
This article is just a natural progression of things as technology moves on. there is nothing to fear here. ASYT may be impacted slightly, if at all. they will make it up with other customers, retrofitting older fabs or equipment, or make the appropriate adjustments to accommodate the change. ARM sales might suffer but integrate indexers and the rest of their factory automation products will still be highly sought after.
Back to the lower costs. Keywords were lower process costs. I take this to me the raw material costs since they are talking about re-using or recycling of chemicals and gases. This is true and there are many examples of lowering the consummable cost of manufacturing. We have sulfuric acid and HF reclamation/recycling/replenishment systems along with other things. The more you integrate processes and use the cluster tool concept, the lower your costs will be for equipment components such as the gate valves mentioned and gas usage. The lower the number of loadlocks the less nitrogen you will use for purging and pumping through chambers.
Do not think for a minute that if a cluster tool or advanced tool like being mentioned uses less components, that the price of that tool will go down accordingly. This is an industry where you pay through the nose for such conveniences. As MTBF or Process Uptime improves, you pay for that. If your spare parts inventory can be reduced, you do not get a break on the service contract. All these efficiencies really translate to better margins for the equipment manufacturer since they do not have to have as much spare parts inventory nor field service staff to cover a multitude of different equipment.
I want look at the cost of a process wafer broken down into 3 basic components (it is rarely done this way these days) of Materials, Labor and Burden with Burden including depreciation and fixed costs. Looking at it this way you would realize that Labor is the cheapest component, followed by Materials with the cost of the wafer substrate itself being one big chunk of the Materials cost. The last time I did wafer cost based on these 3 simplistic components, it turned out that Materials and Labor combined, made up less than 15% of the entire wafer cost. Therefore, I do believe that what is being talked about in the article only relates to the process costs which is more than likely the cost of Materials to process and not necessarily the equipment costs associated with manufacturing process. After all, the $1 million stepper of yesterday are going to be the $7 million DUV systems of tomorrow. And not to be forgotten are the Cost of Ownership for each of these systems. I do believe that the maintenance costs of the DUV stepper will far and away exceed the maintenance costs of the old steppers on a per wafer processed basis.
Lowering of Costs as mentioned in this article has shed new light on this entire subject we have discussed recently relative to some of the new technologies. I am not terribly worried about what is in this article since every new "invention" in this industry usually comes with a higher profit margin for the company producing it<GG>.
BTW-This is one man's opinion and all comments are welcome.
Andrew
|
