Semiconductor Industry Status June 2008

Has Intel artificially inflated microprocessor pricing?

With the recent news that Intel is being fined $1.45 billion dollars by the European Union there has been a number of opinions offered on both sides of the issue. In this article we will present our take on one of the economic issues raised by the case.

As an analyst focused on cost and economics issues in the semiconductor industry I have been analyzing cost per function trends for many years. One chart that I update every year looks at yearly pricing for each of the following: one megabit of DRAM, one million transistors and one million instructions per second of Intel microprocessor power. My data set for these prices goes back to 1970. In the last six years I have also added the cost per one megabit of NAND.

I think that most people reading this article would agree that DRAMs are a highly competitive business with multiple players all aggressively reducing costs and driving down prices. In fact the word “cutthroat” comes to mind. If you plot the price per megabit of DRAMs from 1970 until today on a log plot it falls on a 35% per year price reduction line.

Intel on the other hand is being charged with anti competitive practices and harming the consumer, presumably by stifling competition to artificially maintain higher prices. There are many details to the arguments going on about Intel’s behavior that I am not competent to comment on including whether Intel has broken any laws. What I do believe I am competent to comment on is whether Intel has reduced the price of their microprocessors at a rate that would be consistent with a highly competitive environment. There is no denying that Intel has a dominant market share, but if they are reducing prices at a rate consistent with other highly competitive industry segments I have a hard time believing they are artificially propping up prices.

There is some ambiguity when you start taking almost 40 years of Intel microprocessors and assigning million instructions per second (MIPS) ratings to them. I have been tracking various performance metrics and correlating/converting them to MIPS for many years. My methodology over the entire time period has been consistent so that although the absolute values may be in some doubt, the relative values should be good. When I take Intel’s publically available pricing sheets and calculate the cost per MIPS and then plot it on the same log plot with the DRAM pricing trend I get a very interesting result. The trend is virtually identical! Specifically from 1980 to the present the cost per MIPS has come down at 35% per year. The 35% per year price reduction trend also holds for one million transistors industry wide.

Can I say Intel has or hasn’t broken the law? No I can’t. But what I can say is for almost 40 years Intel has been reducing the price per MIPS of their microprocessors at the same rate that the DRAM manufacturers have been reducing the price of a megabit of DRAM and the overall industry has been reducing the price per million transistors. This at least to me makes a strong statement that microprocessor pricing has not been artificially inflated by a lack of competition.

This opinion is based on an analysis included in the IC Economics Report.

Related Products

IC Economics Report - a full color annual report covering IC Cost and Economics as well as the IC market, materials, capital costs, 300mm and 450mm wafers, design and operations. The IC Economics report is an excellent companion product to the IC Cost Model. A discount is available for purchases of the cost model and economics report. ORDER NOW!