Why DRAMs are Like Steel
Over lunch today I had a conversation with an alum of McKinsey Consulting who remarked that the DRAM business behaved in a way that was similar to the McKinsey Steel Model. For those unfamiliar with this model I found a slideshow HERE that refers to it a good deal. (So far I have not found a tutorial on the model itself, but if anyone knows were to find it The Memory Guy would highly appreciate hearing about it.)
One interesting thing is that this particular McKinsey alum was not the first to point this out to me. About 15 years ago a family friend/McKinsey alum told me exactly the same thing. It seems that the economics of the DRAM business have changed little over the past 15 years, and the McKinsey steel model applies to DRAMs just as well now as it did then.
In a nutshell, the model posits that the market price for a ton of steel is equal to the production cost of the least efficient supplier. Today the DRAM version of this would make Elpida out as that company.
The graphic illustrates this. The columns represent the production costs per ton for a number of steel suppliers A-H. In DRAM this would be the cost per gigabyte of various suppliers. We use gigabytes because that’s all that purchasers care about. They are more than willing to use either 1Gb or 2Gb chips – whichever is cheaper. Clearly producer H has a very large cost disadvantage compared to producer A.
The dotted lines represent market price at various times. When the price is high (as shown in the top dotted line) all producers profit. When prices reduce (the middle line) producers F, G, & H lose money while the others continue to profit. When prices fall very low, as is illustrated with the lowest dotted line, very few companies profit. The companies who don’t profit will be forced to consider exiting the business.
Elpida is now confronted with bankruptcy based on its costs, not just today’s costs, but over the past few years as profits have eluded DRAM makers shrinking each company’s balance sheet in proportion to its cost disadvantage. Taiwanese DRAM manufacturers are also suffering since their balance sheets are weak. The DRAM market will come out of the current market downturn with fewer participants because of the consolidation that stems from this and other market factors. (See the Objective Analysis brief: Why the DRAM Market Must Consolidate, which can be purchased online for immediate download.)
Elpida’s higher costs stem partly from the strength of the Japanese yen. This has forced the company’s production costs to be artificially high, but that’s not artificial enough for Elpida, which is going through a very real bankruptcy proceeding.
The steel model slideshow mentioned above also points out that the profits of the most efficient producer are very low if all makers have relatively similar cost structures. This is certainly the case with DRAM! The small differences between manufacturers, though, are enough to drain the balance sheet of one vendor while bolstering that of another, and that will either allow or disallow each manufacturer’s investment in the next-generation technology. If the company cannot invest in the next generation then it will have to compete at a disadvantaged cost until it is forced to exit the market.
DRAM price dynamics are similar to those of NAND flash, which is also entering a difficult phase. Great detail about this market is available in another Objective Analysis report that can be purchased online: Understanding the NAND Market.