WSJ reports that Taiwan Semiconductor Manufacturing Co., the world’s largest contract manufacturer of silicon chips, would spend $12 billion to build a chip factory in Arizona.
I have been arguing on this blog for sourcing suppliers closer to the customer location (my first research paper was on dual sourcing and near sourcing) and building more plants in the US. So, I view this development as good news, but with a fair amount of caution.
It is frustrating. News reports always seem to score these decisions based on how the political benefit will play out and this WSJ report is no exception. (This is an election year in the US. Which year isn’t an election year? In addition, US Senator McSally is in an unusually competitive Senate race in Arizona).
Let us focus on operational complexity.
We covered a similar issue on the blog before. Recall Honhai Technologies, better known as Foxconn, announced a $10 billion plant to make LCD panels in Mount Pleasant, Wis. in 2017, but the operation has fallen far short of initial ambitions.
In the article, I argued that success depends on three factors:
- Efficiency and Distance to market: The eventual landed cost of manufacturing & delivery to the customer is sufficiently low, because the production facility is located within the consumer market.
- Product Characteristics: Product should be of high margin and healthy demand, to allow for learning period to improve the efficiency at the factory, and also overcome generally higher operational costs.
- Labor Supply: High availability of trained labor in the area of manufacturing.
Just to recall the evaluation on Foxconn (1.) was great and I welcomed it. (2) Foxconn’s choice of LCD TVs products was a terrible choice. (3.) Labor supply turned out to be a problem. Despite all the advantages due to nearby engineering school, and closed GM plant, finding trained labor was a problem!
—
Quick Evaluation of TSMC Plans.
On to TSMC. Can they succeed? First, some quick observations.
This will be a small plant.
TSMC said the plant would make 20,000 wafers a month, making it a relatively small facility for a company that made more than 12 million wafers last year alone. TSMC’s Fab 18 in Taiwan, which currently produces its 5-nanometer chips, was targeted for 100,000 wafers a month when it broke ground in 2018.
They will be making 5mm nano-chips. How long are they going to be in vogue? Well …
TSMC’s plant would likely not be at the leading edge of chip-making technology by the time it begins production, if it manufactures 5-nanometer chips as planned. TSMC has already started making 5-nanometer chips, and has plans to move to 3-nanometer transistors and smaller in the next few years.
1. Distance to market: Good. Arizona is close to downstream demand in California (typically) and other factories in NM and AZ, so this is a good thing. Many US chip companies, including Qualcomm, Nvidia, Broadcom, etc, rely on TSMC for chip manufacturing for many of their production.
2. Product Characteristics: This is truly worrisome. This product choice is essentially a repeat of the LCD problem. By the time this plant is set up for production in 2024, I expect the demand for 4-nm chips to have gone, poof. Essentially it will become a low-margin product, made at low-volumes, and would be shuttered as 5nm goes away. Does TSMC plan to introduce 4nm production in phases?
3. Labor Supply: The supply here for the product is actually better than Mt. Pleasant Wisconsin case (which had poor supply trained labor, and a blind railroaded process). Intel has a nearby plant in Arizona, and given the small size of the TSMC plant, it is possible to acquire Labor competitively. (In any case, trained labor supply will take some years to establish).
I am actually hopeful, that this is an improvement, but I am not convinced about the product choice. Needs a deeper look. I will take return back.
In all of this mix, is Huawei. But, that’s another complicated story. We will return back to this topic soon.