Communications of the ACM
Economic and business dimensions
The Value of Microprocessor Designs
Credit: Intel Corp.
Over the years the leading microprocessor company, Intel, has introduced a steady stream of new microprocessor designs: the 286, 386, 486, Pentium, Pentium II, Pentium III, Pentium 4, and more recently the Multicore design. In the microprocessor industry these designs are called microarchitectures. If there was a market for microarchitectures what would each design sell for? Our research addresses that modern question using economic insights developed almost two centuries ago.
Why estimate the value of these designs? This type of calculation can inform many aspects of firm strategy and valuation. Many companies develop new designs and increasingly outsource the manufacturing of the products. For example, fabless companies like Qualcomm and Broadcomm design chips and do none of their own manufacturing. Foundries do it for them on contract. The value of the fabless firms depends predominantly on the value of designs.
Many factors complicate any attempt to estimate the value of intellectual property associated with such product designs. In microprocessors for instance, consumers are not willing to pay for a new design per se, but for the increase in computing power that comes with a new design.
The classical economist, David Ricardo, had the key insight in 1817.2 Ricardo asked how much would be the rent to a unit of fertile land (say close to a river). Ricardo reasoned that producing a pound of corn in the fertile land is less costly to a farmer than producing a pound on marginal land (the worst land being cultivated, say in the hills). The fertile land requires less effort to achieve the same output. The rent for the fertile land arises from the difference between the labor cost of producing the same quantity of crop on the fertile land and on the marginal land. Charge the farmer a rent higher than this maximum, and the farmer would prefer to move out and start cultivating the marginal land.
Ricardo's logic still applies today, and can help estimate the rent to a new microprocessor design at any point in time. Think of a new design as analogous to fertile land and an old design as the marginal land. A microprocessor made with a new design can compute faster and hence sells for a higher price. To get the same revenue from selling microprocessors with older designs, Intel would need to sell more of the older microprocessors, something that involves more labor cost than making one microprocessor with the new design. The rent for the new design, therefore, is the difference between the cost of producing a dollar of revenue with the new design and the cost using the old design in a given time period. Adding up all such rent across the lifetime of a specific design provides an estimate of the value of the design to Intel.
It is not quite that simple in practice, of course. Increases in computing power can come from factors other than a new design. The decreasing size of transistors used in microprocessors is the leading example. Advances in semiconductor process technology have steadily driven down transistor sizes from three microns (where a micron is a millionth of a meter) in the original 8086 made by Intel to around 0.022 microns in the latest chip. These smaller transistors—by themselves—lead to greater computing power, without any improvements in designs. In other words, a proper estimate must separate the value provided to Intel by new designs from the value provided by the technological transitions to smaller transistors.
To summarize, Ricardo's logic still applies. It can be applied to the measurement of rent to a new microprocessor design at any point in time. A new microprocessor can be defined as a pair of attributes: the design it uses and the semiconductor process technology it was made with. The rent to the combination of design and process technology used in a new microprocessor is the cost savings the new microprocessor provides over the oldest one currently in use. If the new microprocessor uses the same process technology as the oldest one currently in use, then the difference in the cost of production is the rent to the design. On the other hand, if the new microprocessor uses the same design as the oldest one currently in use, the difference in cost of production is the rent to the process technology.
The accompanying figure illustrates the approach, using data from Intel's production for the third quarter of 1997. There were two designs in operation in the quarter, Pentium (P5) and Pentium II (PII). There were process technologies, 0.35 microns and 0.25 microns. There were three microprocessor vintages in production, (P5, 0.35), (PII, 0.35), and (PII, 0.25). The y-axis shows the average cost/average price of microprocessors of each vintage. The x-axis shows the total revenue obtained from each vintage.
What is the message of the figure? The (P5, 0.35) vintage of microprocessors were the oldest ones, and had a high cost/revenue ratio. Microprocessors produced with the same process technology, but with a new design (PII) had a lower cost/price ratio. The latest microprocessors featuring a new design (PII) and a new process technology (0.25) had the lowest cost/price ratio among the three vintages. The vertical distance between the top and the second horizontal lines is the cost saving provided by the new design PII over the old design P5, for each dollar of revenue that Intel obtained by selling the chips that used PII design. Hence the total rent to PII design during the quarter is the area of the rectangle shaded with vertical lines. Similarly, the area of the rectangle shaded with diagonal lines is the rent to 0.25-micron process technology during the quarter.
An estimate of the total value of the design or process technology to Intel comes from constructing similar diagrams for all the quarters in which a design or process technology was in use and by adding these up. The accompanying table shows the estimates we obtained in our study.1
The cost savings from new design is in many billions of U.S. dollars. One can see from the table that Intel's savings from new process technologies was almost three times the savings from new microprocessor designs, indicating the relative importance of new manufacturing technology transitions to Intel.
The table also includes information about the value of specific designs. The Pentium 4 design provided very little value to Intel. This is not surprising due to its problems with overheating, which forced Intel to move to the new multicore designs. The Pentium III was the most valuable design for Intel, reflecting perhaps the high price that new designs were able to command at the height of the Internet boom of the late 1990s. In that period, Intel used the 0.13 manufacturing technology, again the one that we estimated to have provided most value to Intel.
It should be noted that these values are calculated ex-post, after the microprocessors that used these designs were sold on the market. This method is not appropriate for forecasting value prior to any market experience, an important precaution in interpreting these figures. Intel probably spent the most on developing the Pentium 4 among all its designs, an investment that did not pan out for Intel.
Overall, these calculations provide a rough estimate of the value to Intel of intellectual property embedded in new designs. Ricardo's centuries-old wisdom on land rents turns out to be insightful for valuing intangible assets.
References
1. Aizcorbe, A., Kortum, S., and Pillai, U. Measuring Knowledge Capital and its Contribution to Productivity Growth. 2012, Working paper.
2. Ricardo, D. Principles of Political Economy and Taxation. London, U.K., 1817.
Footnotes
The opinions expressed in this column are those of the authors and do not represent those of the U.S. Bureau of Economic Analysis.
The Digital Library is published by the Association for Computing Machinery. Copyright © 2013 ACM, Inc.
Comments
CACM Administrator
The following letter was published in the Letters to the Editor in the November 2013 CACM (http://cacm.acm.org/magazines/2013/11/169039).
--CACM Administrator
The viewpoint "The Value of Microprocessor Designs" by Ana Aizcorbe et al. (Feb. 2013) aimed to analyze the value of microarchitectures in isolation, as though they could be mixed-and-matched with various silicon implementation technologies over the years. This is a nonsensical proposition. For example, a Pentium III microarchitecture could not have been realized in 0.8u technology due to insufficient transistors. A 486 microarchitecture could be instantiated on, say, 90nm technology but would have been too slow to be competitive. And the design trade-offs baked into the 486 pipeline, appropriate for the silicon of the 1980s, would not leverage the much larger transistor budgets of the 1990s and later.
These microarchitectures were not independent of one another, as Aizcorbe et al. implicitly assumed. The Pentium III was the same microarchitecture as the Pentium II but with SSE instructions added. Moreover, both came from the original Pentium Pro P6 microarchitecture. The Pentium-M was also a descendant of the P6. The Pentium 4 microarchitecture was substantially different, but, as chief architect of both P6 and Pentium 4, I can testify that the Pentium 4 was not unrelated to P6.
Aizcorbe et al. also said, "The Pentium 4 design provided very little value to Intel" due to "overheating." Incorrect. The Pentium 4 was designed to the limits of air-cooling but not beyond those limits. It did not overheat nor suffer heat-related reliability issues. Had the Pentium 4 been designed with a much different microarchitecture, Aizcorbe et al. suggested Intel might have earned much higher profits. What other microarchitecture? They shed no light on this question. Neither Aizcorbe et al. nor Intel can possibly know what might have happened had the Pentium 4 been designed another way.
They also missed another important real-world issue: fab capacity and opportunity cost. Chip manufacturers can always lower the price to make demand go up, but the bottom line might suffer. Our goal in the Pentium product line was to design chips that would command high prices but also generate high demand. And we did it in the face of a market that evolved dramatically from the 486 days, when only engineers and scientists could afford PCs, through the early 1990s, when home computers were becoming common, through the late 1990s, when the Internet took off.
Aizcorbe et al. did mention such influences but did not take them into account in drawing their conclusions. Each of these market exigencies influenced the design, as well as the final results, in special ways. Microarchitectures are intrinsically bound to their markets, and their implementation technologies and are not fungible per their assumptions.
Bottom line, Aizcorbe et al. observed that Intel did not earn the same profits on all its products over the years. Some, it turns out, paid off more than others. We can agree on that much. But saying so is not news.
Robert (Bob) Colwell
Portland, OR
---------------------------------------------
AUTHOR'S RESPONSE
We welcome Colwell's comments but disagree on one major point: Our methodology did not mix-and-match different microarchitectures and silicon implementation technologies, as he suggests. We used only those combinations actually produced by Intel at a given point in time. However, our approach does hinge on the assumption that even when a new microarchitecture is available, Intel cannot retrofit all its current capacity to use it. As a result, old and new microarchitectures are used concurrently in production, allowing us to infer the incremental value of the new microarchitecture.
Ana Aizcorbe
Washington, D.C.
Samuel Kortum
New Haven, CT
Unni Pillai
Albany, NY
Displaying 1 comment