# Transamerica Computer Co. v. International Business Machines Corp.

> District Court, N.D. California · October 18, 1979 · 481 F. Supp. 965

URL: https://www.frixlaw.com/law-library/cases/8868473

## Case

- **Full name:** In re IBM PERIPHERAL EDP DEVICES ANTITRUST LITIGATION. TRANSAMERICA COMPUTER COMPANY, INC., a corporation v. INTERNATIONAL BUSINESS MACHINES CORPORATION, a corporation
- **Court:** District Court, N.D. California
- **Decided:** October 18, 1979
- **Citations:** 481 F. Supp. 965; 1979 U.S. Dist. LEXIS 9071
- **Precedential status:** Published
- **Opinion:** Opinion of the court by Schnacke
- **Judges:** Schnacke
- **Cited by:** 8 later opinions in the Frix Law Library

## Citator (automated)

- No negative treatment found by the automated citator. That is not the same as a confirmation that the case is good law; read the citing cases.
- Full citator and citing cases: https://www.frixlaw.com/law-library/cases/8868473

## How later opinions describe it (automated extraction)

- rejecting a “but for” computer-simulated lost profits model as “fanciful projections” that were “predicted only by assuming hypothetical transactions which never were likely to occur and were never even proposed”
- rejecting claim for lost profits based on “hypothetical transactions which ... were never even proposed”

## Opinion text

OPINION
SCHNACKE, District Judge.
I.
Plaintiff, Transamerica Computer Company, Inc. (“Transamerica”), has alleged that certain activities of Defendant, International Business Machines Corporation (“IBM”), violated Section 2 of the Sherman Act which forbids the monopolization or attempted monopolization of any part of trade or commerce. 1
An appreciation of the nature of the computer industry and the parties’ respective roles is helpful to an understanding of, and indispensable to an evaluation of, that conduct.
Computers, like the punched card accounting equipment that preceded them, depend upon the capabilities of the electronic circuits. Because of the laws of physics involved, an electronic circuit, if properly designed, is capable of performing both arithmetic and logical functions. For instance, it is possible to design a circuit that will add two numbers together (an arithmetic function), and another circuit that will compare the result of the addition to a third number and choose between various alternatives on the basis of that comparison (a logical function). The sequence *972 of arithmetic and logical functions that are to be performed is known as a program. With punched card accounting equipment, programs were “hard-wired”, that is, the sequence of functions to be performed by the arithmetic and logical circuits was predetermined, and could only be altered by actually switching wires around within the machine.
In 1951, in response to the needs of the Bureau of the Census, Sperry Rand Corporation introduced the first stored-program computer, the Univac 1. This first computer differed from its predecessors in that it was possible easily to alter the program that determined the sequence of functions to be performed. No longer was it necessary to “hard-wire” the program; the sequence of functions desired could now be indicated to the computer through “softer” and more flexible means. A program could be punched into a series of cards and the content of those cards could then be read into the computer. The computer would turn to this data read in from the cards for instructions as to what arithmetic and logical functions it was to perform, and in what sequence. A program could be altered simply by altering a punched card, and new programs could be carried out merely by causing the machine to read in a different set of punched cards.
IBM offered its first electronic computer in 1953. That machine, like those offered by competitors, utilized vacuum tubes to perform the electronic circuit functions. Before long, this first generation of computers was outmoded by a second-generation whose transistorized circuits performed more economically and more reliably. In 1964, IBM announcéd a series of machines, the System/360; these were the first of the third generation computers. The 360s not only employed improved components (integrated circuits replace transistors), they also relied upon a single general design or architecture for a broad spectrum of machines. That meant that one computer was capable of efficiently performing both scientific and commercial tasks, and, perhaps most important, it meant that customers who outgrew their smaller machines could “migrate” to larger machines without the need to change their existing programs; the whole 360 family of computers was program-compatible to an unprecedented degree. As a result, System/360 was a tremendous commercial success.
In 1970 IBM announced its 370 system, a further significant improvement, and superior to the 360 system.
The new 370 system involved improved central processing units as well as new and improved peripherals. All of the acts Transamerica claims caused it damage were related to the introduction of the 370 system.
IBM is a supplier of computer systems, supplying all, or nearly all of the user’s computing needs. It offers a wide range of services and products, both software and hardware. 2
The hardware of a computing system consists of a central processing unit (“CPU”), which houses the arithmetic and logical electronic circuits, and a variety of peripheral gear.
The functions of peripherals include: storing data for later access by the CPU; feeding data into the CPU (input); and accepting data from the CPU (output). A machine capable of reading data on punched cards and transferring that data to the CPU is considered an input peripheral, while a printer attached to the CPU functions as an output peripheral. Some peripherals perform all three functions. Disk drives and tape drives are examples of peripherals capable of storing data, inputting data, and outputting data. Data is stored on disks and is “randomly” accessed very rapidly by an access arm with the capability of reading or writing data. The access arm can be made to move to a particular track on the disk where the data is to be read or written. Tape drives are used for reading and writing data sequentially. Random ac *973 cess of data on tape reels is impractical because it is so time consuming.
The tape and disk drives which attached to System/360 CPUs were a lucrative part of IBM’s business, so lucrative in fact, that they attracted competition. In the late 1960’s, several companies began marketing copies of IBM’s tapes, disks and printers, which were “plug-compatible” with IBM CPUs. A user could simply unplug the IBM peripheral, substitute the cheaper copy, and plug it into the IBM CPU. The companies providing this new competition became known in the industry as plug-compatible manufacturers (“PCMs”).
The PCMs enjoyed a tremendous success. They were offering equivalent or better performance at a substantial discount. But in order to sustain their growth the PCMs needed funds. Most computer systems were leased rather than sold. This meant that the PCMs were unable to realize a quick return on their capital investment; much of their capital was tied up in ownership of leased machines, an investment that would not be recouped for years. The PCMs needed money to finance the manufacture of machines to meet a growing demand as well as to pay for the engineering costs of developing new products. They sought financing from a variety of sources: the sale of equity and debt securities; bank loans; the use of leasing companies; and more complex arrangements.
Transamerica Computer Company, Inc., was incorporated in late 1967 as a wholly-owned subsidiary of Transamerica Corporation, a large non-bank financial conglomerate with wholly-owned subsidiaries in the insurance, auto rental, motion picture and other businesses. The parent was well financed, with nearly unlimited credit, and desired a “window to the computer industry.” Transamerica made a number of ventures into computer related financing. The transactions central to this case are Transamerica’s purchase of millions of dollars worth of on-lease tapes and disks from two PCMs, Marshall Industries (“Marshall”) and Telex Corporation (“Telex”).
These manufacturers leased their equipment to end users. After it was on lease, the equipment was “sold” to Transamerica under an arrangement by which the manufacturers were obligated to- collect the rents, to service and maintain the equipment, and to remarket it as leases expired. Unlike some leasing companies which buy equipment, find users, collect rentals, etc., Transamerica had no function except to supply capital. In many cases the end users were unaware of Transamerica’s ownership of their equipment, and, indeed, at various times there was considerable doubt as to whether specific equipment, was owned by Transamerica or the manufacturer. The arrangements also contemplated that, after Transamerica had been reimbursed some agreed amount, further rental revenues would be shared between it and the manufacturer. There were two important reasons for this structuring of what was basically a financing arrangement. First, Transamerica expected that, as “owner” of the equipment, it would be permitted to take, for the substantial benefit of its parent, the investment tax credit then available under the tax laws. And, second, the manufacturers hoped to treat the transfers of title to Transamerica as sales, thus increasing their current profit picture, and aiding them in sales of their corporate stock. Neither of these benefits would have resulted if Transamerica had simply lent money to the manufacturer.
Telex and Marshall were but two of the companies which were successful in displacing IBM peripheral equipment by offering it at prices well below those IBM was charging. IBM responded to the PCM competition by offering certain of its own products at substantially reduced prices, and by offering its peripherals for lease on longer and better terms than it had previously. In addition, new CPUs were introduced that were incompatible with the PCMs’ existing peripherals. These actions, and others described- herein, are those that Transamerica contends were the means by which IBM monopolized and attempted to monopolize in violation of Section 2.
*974 Following a seven month trial, a jury was unable to reach a unanimous verdict on any of the issues presented. The parties, before commencement of the trial, foresaw that possibility, and, stipulated that the case would be submitted to the Court for decision in the event of jury disagreement. (It might be well for the courts or Congress to consider whether such an arrangement might be compelled by the trial judge in advance of trial. No one case should be allowed to monopolize the court’s time indefinitely to the exclusion of the rights of other litigants. If, after one long jury trial, it becomes apparent that further jury trials would be inordinately time consuming, without any realistic expectation that the issues would be resolved, the matter should be resolved otherwise. The best method would seem to be by decision of the trial judge on the evidence presented at the jury trial.)
This Court, having heard and fully reviewed all of the evidence, has made the findings of fact and come to the conclusions of law that are incorporated in this opinion.
II.
In order to establish that IBM monopolized in violation of Section 2 of the Sherman Act, 3 plaintiff must prove:
1) that the defendant was in possession of monopoly power in a relevant market;
and either
2) that the defendant has willfully acquired or maintained that power; 4
or
3) that the defendant used its monopoly power, whether lawfully or unlawfully acquired, to foreclose competition, to gain a competitive advantage, or to destroy a competitor. 5
The first essential inquiry must be whether IBM possessed monopoly power; whether it had the power to control prices in, or to exclude competition from some relevant market 6 during the years here in question, 1969 to 1973.
III.
In general, two types of evidence are relevant to an appraisal of defendant’s ability to control price or exclude competition. The first category can be broadly labelled “market structure,” and encompasses evidence that is relevant to an estimation of the nature and significance of constraints on defendant’s power. The second category, “market performance” relates to a comparison between the actual past functioning of the market and the manner in which monopolized or competitive markets are predicted to function.
A. Market Structure
Economic theory teaches that a monopolist, because it is in a position to control price by varying the quantity of goods it produces and sells, will be able to sell its goods at a price higher than would prevail were the market competitive. 7 As a result it will earn supra-normal profits. In competitive markets, two forces are at work which inhibit the ability of a firm to charge more than the competitive price lev *975 el and earn more than a normal profit. First, if a producer raises its price above the price for competing goods, then customer demand for the higher-priced article will fall off in favor of cheaper alternatives, and the producer will be forced to lower its price again. Second, even if the producer were able to charge relatively more for its product for a time, the resultant supra-normal profits would lure new competitors into the market, and the additional supply of competing goods that they offer to customers would cause prices and profits to return to competitive levels. 8 If that first force (“demand substitutability”) and that second force (“supply substitutability”) are sufficient checks upon defendant’s power to control price, then the market is effectively competitive. If not, defendant is a monopolist.
An inquiry into demand substitutability calls for a definition of the relevant market and measurement of defendant’s share of that market. 9 The relevant market is comprised of defendant’s product, products that are fungible, and of those products which the customer would consider to be reasonable alternatives. 10 If a small increase in the price of product A will cause customers to shift their allegiance to product B, then the cross-elasticity of demand is high, the products are reasonable alternatives, and they belong in the same relevant market.
The geographic aspects of the relevant market must also be considered. 11 Thus, if small changes in the price of a product in one location will cause customers to turn to alternatives available in another location, then the quantity of reasonably alternative products supplied at both locations should be considered as part of the relevant market. The relevant geographic market is the area of effective competition within which the seller operates and to which the purchaser can practically turn for supplies. 12
With the relevant market defined, the defendant’s share of that market can be measured. A relatively small market share would lead to the inference that monopoly power was not present since the ready availability of substitutes would defeat attempts to use that power. The larger defendant’s market share, the stronger is the inference that competitors would be unable to effectively check exercises of monopoly power. 13
The distribution of the share of the market not supplied by defendant is also telling. Defendant’s share is more likely to indicate monopoly power if the rest of the market is widely distributed among many small competing suppliers than it would be if the size of competitors and the market share held by them approached defendant’s size and share. 14
*976 Under some circumstances it may be proper to apply the concept of supply substitutability to the definition and measurement of the relevant market and to include within the market any suppliers which might readily and easily enter by producing a reasonably acceptable alternative product.
The greater the barriers faced by a new entrant, the more probable it is that control of a particular market share would enable defendant to exercise monopoly power. 15
Anything that tends to inhibit firms from readily and easily entering the marketplace can be analyzed as an entry barrier. Thus, the capital an entrant would have to invest is a factor, as are the employee skill levels required for a firm to be successful. Product differentiation, product loyalty, or any customer disinclination to accept the product from different suppliers are hurdles the new entrant may have to overcome. And, the existence of economies of scale or a dwindling market demand would make entry difficult.
Within a market, as defined by demand cross-elasticity considerations, economically significant submarkets may exist. 16 Submarkets are zones of .actual or potential competition 17 that are sufficiently distinct from the larger market that one firm could exercise the power to control price or the power to exclude competition within them. 18
Practical aids in identification of such zones of competition include industry or public recognition of the submarket as a separate economic entity the product’s peculiar characteristics and uses, unique production facilities, distinct customers, distinct prices, sensitivity to price changes, and specialized vendors. 19 If these indicia indicate a submarket exists, market structure and market performance should be examined to determine whether defendant is in possession of monopoly power. 20
B. Market Performance
Evidence of the actual operation of the market or submarket can also provide valuable clues as to the existence of monopoly power. If defendant has, in the past, successfully controlled price or excluded competition, that is direct and convincing evidence that it had the power to do so. 21
Less direct evidence can also be persuasive. Many courts have examined defendant’s profit record on the theory that persistent excess profits are inconsistent with the competitive model and attributable to the possession of monopoly power. 22 Mo *977 nopoly is also expected to result in technological stagnation. Without a competitive spur, the monopolist is thought not to have any incentive to innovate. Thus, the rate of technological progress in the industry is relevant. 23
In a competitive industry firms are expected to be “price takers.” That is, they will be unable to affect price by variations in the quantity of goods they offer for sale. Where it appears defendant was a “price maker,” able to choose among a range of. price options in order to achieve its profit goals, or able to disregard cost and set price according to utility to the consumer (functional pricing), it may indicate monopoly power. 24 , 25
Market history, in terms of entry and growth of new firms or exit of failed firms, and whether concentration is increasing or decreasing are additional factors that can indicate whether the competitive process is functioning properly or not. 26
Transamerica asserts that IBM had monopoly power in three markets:
1. The manufacture and placement of general purpose electronic digital computer systems (“systems market”);
2. The manufacture and placement of tape drives and their controllers plug-compatible to IBM CPUs (“tape market”); and
3. The manufacture and placement of disk drives and their controllers plug-compatible to IBM CPUs (“disk market”).
C. The General Purpose Systems Market
Computer systems are the functioning combination of hardware (CPUs and peripherals) and software that a user may employ to satisfy data processing requirements. Transamerica draws a distinction between general purpose systems and special purpose systems. Special purpose systems are those that are designed for .and dedicated to specialized applications. For instance, a special purpose computer system might be utilized to control a manufacturing or a chemical process. General purpose systems, on the other hand, are those that are capable of easily and economically being adapted to a variety of data processing applications. A general purpose computer system might be used to process a company’s payroll, to update its books of account, or to allow its engineers to solve complex mathematical problems while sitting at a remote terminal. Some (but not all) general purpose systems are capable of performing all these tasks concurrently. By definition, all general purpose systems allow the user to switch easily from one task to another.
This Court accepts the view that the market for general purpose computer systems should be analyzed separately from the market for special purpose computer systems. The systems sold to satisfy these different needs are not readily interchangeable and different suppliers specialize in each of the markets.
Transamerica’s definition of the systems market is, however, too narrow. Only suppliers of complete systems are included, suppliers of parts of systems are ignored. Firms which offer the user parts of systems have become a major competitive force in the computer industry, and, because they significantly constrain IBM’s power to control the price of the systems it sells, a *978 market definition that ignores them- is incorrect.
At one time the only option a general purpose systems user had was to choose between suppliers of complete systems. If an IBM user was dissatisfied with IBM’s peripheral equipment or with IBM’s software, the only recourse was to remove the entire IBM system as a unit, and replace it with an entire system from another manufacturer. Since that time, literally thousands of companies have entered the systems market, not as suppliers of entire systems, but rather as specialists providing users with replacements for parts of their computer systems. IBM did not charge the user one price for an entire computer system, rather they priced each component of their systems individually. That meant that companies which focused on replacing parts of a system could offer the user a better price on that part and the user was still free to acquire the rest of the system from IBM.
The PCMs are perhaps the best example of such specialists. Some PCMs got their start by copying IBM’s tape drives and offering them to IBM’s systems customers. Other PCMs specialized in replacing IBM’s disk subsystems with their own versions. Still others recognized a competitive opportunity in supplying printers for use on IBM’s systems, and some PCMs competed with IBM for sales of the electronic memories attached to IBM CPUs.
Entry barriers proved remarkably low, and the PCMs proliferated. Today, a user could replace practically every part of an IBM system with hardware from a PCM. The same is true for software. Software firms offer programs to users of IBM systems that are reasonable alternatives to the programs available from IBM. And the entry barriers faced by an entrepreneur with a software package to sell are truly insignificant.
A market definition should “recognize competition where, in fact, competition exists” 27 , and should include all significant competition even though that competition differs in form or nature. 28 Transamerica’s systems market definition fails this fundamental test. By assuming that only manufacturers offering complete systems compete with IBM, Transamerica ignores the fact that IBM systems users had other reasonable alternatives; they could, replace parts of their IBM systems. In many ways, partial replacement was a more significant constraining force than total replacement. Customers found partial replacement attractive because it did not involve a scrapping of the user’s investment in programming and personnel skills, and it allowed users to reduce data processing costs without abandoning entirely the security of the IBM fold. Partial replacement of parts of IBM systems was a significant constraining force on IBM’s market power during the 1969-1973 time period. 29 Billions of dollars worth of equipment was being replaced piecemeal by IBM systems users. A market definition that ignores this cannot be accepted.
Some minicomputers (minis), small, relatively inexpensive computer systems should also have been included in Transamerica’s general purpose systems market definition. Transamerica included the smaller IBM systems, but excluded the thousands of minis that were directly competitive with those small IBM systems. That exclusion would have had more validity in the past. When they were first introduced minis were limited, or at least thought to be limited to special purpose applications. But things changed. By May, 1970, Computerworld, a widely read industry newspaper, recognized that the “. . . new minicomputers and microcomputers showed drastic changes in design that make them truly general pur *979 pose . . . . For the first time, they offer a reasonable alternative to ‘full-sized’ systems.” Advertisements by minicomputer manufacturers emphasized that theme, and apparently users got the message. In August of 1972, Computerworld reported:
“They [minis] are being used as true general purpose computers, running a range of applications such as order entry, sales analysis, payables, receivables, general ledger, inventory control, payroll, price ticket printing and credit authorization . .
During the relevant time period minis were being sold as reasonable alternatives to the smaller general purpose systems offered by IBM. When connected together, several minis were a reasonable alternative to the acquisition of larger systems. And minis, whether used as data collection devices, intelligent terminals, pre-processors, or stand-alone systems, could ease the workload of larger general purpose systems, thereby providing the user with a viable alternative to the acquisition of a more powerful system. The competition minis provided was significant, and their exclusion from the systems market definition was incorrect.
IBM also faced other constraints not contemplated by Transamerica’s systems market definition. Service bureaus buy computing systems and make them available to others by renting out time on them or doing work on them for the benefit of end-users. Time-sharing companies make their computers available to many users who access them concurrently via remote terminals. In both cases, users may turn to such companies during peak periods, or for the specialized services they offer. Leasing companies are financial intermediaries; they buy computer systems, or parts of computer systems, from manufacturers and lease them to users. Service bureaus, time-sharing companies and leasing companies all provide significant competition for IBM. However, their exclusion from the market definition is analytically correct. These companies all buy computers from systems manufacturers, add some financial or technical services, then turn around and offer the computers to the end user. Their basic product is computing power. Their basic raw material is a computer system. Since they all must buy equipment from manufacturers, if one manufacturer has the power to control price at the time of the initial sale, it has the power to indirectly control the price at which the computer re-enters the market with the added services. Where it is possible that a defendant might indirectly control the price at which goods reenter the market, that alternative source of goods should be excluded from the market definition. 30 But the nature and the volume of the business done by service bureaus, time-sharing companies, and leasing companies need not be totally ignored. Although exclusion from market definition is called for, the constraints these companies provide decrease the likelihood that defendant’s market share reflects monopoly power.
This Court agrees with Transamerica’s contention that the concept of supply substitutability does not affect the systems market definition since there has been no evidence that any firms are in a position to readily shift their production facilities into the manufacture of general purpose systems, and also agrees that the United States constitutes the relevant geographic market.
Transamerica has made no effort to measure IBM’s share of the market as this Court would define it, i. e., including minicomputer manufacturers, software suppliers, and PCMs as competitors. Nonetheless, the evidence of IBM’s market share, in the market as defined by Transamerica, will be examined for whatever light it might shed on the issue of IBM’s market power.
The most common measurement of market shares is a comparison of the competitors’ annual sales. 31 Here, Transamerica, *980 however, has used what is called the “installed base” method.
Included in each company’s installed base for any year are all computer systems ever leased or sold by that company which are still in use in that year. Those systems, no matter how old they are, are valued at their original purchase price or their current purchase price, whichever is higher. Transamerica insists that this method is appropriate because the general purpose computer systems market is predominantly a lease market. Any measurement that failed to recognize past activity, they argue, would ignore the substantial revenues IBM enjoys from machines leased and shipped in the past and ignore the substantial hold IBM has on those customers due to software lock-in. 32
The computer industry is characterized by cyclical development or “generations” of computer equipment. These cycles can be expected to produce sharp fluctuations in one company’s share of annual shipments, and the tendency of the installed base method to smooth out these swings is another advantage claimed for it.
Installed base is a method traditionally employed by the computer industry, including at one time IBM itself, and this too is a reason advanced for its adoption here.
IBM counters that installed base reflects more history than current market power. IBM is right. IBM has been successful in the computer field almost since the industry’s inception. Thus, inclusion of all IBM machines still in use distorts their share. The successes of newer companies (such as the PCMs) are camouflaged by the large installed bases of more established manufacturers.
The installed base method credits IBM with purchased machines from which it no longer enjoys any revenue, and even with machines that have come back into the market to compete against it. Machines sold in an earlier year to leasing companies, time-sharing companies, service bureaus, and used machines are all attributed to IBM’s market share in later years. Transamerica insists this is proper because IBM, as the original manufacturer, indirectly controls the prices at which such machines and services can be offered. While such reasoning supports an exclusion of this competition from market definition, it does not warrant an attribution of the value of their capital equipment to the manufacturer’s market share except in the year of the initial sale. 33
The lease nature of the business is an insufficient reason for abandoning an annualized approach. By 1973, less than half of IBM’s general purpose systems were leased. And if software lock-in, or any other factor, persuades users not to switch vendors, then the future successes that installed base would predict will ultimately be reflected in annual shipment data. One of Transamerica’s industry market share experts, recognizing the infirmities of installed base, recently discontinued his use of that method.
It is not necessary to discard all of Transamerica’s data because the installed base measurement method is unacceptable. Annual shipments can be approximated from the installed base figures in evidence by subtracting one year’s installed base from the following year’s. This will yield a figure for net shipments. Net shipments appear to be a more desirable indicator then gross shipments because many computers are shipped to users who return their older machines to the manufacturer. Cyclical *981 fluctuations can be accounted for by averaging the share figures for several years.
In some of its data Transamerica has included the value of equipment placed by the PCMs as part of IBM’s share. This is clearly incorrect, and Transamerica has offered no convincing rationale for it. It would be possible to rehabilitate this data by extracting from IBM’s share the value of disks and tapes supplied by PCMs. However, the value of PCM supplied printers, electronic memories, and other gear cannot be separated out, providing still another reason for viewing the resulting figures cautiously.
IBM’s share of net shipments (less all PCM products) has been calculated to be: 65.4 percent; 67.9 percent; 51.4 percent; 56.4 percent; 56.5 percent; 30.6 percent; and 64.4 percent, for the years 1969 through 1975 respectively. Over the entire seven-year period IBM shipped 57.4 percent of all general purpose systems (net). The next largest share was shipped by the PCMs as a group with 14 percent over the seven years. Burroughs was the second largest manufacturer of systems with 7.7 percent, followed by Univac with 4.9 percent, and Honeywell with 4.6 percent. 34
Even in a properly defined market these market share statistics would not be overwhelming. Other ' courts have doubted whether sixty percent of a market would be enough. 35 But while a finding of monopoly power is not foreclosed by IBM’s share if the setting is conducive to the exercise of such power and if other indicators confirm its presence, 36 such a finding is certainly not compelled by these market shares.
The distribution of the remainder of the market serves to enhance IBM’s power. No single manufacturer approaches the success IBM enjoys. However, their shares are growing while IBM’s is shrinking. IBM’s share tends generally downward, indicating that monopoly power is not present, or, if present, that the monopolist’s grip is weakening.
Consistent supra-normal profits may be attributable to an ability to control price. Here IBM’s profits have been very substantial. Its after-tax return on equity during the 1964 to 1973 period was consistently 30 to 70 percent above the Fortune 500 median, while its after-tax return on sales doubled and sometimes tripled the industrial giants’ median. But the inference that a defendant that enjoys healthy profits only does so because of an unhealthy market structure is not a strong one. Good management, superior efficiency and differences in accounting provide explanations that are just as plausible, and none of those explanations is inconsistent with an effectively competitive market.
One thing excessive profits can be expected to result in is increased entry by firms hungry to participate in the rich rewards. Transamerica says that this has not occurred because the entry barriers faced by potential entrants are insurmountable. It is true that enormous barriers are faced by firms wishing to enter this market as suppliers of complete computer systems. Between $500 million and $1 billion in capital would be required; a national network of sales and maintenance offices staffed with highly skilled personnel is imperative; the entrant would have to fund research and development projects; a broad product line must be offered; and customer loyalty due to software lock-in would have to be overcome. Because such a large commitment would be involved for entry as a systems supplier, it is not surprising to find that there has been no new entry on that basis since the early 1960’s.
The flaw in Transamerica’s entry barrier analysis is a consequence of its market definition error. Competition and entry in this market does not occur on only a full-system *982 basis. Entry could, and did occur where companies recognized profitable opportunities in less inclusive offerings. Storage Technology Corporation, one of the most successful PCMs, was started by a bright engineer with $225,000 in equity capital, and experienced phenomenal growth. It was not alone. In 1970, International Data Corporation, (IDC) a company specializing in statistical studies of the computer industry, reported:
“Just 15 years ago, the number of companies in the infant industry could be counted on the fingers of one hand; today well over 4,000 companies produce computers, related equipment and supplies, or offer data processing and programming services.”
IDC thought there had been “an explosive proliferation of options for the computer user.” In 1971, Transamerica, in a presentation to its parent Board of Directors, agreed, saying:
“The frequency of entry into the data processing industry is quite high compared with any other industry, and there are usually at any one time, more competitors than anyone would really like to have.”
The new entrants were growing. From 1967 to 1975 the value of PCM equipment attached to IBM systems increased from $15.8 million to $2.6 billion. In the same period, leasing companies jumped from holding 8.3 percent of the IBM CPUs to 26.7 percent. And the market as a whole was growing; most of IBM’s full-systems competitors participated in that growth. Digital Equipment Corporation, (DEC) for example, was founded in 1957 with DEC initial capitalization of $70,000. By 1977 it had computer related revenues of $1.1 billion. One commentator has referred to IBM’s systems competitors as the “billion dollar dwarfs.”
This is not the story of a stagnant, dominated industry. There is no doubt that the pace of technological progress in the computer industry is extraordinary. Commentators are fond of saying that had the auto industry kept the same pace over the last 30 years, a Rolls-Royce would cost $2.50 today and would have an EPA gas rating of 2,000,000 miles per gallon.
There is some evidence that IBM sought to maximize profits by choosing a price for its products from a range of alternative prices. But that is not necessarily inconsistent with an effectively competitive market. The range of price options available to IBM was never so broad that it could set a price without regard to competitors' prices. And, in most cases, the range was only theoretical, because competitive pressures, not IBM preference, dictated which price had to be adopted.
This market was not monopolized. IBM did not control prices, it reacted to the prices set by others. IBM’s market share (if the market had been properly defined) was below 57 percent and falling. Entry was easy for anyone with a good idea and the courage to exploit it, and new entrants, along with the old, grew and prospered. This Court finds that Transamerica has failed to prove that IBM had monopoly power in the general purpose systems market.
Nonetheless, Transamerica’s other contentions will be considered on the assumption that IBM did have such monopoly power.
D. Tape and Disk Markets
The hardware of a computing system was previously described as consisting of a CPU and peripherals. Further refinement of that over-simplification is necessary for an understanding of the peripherals markets alleged to be relevant.
Certain control functions must be performed if peripheral gear is to function properly. The access arm of a disk drive, for example, must be positioned over the proper track before data can be read from, or written onto that track. These control functions can be performed by the arithmetic and logical circuits of the CPU, in which case the control function is said to be “integrated” into the CPU. But when the CPU is tied up with control functions, it is not free to perform more sophisticated tasks, *983 and so an alternative design was incorporated into the architecture of System/360. Most of the peripherals attachable to System/360 CPUs were not integrated. The control functions were performed by two independent “boxes” or machines with arithmetic and logical circuits: channels and control units.
A limited number of channels could be attached to most System/360 CPUs. Channels funnel data into the CPUs’ main electronic memory, and take from the memory data that is to be written onto any of the peripheral devices that may attach to the channel, and send it to its destination. Channels perform control functions common to a variety of peripheral gear.
It is possible to attach a number of control units to each channel on the system, and the peripheral devices attach to the control unit. Control units perform the control functions that are required by the specific type of peripheral that attaches to it. Thus, one type of tape drive, expecting specific control functions to be performed, only attaches to a specific type of control unit. Other types of tape drives require a different control unit, and the same is true of disks. But various types of control units can be attached to the same channel.
PCMs typically supplied a combined package of disks and disk control units attachable to, or “plug-compatible” with, IBM channels. This meant that the user could remove the IBM control unit and disk combination (known as a “subsystem”), plug in the PCMs’ replacements and the computing system would continue to function with very little inconvenience. That was so because the PCMs had duplicated the “interface” between the IBM channel and the IBM disk control units. An interface has both physical aspects (in terms of the number of wires' involved and their arrangement in a connector), and protocol aspects (what a given pulse sequence on a particular wire is understood to indicate by the machine receiving the pulse).
Unlike PCM disk suppliers who provide subsystems, PCM tape suppliers concentrated initially on tape drives and generally did not sell tape control units. Thus, the interface that they duplicated was the interface between IBM tapes and IBM control units.
Although many computer manufacturers designed their systems with this CPU-channel-control unit-peripheral hierarchy of control functions, no other systems were designed so that the interface between the channel and control unit (disks) or the interface between the control unit and peripheral device (tapes) was precisely the same as the corresponding IBM interface. Thus, tape drives and disk drive-control unit combinations developed for use on IBM’s System/360 could not, without some interface modifications, be plugged into another computer manufacturer’s system, and vice versa.
The two peripherals markets alleged by Transamerica consist of tape drives and tape control units, and disk drives and disk control units that can be plugged into IBM CPUs without any modifications, i. e., that are plug-compatible.
These peripherals markets will be examined together. Both markets, if they can be so described, became economically significant in the late 1960’s, when PCMs began duplicating the peripherals that would attach to IBM’s computers. Prior to that time some of the PCMs supplied peripherals to IBM’s general purpose systems competitors, who then incorporated them into the systems they offered. When the systems competitors took to manufacturing their own peripherals, the PCMs turned their attention to marketing peripheral replacements to IBM users.
Having already accepted a definition of a systems market that encompasses all the economic activity on the peripherals front, perhaps “submarket” would be the proper term to apply here. In any event, the Brown Shoe submarket indicia can provide valuable clues as to whether plug-compatible tapes and control units and plug-compatible disks and control units are sufficiently distinct that they could be monopolized.
*984 Industry recognition of the peripherals markets as separate economic entities could hardly be more evident. IBM studied the “tape market” and “disk market” separately and referred to them repeatedly by those or similar appellations. Others in the industry, including witnesses called by both sides, referred to and treated the tape and disk markets as separate economic entities.
The peculiar uses and characteristics of the products involved have already been spoken of. Because the interfaces between IBM CPUs and control units and between IBM control units and peripherals were unique, the tapes or disks described as plug-compatible to IBM CPUs could only be used with those CPUs. Users with systems provided by other manufacturers could not use them at all, and IBM users could use nothing else. Thus, customers for plug-compatible tapes and disks were readily identifiable and distinct.
The production facilities for tapes were different from the production facilities for disks, and both were different from the production facilities for other components of the system. IBM’s main disk facilities were in California; its main tape facilities in Colorado; and its CPUs were manufactured elsewhere. However, the production facilities of independents and systems manufacturers producing non-IBM-plug-compatible tapes and disks were probably not so different from the facilities for the corresponding plug-compatible peripherals. This is so because the tapes and disks were similar in almost all respects save for the unique interface electronics.
Prices were distinct. In setting its prices IBM was less influenced by the prices of non-plug-compatible tapes and disks than by the PCMs’ prices. Users were price sensitive. Lower prices for equivalent or superior performance was the PCMs’ selling point. With it they managed to attract millions of dollars worth of business away from IBM.
Finally, the PCMs were specialized. Peripherals were their major activity. Many supplied only plug-compatible tapes or only plug-compatible disks. Those that supplied both, made the combination only after experiencing an initial success with the one device or the other. The Brown Shoe sub-market criteria indicate that the peripherals markets are capable of being dominated, and the markets should be examined further.
IBM attacks these markets as narrow for three reasons: (1) they ignore the interchangeability between tapes and disks; (2) they fail to include as potential suppliers either the systems manufacturers who make their own peripherals, or the independents who manufacture peripherals for them; and (3) they fail to account for the constraint of systems competition on peripheral price discretion.
IBM's first argument, that tapes and disks are interchangeable, is without merit. In configuring a system, users have several alternatives. Various types of media are available to store and access data. Principal among these are tapes, disks and electronic memory. In that order, data is accessed by the CPU with increasing rapidity, but also with increasing cost. Electronic memory is so expensive that it is only used as a temporary storage area for data. Data is read into the electronic memory (from a tape or disk perhaps) where the CPU can readily access and work with it, then written from electronic memory onto another medium for more permanent storage. Electronic memory is not really a peripheral, rather more a part of the CPU itself, and is not a reasonable alternative to either tape or disk.
For some applications a disk must be used. Where the CPU must access data frequently, but not necessarily in a serial fashion, a disk is indispensable; tapes are not an alternative regardless of the price differential. Applications where many users access a central computer through remote terminals (such as an airline reservations system) are only feasible if the data for them is kept on disk (or other direct access device). This is so because every data record on a disk pack can be reached within a fraction of a second by a lightning-quick access arm.
*985 To reach a particular record on a tape, the tape reel must be spun until the proper spot is reached. It might take as long as one minute to reach a particular data record, ages by comparison with disk speeds. However, if the records on a tape reel are sorted so that the CPU can access them seriatim, then the time required to access records in a tápe file is quite acceptable. So, tapes are typically used only for data which can be accessed serially, and only disks are used when data must be accessed randomly. Still, there is some overlap.
For some serial data files either a tape or disk can be used. Because the cost of storing data on disks is much greater than the cost of tape storage, only with relatively short files would the user consider tape and disk to be alternatives. Most system users employ both disks and tapes.
Indications are that the incidence of actual substitution between tape and disk is not high. Neither IBM nor the PCMs considered tape prices when pricing disks or vice versa. IBM’s Financial Procedures Manual calls for calculation of the impact on existing products’ profits caused by introduction of a new product. If tapes and disks were reasonable alternatives, one would expect that a new tape would impact the profitability of old disk products and that new disk devices would impact the profits of existing tape products. Yet only rarely were such impact calculations made. The manual also calls for competing products to be identified. Tapes were not so identified when disks were introduced, nor were disks, when tapes were introduced.
The head of IBM’s Computer Division testified that slight variations in the price of disks would not affect the demand for tapes; that variations of 40, 50 or 100 percent in price would be required. That is not the high degree of price cross-elasticity that is required if two products are to be considered reasonable substitutes and placed in the same market.
IBM’s second argument for a broader market definition, supply substitutability, calls for the inclusion of firms with a high cross-elasticity of production. This contention would require a finding that producers are capable of quickly and easily entering the market to produce alternatives. That finding cannot be made. Producers of non-plug-compatible peripherals would have to change the interfaces in their devices to cope with IBM’s CPUs. Interface changes would cost hundreds of thousands, if not millions of dollars, and might require a year to carry out. Under these circumstances it would be improper to include this potential competition in the peripherals market definition.
IBM’s third attack, emphasizing the influence of systems competition, has merit. The cost of peripherals accounts for 50 to 70 percent of the total cost of a typically configured general purpose computer system. Efforts made at controlling price in the peripherals submarkets would be felt by systems market competitors because control of peripheral prices translates indirectly into control of systems prices. If the systems market were effectively competitive, one would expect that the pressures exerted by that competition would be sufficient to defeat attempts to wield monopoly power in a peripheral submarket. Thus, the existence of monopoly power in a peripherals market must be predicated upon a finding of monopoly power in the systems market.
Transamerica’s expert economist agreed. He testified that by definition IBM could not have monopoly power in the peripherals market if it did not have monopoly power in the systems market. This Court’s conclusion that IBM lacked systems market monopoly power is thus fatal to Transamerica’s peripheral markets claims. Nonetheless, this inquiry will proceed under the assumption that IBM had a systems monopoly.
As with the systems market, Transamerica combined a narrow définition of the peripherals market with a distorting share measurement technique, the installed base. Installed base measurement makes even less sense here than it did in the systems market. For many years, IBM faced absolutely no competition for the tapes and disks that attached to its CPUs. As a re- *986 suit, it built a tremendous installed base of those peripherals. Even if the PCMs had won every sale from 1968 onward, many years would have to pass before the installed base figures would show PCM competition was significant. And the software lock-in argument rings hollow here. By definition, every product in these markets is plug-compatible; none of the competitors enjoys any advantages traceable to software lock-in.
Transamerica’s mistaken failure to recognize systems competition in its market definition invalidates its share measurement data. However, the net shipments of peripherals in the market, as Transamerica has defined it, can be calculated in the same manner as net shipments were calculated for the systems market. The installed base at the end of one year is subtracted from the installed base at the end of the following year to indicate net shipments over that year. Thus, the 1969 net shipment figures were calculated by subtracting the 1968 year end installed base from the 1969 year end installed base. Years prior to 1969 are ignored. Before that time PCM competition was insignificant and IBM’s net shipments were almost 100 percent of all net shipments of tapes and disks that hooked onto its CPUs.
The following two tables show the total net shipments of peripherals (drives and control units) for the years 1969 through 1975 as derived from the installed base data. 37 IBM’s share appears both in dollars and as a percentage of total. That portion not attributable to IBM was divided amongst the PCMs. Dollar figures are shown in terms of thousands of dollars of monthly rental value.
TAPE MARKET
1969 1970 1971 1972 1973 1974 1975 1969-75
Totals 5760 737 519 -19 1443 4144 2569 15155
IBM 4547 105 -587 -470 544 2409 1618 8167
IBM% 78.9% 14.3% * * 37.7% 58.1% 63% 53.8%
DISK MARKET
1969 1970 1971 1972 1973 1974 1975 1969-75
Total 7838 15594 627 5204 10376 21738 16772 78158
IBM 7248 13423 -3095 3809 7155 18409 13240 60237
IBM% 92.5% 86.1% * 73.1% 69% 84.7% 78.9% 77.1%
* —Loss
It would be difficult to conclude that a company possessed monopoly power in a market where its net shipments were negative at the very time it allegedly used or maintained that power. That is the case here. In 1971, IBM was losing ground to its PCM competitors. That was a relatively lean year for everyone, but more so for IBM than for its competitors. The PCMs were increasing ' their customer base at IBM’s expense.
. Even the aggregate numbers are unconvincing. IBM’s mere 53.8 percent of the tape market suggests monopoly power is not present. The 77.1 percent figure in the disk market would support a monopoly finding, but the significance of this market share is diminished by the realization that IBM had 100 percent of the market only a few years before, and the failure of the market definition to account for systems competition means these numbers must be viewed circumspectly. These share figures are also inflated by Transamerica’s inclusion into the market of products which the PCMs never duplicated, and areas of the country where PCMs chose not to compete.
*987 Entry barriers around these markets are extraordinarily low. The market definition excluded potential suppliers (such as systems competitors and their independent peripherals sources) because they were incapable of quickly and easily supplying substitute products. But the hurdles those potential suppliers must clear are far lower than those new systems market entrants face. Firms already producing peripherals have the production facilities, the trained staffs, and the research and development expertise needed; all that remains for them is to study the IBM interface and duplicate it. Two of the most prominent systems competitors, Univac and CDC, have done just that. Both manufacture peripherals for their own computer systems, and also manufacture IBM plug-compatible peripherals that compete in the markets at issue here. And many PCMs market their wares to other systems manufacturers as well as to IBM end-users. While the cost, in time and money, of developing an interface is sufficient to exclude systems manufacturers and independents from the market definition, it does not provide a formidable entry barrier around the market as defined.
The entry barriers faced by completely new entrants into the peripherals market have already been discussed in relation to the systems market. Those barriers are low, as is demonstrated by the history of entry into peripherals manufacture. PCMs, attracted by IBM’s high profits on its peripherals, stormed these markets and flooded them with cheaper copies of the IBM gear. IBM, unable to maintain its relatively high price, was forced to react with lower prices and improved products. This picture is consistent with the dynamics of an effectively competitive market, not a monopolized one.
This Court will not be misled by the share figures propounded by Transamerica. Those calculations ignore the significant constraint .of systems competition, and measure the markets as an aggregation of all peripherals in use, when there is no reason for doing so. Accurate share figures cannot be determined from this record, but the data which can be rescued indicates that IBM did not have market control, and that its share was not dominant. Entry barriers are low, and entry has been easy, frequent, and successful. Even if a systems monopoly were not a prerequisite to peripheral market control, or even if IBM had a systems monopoly, this Court would find, as it does here, that Transamerica has failed to prove that IBM had monopoly power in either the tape or the disk market.
IV.
PRICING CONDUCT
A defendant not in possession of monopoly power may nonetheless incur antitrust liability by attempting to monopolize. Transamerica alleged that IBM attempted to monopolize by its pricing conduct, by its design conduct, and by its long-term leases. IBM’s conduct will be examined first under the assumption that IBM did possess monopoly power. Following that, the conduct that would have violated the monopolization provision of Section 2 will be re-examined to determine whether, in light of the absence of monopoly power, an attempt to monopolize has been proved.
IBM responded to the lower prices and increased successes of the PCMs by offering some of its own products at substantially reduced prices. Transamerica claims that those prices were predatory and violated the antitrust laws.
The concept that prices can be used as a means of acquiring or maintaining monopoly power is not new; it has long been recognized that financially powerful firms might coerce or destroy their rivals by pricing at unremunerative levels. Attempts to distinguish between predatory pricing and legal price behavior, by comparing defendant’s costs with its prices, are new.. In an article published in 1975, Professors Areeda and Turner suggested that pricing conduct should be conclusively presumed legal if price levels exceeded either defendant’s average variable cost or defendant’s short-run *988 marginal cost. 38 Reaction to the Areeda and Turner proposal has been mixed. Courts have embraced it; academicians have criticized and rejected it.
An understanding of the relationship between these cost measures is critical to an appreciation of the effect that imposition of the Areeda and Turner rule would have. All three kinds of cost (average cost, average variable cost, and marginal cost) will at first decline, then later increase as a firm’s output is expanded. Average cost (which includes both fixed and variable costs) will always be greater than average variable cost (which only includes variable costs). At low output levels, marginal cost will be less than both of the other cost measures. As output is increased, marginal cost will rise until at some point it will be greater than average variable cost (but less than average cost). If output is increased still further, marginal cost will increase to the point that it exceeds both average cost and average variable cost. 39
A monopolist has the power to control market price by varying the quantity of goods it offers for sale. As that quantity is varied, the monopolists’ marginal, average variable, and average cost will also vary. If market demand would permit it, 40 the monopolist would have the following options: 1) setting a market price in excess of both average cost and marginal cost; 2) setting a market price below average cost but greater than marginal cost; or 3) setting a market price below both average cost and marginal cost. Areeda and Turner would conclusively presume the monopolist’s conduct legal if it exercised option 1) or option 2), and would conclusively presume conduct illegal only if option 3) was chosen.
IBM argues that the Areeda and Turner rule is the law of the Ninth Circuit and is binding upon this Court. Transamerica maintains that prior cases are distinguishable, and that a monopolists’ pricing is illegal if it sets a market price below the point at which it would maximize profits if the purpose and effects of that action are to unnecessarily exclude competition. Neither standard is appropriate in this case. If a monopolist, in response to actual or threatened entry into a previously controlled market, reduces prices below its full cost, there are strong implications of conduct that unreasonably restricts competition. Therefore, and for reasons hereinafter stated, the IBM conduct will be judged against a “full cost” standard.
A. Predatory Pricing Case Law.
In 1976, the Ninth Circuit Court of Appeals, in Hanson v. Shell Oil Co. 41 (“Hanson ”), held that a plaintiff’s failure to show that a defendant’s prices were below its marginal or average variable costs was a failure as a matter of law to present a prima facie predatory pricing case under Section 2. One year later, in Janich Bros., Inc. v. American Distilling Co 42 (“ Janich ”), the Ninth Circuit, relying heavily on Areeda and Turner, upheld a directed verdict against a plaintiff whose proof suffered from the same insufficiency. If those cases are binding upon this Court, in this case, then, as IBM suggests, Transamerica’s predatory pricing allegations must fall as a matter of law, for there was no evidence of *989 any IBM price below marginal or average variable costs.
Both Hanson and Janich involved non-monopolists, charged with attempting to monopolize. In attempt cases, a plaintiff must establish specific intent to control prices or destroy competition, predatory or anticompetitive conduct, and dangerous probability of success. But it may be misleading to list those three as separate and independent elements. In fact, there is a great deal of interrelation between them. Specific intent may be established by direct evidence, such as memos or statements made by corporate officers, or alternatively, specific intent can be inferred from the defendant’s conduct. 43 Dangerous probability can be shown by proof of substantial power in a relevant market, or it can be inferred from conduct which independently violates Section 1 and is of a kind “clearly threatening to competition or clearly exclusionary”, or it can be inferred from specific intent. 44 In extreme cases only conduct need be proven directly; specific intent can be inferred from the conduct and dangerous probability can be inferred from the specific intent. 45
A firm that prices its product at levels above marginal or average variable cost is not necessarily engaged in clearly exclusionary conduct. A firm that prices its products below those levels is. These standards then, are appropriate for judging the pricing conduct of a firm charged with attempting to monopolize where no independent evidence of specific intent and no independent evidence of dangerous probability (relevant market proof) is admitted. Hanson and Janich were such cases, and the application of very strict standards to plaintiff’s a proof was proper in those circumstances.
However, less egregious conduct may violate Section 2 if engaged in by a monopolist 46 or by one attempting to monopolize if there is also direct proof of specific intent or direct proof of a dangerous probability of success. 47 Hanson and Janich are deemed therefore to leave open the appropriate standards to apply to. cases with different facts.
The same is true of CalComp. 48 IBM thinks CalComp foreclosed any further inquiry by adopting the Areeda and Turner marginal cost standard. CalComp dealt with the pricing conduct of a monopolist and decided that if prices were reasonable they did not violate the law, and that price reductions which responded to lower prices of competitors, but still left defendant’s products “substantially profitable”, were reasonable. 49 However, CalComp carefully limited its pronouncements to the facts of the case before it. The court recognized that on a different showing, even profitable prices might violate Section 2 if set by the monopolist to discourage new entrants (limit pricing), and also held out the possibility that other aspects of a defendant’s conduct might make prices in excess of marginal cost predatory. 50 Those “refinements” are at odds with the fundamentals of the Areeda and Turner analysis, where it was argued that the prices should be conclusively presumed to be legal if they exceeded marginal or average variable cost. 51
It does not seem warranted to conclude that the CalComp opinion mandates adoption of a conclusion it flatly contradicts.
B. Prices Above Average Cost Should be Conclusively Presumed Legal.
Theoretical “perfect” competition does not contemplate stasis. Economists realize *990 that the prospects of supra-normal profits will induce entrepreneurs and innovators to supply new markets or invent new products. For a time, the innovator will be the market’s sole supplier, in possession of monopoly power, and able to charge a profit maximizing price, with returns well in excess of “normal” profits. This is the reward that accrues to those whose ventures into new fields are successful.
Charging a monopoly price and earning monopoly profits is in no way anticompetitive. To the contrary, it can provide the competitive incentive that keeps an economy dynamic and innovative. But the monopoly power that accrues to the innovator is not expected to be permanent. High profits will attract imitators. Initially, the imitators will price their products incrementally below the innovators’, sufficiently low to attract customers for the goods they can produce, sufficiently high to preserve as much profit to themselves as possible. Eventually the market’s supra-normal profits will attract enough new entrants so that the price will be driven down to the point where all suppliers are covering their costs and enjoying no more than a normal profit on their investment (price.will equal average cost). This state of perfect competition will continue until another innovator starts the innovation-monopoly-imitator-competition cycle over again.
However, the monopolist may seek to disrupt that competitive process in order to preserve to itself the benefits of monopoly. For instance, when faced with competition, the monopolist might choose to price so low that "the new entrant, no matter how efficient, is driven out of business. Once that is accomplished, the monopolist can return to profit maximization and supra-normal profits. Extinguishing the competitive threat in that fashion also serves to raise entry barriers. Potential entrants will be on notice that even if they have a good product, efficiently produced and marketed, the monopolist will not allow them to sell at a price which returns a profit.
The great problem in fixing a legal standard by which pricing conduct should be measured is that it is extremely difficult to distinguish between a monopolist’s price reduction that is a normal, expected component of the dynamic competitive process, and the predatory, undesirable conduct just described. If the law is overzealous in guarding against predatory pricing, it may well inhibit the competitive dynamics it seeks to promote.
It is tempting to consider the monopolist’s intent in order to distinguish between, pro and anticompetitive price reductions. But where is the line between illegal, undesirable, “predatory intent” and the legal, desirable intent to prevail in the competitive struggle? Appropriately low prices are to be encouraged, not discouraged. The punitive impact of the antitrust laws must not be permitted to compel high prices. Unless some objective, understandable standard is established for the guidance of businessmen, they must either forego competitive price decreases or risk punitive damages that might turn on some careless word once spoken in a board room. 52 Businessmen must have notice of what is violative of the law. A test based strictly on intent would not serve that goal.
The first issue to be addressed in adopting a cost-based test is whether any range of pricing conduct should be beyond the reach of the law. A “free zone” of price activity would preserve competitive incentives, but could also sanction undesirable conduct. If monopolists are allowed with impunity to lower price to a level equal to their average' cost, then the monopolist might lower its price from the profit maximizing level to a level competitors cannot meet in order to discourage entry and preserve the monopoly (limit pricing). Or, entry could be defeated and discouraged by temporary reductions to average cost levels, followed by a return to monopoly pricing once the threat had been vanquished. In *991 either case, the competitive process would be thwarted, and the monopolist could preserve its status without risking antitrust liability.
Areeda and Turner considered these possibilities, but concluded, as this Court does, that preservation of incentives is more important than the minimal threat to competition presented by a free zone above average cost. Only less efficient competitors would be eliminated by price manipulations above the average cost level. Competitors who are as efficient as the monopolist will continue to make at least normal profits, while the more efficient will profit more. 53
It would be all but impossible to distinguish between above cost limit pricing conduct and a monopolist’s pro-competitive reaction to lower priced competitors. 54 One external characteristic is common to both cases, a lowered price. Any attempt to attach liability to the one will surely inhibit the indistinguishable other.
Even a monopolist is permitted to compete in the battle for trade. 55 And the preservation of competitive incentives, even for that monopolist, is deemed such an important consideration that monopolization is condoned where condemnation would provide a disincentive to compete. 56 Protection of relatively inefficient competitors is a lesser consideration. 57 Monopolists can compete by being more efficient than their rivals, 58 and the most significant manifestation of that efficiency is a lower price. If the monopoly is attained or preserved because the monopolist is profitable at price levels where others are not, so be it. Lower prices and increased efficiency are to be fostered. This Court agrees with Areeda and Turner that price reductions which re-suit in prices that exceed defendant’s average cost should be conclusively presumed legal. 59
C. Prices Below Average Cost Should not be Conclusively Presumed Legal.
Efficiency is no longer the sole determinant of survival once price is permitted to drop below the monopolists’ average cost; cash reserves become paramount. If price is insufficient to cover average cost, the monopolist will be losing money, less efficient rivals will be losing money, equally efficient rivals will be losing money, and quite possibly, even more efficient rivals will be losing money. Average cost is the breakeven price level. Competitors with average costs equal to the monopolist’s, break even when the monopolist does. Competitors with average costs lower than the monopolist’s are more efficient and are making a profit when the monopolist is just breaking even. If a monopolist is permitted to set price as low as marginal cost with impunity, it can drive from the market all competitors whose average costs exceed the monopolist’s marginal costs unless their staying power is as great as the monopolist’s.
The difference between these cost levels can be quite significant. Evidence in the case at bar suggests IBM’s marginal costs or average variable costs might be as little as 50 percent of their average costs. If that is so, then IBM could destroy rivals who are twice as efficient (their average costs are one-half of IBM’s) simply because IBM had a bigger bankroll, and fear no antitrust liability under the Areeda and Turner rule.
Courts which have adopted the Areeda and Turner rule have not always considered *992 its potential for abuse. In Int’l Air Ind., Inc. v. American Excelsior Co., 60 the court recognized that equally efficient rivals could be driven out .but incorrectly assumed that more efficient competitors were safe. 61 The court in Weber v. Wynne 62 acknowledged the threat to equally efficient firms, but said nothing of the fate of the more efficient. 63 The Court in Hanson thought only the less efficient were endangered:
“If its prices were above its [marginal or average variable] costs, and nevertheless Shell’s did drive Hanson out of business, this can only be because Hanson was so inefficient that at prices at which Shell could make a reasonable profit he could not.” [64]
And finally, CalComp perpetuated that misconception in this Circuit:
“The thrust of this analysis is that price reductions up to the point of marginal cost are consistent with competition on the merits, since in this case only less efficient firms will be disadvantaged, while a firm pricing below marginal cost by definition incurs losses, so that competition on the basis of efficiency in this situation is frustrated.” [65]
Average cost is the point to which the. normal forces of competition will tend to lower price. To repeat what is self-evident, if a firm sells below' its average cost it is incurring a loss, equally efficient firms are incurring a loss, and more efficient firms (if their average cost is lower than the monopolist’s average cost but greater than the price) will also be incurring a loss. Only firms able to withstand losses for as long as the monopolist decides to inflict them will survive, others will perish. If a monopolist is permitted to set a price below its average cost, competition on the basis of efficiency is frustrated, and competition on the basis of wealth replaces it.
Areeda and Turner have opted to place a potent weapon in the hands of the monopolist. Their reasons for doing so should be carefully examined. They have identified four major reasons supporting their position: 66
(1) it will deter frivolous suits;
(2) a price floor above marginal cost will tend to preserve inefficient rivals;
(3) short-run welfare is maximized by pricing at marginal cost; and
(4) even if long-run welfare is not maximized, long-run consequences are too speculative to incorporate into a legal rule.
The first justification smacks of overkill. Surely it is not necessary to go so far as to allow a monopolist to destroy its competitors in order to protect that monopolist’s incentive to price competitively. A prophylactic rule is required in order to preserve incentives, but sanctioning all price levels above average cost serves this purpose as well. 67
Areeda and Turner’s second reason is interesting, but highly questionable. Apparently, they feel it is the monopolist’s function to destroy less efficient competition, 68 and that survival of inefficient firms is *993 undesirable. Even if there is some value in the elimination of competitors who are less efficient than the monopolist, this reasoning overlooks the tendency of a price floor at marginal cost to destroy equally efficient and more efficient rivals unless they have pockets as deep as the monopolist’s.
The welfare maximization argument is really the most significant. What is meant is simply that a monopolist with excess capacity, i. e., average cost exceeds marginal cost, will be making more goods available to the consumer at lower prices if allowed to price as low as marginal cost than the monopolist would if forced to utilize less of its capacity by producing less and pricing in excess of average cost. Social resources are wasted because there is a divergence between the cost of the incremental output (marginal cost) and the value of that output to consumers. And, since excess capacity exists, it would be a waste of social resources to attract new entrants who would construct even more facilities. 69
That short-run welfare maximization argument is short-sighted. While it is true that resources are most efficiently utilized and that consumers benefit when the monopolist prices at marginal cost, such beneficence cannot be expected to continue. Once the competitive threat has been extinguished, the monopolist will return to higher prices and profits. When that happens, society will suffer a greater welfare loss. Fewer goods will be produced, efficient capacity will be idled, and consumers will be paying much more than incremental production costs. Professor Scherer recognized that the inclination of monopolies to maximize their profits at the consumers’ expense would mean that welfare would be maximized if the monopolist’s price reduction was checked at some point above marginal cost. But he was unable to quantify precisely a limitation on price that would maximize long run welfare, i. e., a price rule that would insure that the discounted welfare loss from the anticipated future price gouge would be most offset by the temporary benefits of lower prices. 70 Professor Williamson is also suspicious of Areeda and Turner’s approach. In his opinion negligible benefits would flow from temporary price cuts to the marginal cost level, if followed by a return to monopoly pricing. 71
Areeda and Turner do not argue with Scherer’s long-run welfare analysis. They acknowledge that the net long-run consequences of marginal cost pricing might be adverse because of its exclusionary effect, but reject consideration of long-run effects because they are difficult to determine (their fourth reason for adopting a marginal cost standard). 72 Indeed, the welfare maximization problem is complex. In addition to the return to monopoly prices scenario, Scherer described several other cases, some of which would find welfaré maximized with price limitations above marginal cost, and some with price limitations below that point. 73 However, ease of application is a poor argument for adopting a rule that admittedly ignores important considerations of economic efficiency, especially when the main justification for that rule is economic efficiency. 74
In any event, the marginal cost test is not easy to apply either. In fact it is frequently impossible to apply.
It is almost too easy to criticize rule proposals in this area as administratively complex. The initial Areeda and Turner article has engendered much academic discussion and several counter-proposals. It seems that every attempt to formulate rules has been criticized by proponents of other rules as an unworkable solution. Areeda and Turner fare no better. Marginal cost is not *994 a cost recognized by accountants. It is a figment of the economist’s imagination. It is not recognized in the books of account, and thus, it is unlikely that any evidence of marginal cost could be easily developed. Areeda and Turner recognized that in their initial article, and suggested average variable cost could be used as a proxy, 75 and that in situations of strained capacity an average cost test could be applied. 76 But they seem to have retreated from that position of late, perhaps as a result of pointed criticism of the logic and feasibility of average cost and average variable cost tests. 77 They now emphasize that theirs is a short-run marginal cost standard, and that the other measures are simply approximations of marginal cost. If plaintiff can show that marginal cost is significantly higher than average cost, or if defendant fails to show that average variable cost is not significantly below marginal cost, then prices in excess of the surrogate standard are not safe under their rule. 78 The burden is always on someone to offer marginal cost evidence, but the data is not there.
Areeda and Turner’s justifications are not convincing. They argue that a marginal cost test will optimize social welfare. Then they admit that it will not. They argue that a marginal cost test is easier to apply then a long-run welfare maximizing test, then they suggest surrogates because marginal cost data is impossible to come by. But even if their theory had merit, if all the world’s economists were of one voice, and like Areeda and Turner, placed their faith in the monopolist to maximize social welfare by eliminating competitors from crowded industries through temporary provision of more and lower priced goods, the Congress and the courts have already placed their faith elsewhere. The goal of welfare maximization through proper resource allocation is to be accomplished by a system of effective competition, not by reliance on the presumed beneficence of a monopolist. 79 In the words of Justice Black, the Sherman Act “ . . . rests on the premise that the unrestrained interaction of competitive forces will yield the best allocation of our economic resources, the lowest prices, the highest quality and the greatest material progress . . ,” 80 The economic rationale behind the act cannot be overcome by arguments that the monopolist has provided economic benefits. 81
More than just economic considerations impelled the Congress to adopt competition as the Acts’ unequivocal policy. Chief Justice Warren said:
“. . .we cannot fail to recognize Congress’ desire to promote competition through the protection of viable, small, locally owned businesses. Congress appreciated that occasional higher costs and prices might result from the maintenance of fragmented industries and markets. It resolved these competing considerations in favor of decentralization.” [82]
Economic considerations aside, justifiable apprehension of excessive economic power concentrations underlie the law’s aversion to monopolies. 83
*995 Areeda and Turner have made a policy judgment. The economic analysis used to justify that judgment is incomplete, and the judgment itself stands contradicted by the economic, political, and social policies of the Sherman Act.
A conclusive presumption of the legality of an unprofitable low price, merely because it is above marginal cost, a cost which is all but incapable of proof, would truly be a “defendant’s paradise.” 84 This Court rejects it.
D. Prices Below Average Cost are Illegal if They are Unreasonable.
There are no simple bright line tests that are universally applicable. Williamson proposes a complex set of “per-se” rules, a different rule to be applied depending upon, whether the dominant firm is reacting to new entrants or already established firms, whether a short, long or intermediate-run is involved, and whether the demand function is normal or conditions of chronic excess supply are found. 85 Scherer calls for evaluation of the many criteria that would affect long-run welfare maximization, examined in the light of the monopolist’s intent and the structural consequences of its pricing behavior. 86
Professor Schmalensee recently attempted to apply the Williamson rules to the facts of a monopolization case heard before the Federal Trade Commission, but found them imprecise, difficult to apply, and based on an economic model probably at variance with real world conditions and certainly at variance with the facts of the case he discussed. 87 Having already rejected Areeda and Turner’s approach as incapable of being “defensibly applied to all, or even most, cases,” Schmalensee concluded that Scherer’s rule of reason analysis was the only economically defensible policy choice and that an average cost test could be employed to dismiss patently groundless charges of predation. 88
This approach is not wholly without precedent. In a thoughtful concurring opinion to the case discussed by Schmalensee, Commissioner Pitofsky distinguished Ninth Circuit cases and applied the same rule. 89 The average cost standard adopted here also finds support in several trade rules promulgated by the Federal Trade Commission, 90 in state statutes prohibiting below cost sales, 91 in a predatory pricing rule propounded by Professor Posner 92 in a recommendation by the National Commission for the Review of Antitrust Laws and Procedures that the marginal cost standard be abandoned in favor of a more flexible analytical approach, 93 and finally by the Supreme Court’s holding in Utah Pie Co. v. Continental Baking 94 that a jury could infer price cuts were predatory if price was below average cost. 95
*996 If the monopolist chooses to price below average cost, something unusual is taking place. A decision has been made to sell goods at a loss, and such activity is worthy of investigation. It may be a predatory act with the design and effect of starving out competitors, or it may in some way be justified.
. Prices below average cost would be reasonable if the monopolist was merely liquidating excess, perishable, or obsolete merchandise. 96 Prices below average cost would be warranted if shrinking demand forced the monopolist to minimize its losses by selling at the best price-cost relationship available to it, 97 or where the industry suffers from chronic excess capacity. 98 And the same might be true of promotional pricing, meeting competitors’ prices, or even occasional price wars, if carried out under appropriate circumstances. 99 Intent evidence can prove helpful here. The monopolist’s own evaluation of the situation, whether it thought it was cutting losses or cutting throats, can help to clarify the nature of the acts undertaken.
A monopolist’s prices should be judged according to whether or not they are reasonable, and prices above the monopolist’s average cost should be per se reasonable. That standard will be applied to the facts of this case.
E. IBM’s Pricing Conduct.
IBM did not simply reduce the prices of its existing products, instead it responded to PCM competition by introducing several “new” products, all of which were, to some degree, repackaged versions of prior offerings. By offering the new products at lower prices, while maintaining higher prices on the older versions, IBM managed to avoid the disastrous effect an across-the-board price reduction would have had on revenues from installed machines, and yet still had a vehicle with which to challenge the PCMs. It is the prices of these new products that Transamerica claims were predatory.
In 1964 IBM introduced a disk drive known as the 2311. In 1965 an improved version, the 2314, was announced. At first, 2314-type disks were only available in a package of eight disks or “spindles” (plus one spindle as a spare). Later, IBM offered a one spindle version (the 2312), a two spindle version (the 2318), and a four spindle version (the 2313), all of which could be connected to the CPU through a control unit (the 2314). The functional characteristics (access time, capacity, and data transfer rate) of all versions were practically identical, and all are known as “2314-type disks. The PCMs aso marketed 2314-type disks. Most PCMs offered some functional improvements (principally a reduced access time), all offered only single spindle versions, and all offered lower prices than IBM.
In September of 1970, IBM announced a new computer system, the System/370, Model 370/145. At the same time, a three-spindle version of the 2314-type disks was announced for use on that new system.. This new version was known as the 2319A and was code-named “Mallard.” It connected directly to the System/370 Model 145 CPU (and later to the System/370-135) by means of an integrated file adapter (“IFA”) which eliminated the need for a control unit. The 2319A was available at a monthly rental of $1,000, a price substantially less than IBM’s one, two and four *997 spindle versions (if comparison is made on a per spindle basis, i. e., monthly rental divided by number of spindles), and a price Transamerica claims was predatory.
Three months after the 2319 announcement, in December of 1970, IBM announced a three-spindle version 2314 for use on the older System/360 CUPs. This was known as the 2319B. It attached to the 360 CPU through a control unit (2314B), and a channel. At the same time IBM announced the 2319A2, a second set of three spindles for attachment on the 370/145 CPU behind the first set (the 2319A). The 2319B and the 2319A2 were both priced at $1,000 (monthly rental). Transamerica claims that price was predatory.
In July of 1971, still another three-spindle 2314-type was announced: the 2319A3. Unlike the other versions, the 2319A3 (code named “Zoom”) had its own power supply. It was to be used exclusively on the 370/135 CPU, where power constraints had previously limited configurations to a maximum of five spindles. Zoom extended that maximum to eight. Because of the extra power equipment, the 2319A3 was priced at $1050 monthly rental, slightly higher than the other 2319s. Transamerica also claims this price was predatory.
In 1964 and 1965, IMB announced various models of the 2401-type, tape drives and 2803 control units. Improved tape drives, the 2320s, were announced in 1968. In November of 1970, IBM announced an improved control unit, the 3803, and three new tape drives to go with it, the 3420 Models 3, 5 and 7. The functional characteristics of the new tapes (data transfer rate, tape speed, and tape density) were practically identical with those of the 2420. The combined tape program was code-named “Aspen.” Transamerica claims that the Aspen prices were predatory.
At the same time as the Aspen announcement, November of 1970, IBM announced a new model 2401 tape drive, the 2401 Model 8. Two control units for that model, the 2803 Model 3 and the 2804 Model 3, were also announced. This program was codenamed “Mandan.” The 2401 Model 8 was almost identical to the seven-track version of the 2401 Model 2, but the price was substantially lower. Transamerica claims that the Mandan prices were predatory.
IBM analyzed the financial aspects of each of these programs (the 2319A, the 2319B/A2, the 2319A3, Aspen, and Mandan) prior to announcement in order to predict the costs, revenues, and profits that could be expected. A common procedure was followed in all pre-announcement financial analyses. The Product Manager, an employee of the engineering division, identified the forecast assumptions. These included a description of the product, competing products, follow-on products, delivery schedules, etc. The forecast assumptions were then used to project the number of units that would be sold and leased at various price levels (the demand function), and from this the program revenue at those prices could be calculated. Direct costs (those that could be readily identified with the product in question) were estimated by the manufacturing division. Indirect costs were assigned to the products through three apportionment methods. By allocating a portion of the indirect costs (such as corporate overhead) to each product, as well as the direct costs, IBM followed a full-cost philosophy of financial analysis and hoped thereby to be able to predict a program’s profitability. Full costs, direct plus indirect, were subtracted from revenues, to give predicted program profits.
Each of the product programs at issue here was expected at announcement to return substantial profits to IBM. Zoom (the 2319A3), by far the smallest program, was projected to return a $900,000 profit, or 415 percent of revenue; Mandan was expected to produce $29.2 million in profits, or 33.6 percent of revenue; Aspen was projected to bring in $473.3 million in profits, or 26 percent of revenue; the 2319B/A2 program was predicted to bring in a $71.6 million profit, or 31.5 percent of revenue; and finally, the 2319A was projected to bring in $55.2 million in profits, 31 percent of its revenues. It is only through the application of several “adjustments” to the IBM an *998 nouncement financial analyses that Transamerica is able to claim that any of the prices at issue here were below average cost. Each of the adjustments Transamerica suggested is discussed below.
1. The Revenue Apportionment Adjustment.
As mentioned above, IBM uses three apportionment techniques for determining what indirect costs a particular product or program should bear. The indirect costs that are most closely associated with engineering are apportioned among products according to the relationship between that product’s direct engineering costs, and total direct engineering cost. Similarly, indirect costs that are associated with production are allocated to the ratio between the product’s direct production costs and total direct production costs. Finally, indirect costs that are not otherwise apportionable, such as marketing, general and administrative expenses, are apportioned to products according to the relationship between that product’s revenues and total revenues.
IBM allocates over 50 percent of a product’s costs through the revenue apportionment method. Transamerica complains that this method provides a monopolist with too much flexibility. They say that where a price cut is involved it permits the monopolist to drastically reduce price and still not be pricing below cost because reduction in price means a reduction in revenue and thus a reduction in revenue-apportioned costs. When revenue is reduced on one .product, some of the costs previously associated with it are automatically transferred to other products. Transamerica characterizes all the products at issue here as disguised price cuts: The'2319s as disguised price cut versions of the 2314; Mandan as a disguised price cut on 2401 Model 2; and Aspen as a disguised price cut on the 2420.
Through its expert accountant Peterson, Transamerica invokes “generally accepted cost accounting principles,” and adjusts the revenue-apportioned costs assigned to each of the products at issue. Assuming that he was in fact dealing with price cuts, Peter-. son increased each new program’s revenue-apportioned costs by multiplying the costs assigned at its announcement by the ratio between the “old” price and the lower “new” price.
The basis for Mr. Peterson’s adjustment is suspect. The phrase “generally accepted cost accounting principles” does not appear in accounting literature. It, like the adjustment it supposedly requires, appears to have been invented for purposes of this lawsuit.
The adjustment is predicated on the existence of a price cut when, at least in some circumstances, different products are clearly involved. And, it assumes that the cost allocated to the product at the old price was somehow “correct” and should be maintained, when in fact it is precisely the inability to associate these expenses with any particular product that makes an apportionment method necessary. It would be equally logical (or illogical) to assume the previously assigned costs were too large and the new ones are “correct.”
Revenue apportionment is a generally accepted and commonly used method of indirect cost allocation. IBM used it consistently long before making the cost analyses at issue here. Peterson’s adjustment would abandon that consistency, and with it any assurance that full costs are recovered. Increasing the costs assigned to one product without somehow decreasing the costs that other products must bear would lead to an over-recovery of costs, and make the full-cost recovery rationale and program profit predictions meaningless.
All methods of indirect cost recovery are arbitrary and flawed in some respects. The flexibility of revenue apportionment is apparent, yet this Court will not attempt to “correct” that characteristic by mandating the use of a patch work system based on one accountant’s views of the facts of this particular case. To do so would not only be bad accounting, it would be bad law. Businessmen would have no notice of what was required to be done if arbitrary adjustments were allowed to be applied after the fact to their reasonable and consistent cost accounting.
*999 This is not to suggest that under other circumstances cost accounting adjustments might not be warranted. Where a defendant itself abandons consistency in attempts to circumvent the law, adjustments might be appropriate. That, however, is not the case here. The revenue apportionment adjustment is rejected.
2. The Manufacturing Cost Adjustments.
All the 2319 programs were reuse programs. For instance, 80 percent of the 2319As were not newly built, but were built by reconditioning and slightly altering used 2313s that had come off rent. Mandan was a total reuse program; all the units involved were reused, off rent, tape drives and control units. In the announced financial analyses for these products, IBM assigned a full manufacturing cost to those units that were to be newly built, but the reused units were costed at the cost of conversion efforts plus their remaining undepreciated net book value. Since IBM uses an accelerated depreciation technique (sum of the years’ digits), and since the 2313s had already been depreciated for three years (of a total four year depreciation life), the net book value used to represent the manufacturing cost of the 2319s was only 10 percent of the cost of building them new. With Mandan, the reused units were even older, and, since their net book value was insignificant, the only manufacturing cost reflected was the minor expense involved in reworking the units.
Peterson testified that this was in conflict with the “benefit principle”, which he termed the single most important “generally accepted cost accounting principle.” He suggested two alternative adjustments. The first would simply account for all machines at the cost of building them new, whether that was in fact done or whether they were “retreads.” The second, more complex, adjustment would require the reworked product’s cost accounting to bear a manufacturing cost equal to the new cost multiplied by the ratio between the expected rental life of the new product and the total lives of the old and the new products. In effect, the second adjustment would be the same as using net book value, had IBM depreciated over the total life of both products using a straight-line method.
IBM’s Financial Procedures Manual calls for using new cost in some situations, scrap value in others, and permits net book value to be used in yet other cases. IBM argues that the manual would have permitted the use of scrap value and so their use of net book value was conservative. Transamerica interprets the manual differently, claiming it mandates the use of new cost. Mr. Peterson contradicted that. He testified that IBM complied with the manual in its handling of reused products. This Court agrees with Mr. Peterson, and finds that IBM followed its cost accounting procedures consistently. Until some cost accounting principles become generally accepted (and others generally rejected), it seems that a reasonable procedure, consistently followed, should not be vulnerable to subsequent manipulations. IBM’s depreciation method is reasonable and generally accepted. The cost accounting for the original product reflected its full cost, and use of net book value is a reasonable means of apportioning some of the cost to the new product. Deviations from, consistent procedures in this area would cause the same inconsistencies as would be created by the revenue apportionment adjustment. Accordingly, this Court rejects the manufacturing cost adjustments.
3. Shorter Rental Lives.
The programs at issue here were predominantly leasing programs. New units were expected to be sold. Consequently, the length of time forecasters predict units to be producing rental revenues, the average rental life, is a critical factor in determining a program’s profit. IBM predicted a 48 month life for the 2319A, 50 months for the 2319B/A2, and 58 months for Mandan. No witness ever opined that IBM should have chosen shorter lives, but Peterson produced alternative profit Calculations based upon much shorter (and therefore less profitable) *1000 lives for these products. The apparent inference was that the lives chosen at announcement were unreasonably long, and that the shorter lives should have been substituted.
This Court refuses to make that inference. Although the alternative lives presented were found in IBM documents, Transamerica has invariably chosen the shortest lives reflected in any IBM document while ignoring longer lives for other products in the same document, and ignoring longer lives for the same product in other documents. The shorter lives all appear in documents dated well after announcement and can be explained in part by the intervention of unexpected increased PCM successes, in part by the intervening funding of impacting IBM programs, and in part by the unreliable nature of the documents themselves, at least as far as the reliability of the lives contained therein is concerned. There is no evidence that anyone in IBM contemplated shorter lives at any time prior to the announcement dates. One of Transamerica’s witnesses, a man experienced in the financial aspects of computer leasing, testified that he thought the most reliable estimate of life would be that made by IBM, and that he didn’t know how he could go back in hindsight and assign a different life. Absent some evidence that shorter lives were, in fact, required, this Court must decline to substitute shorter rental lives.
4. Cost of Capital Adjustment.
'As already indicated, this Court would consider prices to be per-se reasonable only if they exceeded average cost, where average cost is defined to include a reasonable return on capital invested. If a company is thinly capitalized, most of these capital costs will be reflected as interest on the books of account. Conversely, if a company depends upon equity capital to fund its programs, no interest expense will appear, and the opportunity cost of the capital invested in the program will have to be estimated and imputed. Unless all capital costs are in some way included in the cost-price equations, a per-se sanctioning of price in excess of average cost would favor strong firms, able to use equity capital over weaker firms depending on borrowings.
Estimating capital cost is not an impossible task. It requires some estimate of the amount of capital invested in a program and of the duration of that investment, and some basis for finding a “normal” rate of return on such an investment. Transamerica proved none of these. No evidence of invested capital has been introduced and no witness has ventured an opinion as to what the cost of capital ascribable to any program might be. Reliance instead is placed on a “Program Evaluation Technique” (“PET”), a method of estimating cost of capital described in IBM’s Financial Procedures Manual.
Transamerica first revealed it was relying on PET to indicate IBM’s program revenues were insufficient to cover its cost of capital during counsel’s closing argument. As a result, no witness has had an opportunity to be examined or cross-examined as to the propriety of such an approach. Some flaws are apparent, however.
The method applied by Transamerica is the least sophisticated of several PET options. It employs a matrix with which one can determine a current value of profits (as a percent of revenue) if one knows the rental life of the program, and the program’s expected (undiscounted) profits percentage. For instance, if a program with a 36 month rental life were expected to yield a 20. percent profit, then the matrix would indicate that when discounted to present value the profit returned would be only 8.8 percent. The PET explanation describes the .cost of capital as a cutoff rate established by management, and sets that rate at 7 percent. Assuming that a 7 percent return is a normal return, then the hypothetical program’s profits exceed the cost of capital invested (8.8 percent exceeds 7 percent).
But there is a caveat that concludes the description of the PET matrix which warns against its use if the program being analyzed deviates from the pattern the matrix *1001 developers assumed was typical. Clearly they have made some assumptions as to the level of investment, and the annualized cash flow of the program. There was no showing here as to what those assumptions were, or that the investments in any of the programs at issue are similar. Moreover, even if that evidentiary gap could be bridged, the matrix shows that programs with profits in excess of 25 percent more than cover capital costs at all rental lives. The programs at issue here, once the specious adjustments rejected above are ignored, were all expected to earn profits in excess of 25 percent and thus in excess of capital costs. 100
Transamerica’s cost of capital adjustment suffers from a failure of proof. It was plaintiff’s burden to establish this element of its case. This Court will not attempt to estimate that which plaintiff has failed to quantify.
5. FTP Adjustment.
FTP (Fixed Term Plan) was a long-term lease plan offered by IBM on peripheral equipment. FTP was announced in May of 1971. Prior to that time customers could either buy peripheral equipment, or lease it on a month to month basis. With FTP, IBM offered customers a one year lease at an 8 percent discount from the monthly rental price, or a two year lease at a 16 percent discount. Transamerica offered no evidence that FTP caused a decrease in the profits of any product. Indeed, all the evidence indicates that IBM projected that FTP would increase revenues and profits over the levels expected if FTP were not announced.
Transamerica’s argument for an FTP adjustment to the profits of the programs at issue here, did not surface until after the close of evidence. They compare an IBM projection of program profits at one time (before and without application of FTP) with another and later profit projection which included the application of FTP. They conclude from the comparison that FTP caused revenues and profits to decline, and seek to apply the computed decline pro-rata to the revenues and profits of the products in issue. This method is unacceptable. It fails to give effect to other circumstances which changed between the two projections, and, more importantly, it ignores the fact that the second projection, specifically designed to compute the effect of FTP under the circumstances existing at that time, showed that FTP could be expected to increase both revenues and profits. Unrelated forecasts cannot be so manipulated.
FTP undoubtedly created a new price-cost relationship, one not reflected in the announcement financial analyses, and one probably unanticipated at that time. But there is evidence to indicate that FTP reduced costs and increased revenues; there is no convincing evidence to the contrary. Accordingly, this Court refuses to decrease projected profits as a result of FTP.
6. Additional Evidence that Prices were Reasonable.
This Court has concluded that the IBM announcement financial forecasts for the 2319s, Aspen and Mandan were reasonable, and that IBM reasonably anticipated that those prices, even after FTP discounts were applied, would exceed average costs. 101 Although profit expectations at the product’s *1002 introduction may be the primary consideration in analyzing a firm’s pricing behavior, 102 evidence of actual profitability is also relevant.
IBM’s financial analyses are updated at various times after a product’s announcement. These updates reflect the realities of marketplace experience with the product, and thus more accurately reflect its profit picture. Financial analyses of the 2319A and 2319B done one year after announcement showed those products to be expected to produce profits (as a percent of revenue) of 40.2 percent and 34.5 respectively. One year after announcement, Mandan was expected to bring in 25.8 percent in profits, and Aspen 17.4 percent. A little over a year after that, Aspen’s profits had improved to 20.5 percent. A financial analysis of all the 2319s combined and updated through 1978 indicated that the combined programs yielded 28.4 percent in profits. This evidence confirms the findings that IBM reasonably expected substantial profits on all the peripherals at issue; that the prices (including FTP) were above cost, and reasonable; and that the prices were legal.
Even if those prices had been slightly below cost, and their reasonableness thus open to inquiry, this Court would not hesitate to find them reasonable. The 2319s and Mandan were programs which reused third generation equipment at a time when fourth computer generation had already been announced. That older equipment was soon to become obsolete, and it is reasonable for a manufacturer in such circumstances to reduce price and squeeze as much ‘revenue as possible from equipment before demand dwindles.
Reasonableness is also indicated by comparisons between the PCMs’ prices and IBM’s. Such comparisons indicate that IBM’s prices did not undercut the PCMs’ prices, they did not even equal them, they merely closed the gap; the PCMs’ prices were still lower.
Shortly after the announcement of FTP, IBM announced price increases on some of its CPUs. Transamerica suggests that this was done to offset profits lost on peripherals as a result of FTP, and was a use of monopoly power in one market to gain a competitive advantage in another. The evidence indicates otherwise. First, FTP was expected to increase long term profits. Second, if a competitive advantage is gained by lowering price to a level in excess of average cost, the competitive advantage is efficiency related, not monopoly related, and equally efficient competitors will continue to make at least a normal profit. If the monopolist foregoes some excessive profits but still prices above average cost, it is immaterial how, or whether, the foregone profits are recouped. Conversely, if price is set below average cost, recouping losses by monopoly pricing in another market would be a strong indication that the lowered price was unreasonable. The critical issue is whether price exceeded average cost. If, as here, it did, the price is reasonable, will eliminate only the less efficient, and does not violate the antitrust laws.
This Court finds that all the prices at issue here exceeded average cost, were reasonable, and were legal.
V.
DESIGN CONDUCT
It is not difficult to imagine situations where a monopolist could utilize the design of its own product to maintain market control or to gain a competitive advantage. For instance, the PCMs were only able to offer IBM’s customers an alternative because they had duplicated the interface, the electrical connection between the IBM System/360 CPU and the IBM peripheral (or peripheral subsystem). Had IBM responded to the PCMs’ inroads on its assumed monopoly by changing the System/360 interfaces with such frequency that PCMs would have been unable to attach and unable to economically adapt their peripherals to the ever-changing interface designs, and, if those interface changes had no purpose and effect other than the preclu *1003 sion of PCM competition, this Court would not hesitate to find that such conduct was predatory.- Or, if a monopolist frequently changed the teleprocessing interface by which its computers communicate with remote terminals in such a way that its terminals would continue to function while others would fail, and, if the only purpose and effect of the change was to gain a competitive advantage in the terminal market (where the monopolist lacked monopoly power), that use of monopoly power would be condemned.
It is more difficult to formulate a legal standard for design conduct than it is to imagine clearly illegal situations. Any such standard must properly balance a concern for the preservation of desirable incentives with the need to prevent monopolization by technology. Like pricing, equipment design can have pro-competitive as well as anti-competitive aspects. Truly new and innovative products are to be encouraged, and are an important part of the competitive process. For this reason, the acquisition or maintenance of monopoly power as a result of a superior product does not violate the Sherman Act. One court has even suggested that where there is a valid engineering dispute over a product’s superiority the inquiry should end; the product is innovative and the design is legal. 103 That view, probably the result of a concern for the creativity that has characterized the history of computers, is overproteetive. It ignores the possibility that a superior product might be used as a vehicle for tying sales of other products, and would pronounce products superior even where the predominant evidence indicated they were not.
Another approach would be to examine the designers’ intent. If a technological design were chosen for an illegal purpose (such as to effectuate a tie) and if that purpose was fulfilled, it would be illegal. 104 If that standard were to apply only where the intent was solely an illegal one, creativity would not be stifled. But usually many results are intended, and if only one, even the predominating, intent is illegal, and thus punished, legitimate incentives will be imperiled. Discerning corporate intent is seldom easy, and, in any event, the law against monopolization is much more concerned with the effect of conduct rather than with its purpose.
A more generalized standard, one applicable to all types of otherwise legal conduct by a monopolist, and one recently adopted by the Ninth Circuit, must be applied to the technological design activity at issue here. If the design choice is unreasonably restrictive of competition, the monopolist’s conduct violates the Sherman Act. 105 This standard will allow the fact-finder to consider the effects of the design on competitors; the effects of the design on consumers; the degree to which the design was the product of desirable technological creativity; and the monopolist’s intent, since a contemporaneous evaluation by the actor should be helpful to the factfinder in determining the effects of a technological change.
A. Interface Changes.
The PCMs were engaged primarily in replacing only peripherals attached to IBM systems. To the extent that IBM could prevent (or make inordinately expensive) the physical attachment of such gear, it had the power to eliminate competition. Transamerica complains that this was what was done by interface changes made when the new products at issue here were designed.
The interface between the Aspen tape drives (3420 Models 3, 5 and 7) and the Aspen control unit (3803) differed significantly from the interface utilized by the immediate predecessor tapes (2420s) and *1004 control units (2803s). PCMs wishing to attach their 2420 copies to the new and improved control unit were effectively precluded from doing so. The cost of adapting a 2420 to the new interface format were prohibitive. Transamerica charges that the interface change was unnecessary, and that its only purpose and effect was to preclude PCM competition for tapes attachable to the 3803.
The evidence is to the contrary. Not only was the 3803 a superior product (as Transamerica concedes), the interface itself was a superior product. It had fewer wires and connectors. This meant switching between tapes and control units could be done more simply (fewer switch points), enabling IBM to integrate the switching mechanism into the control unit and thereby eliminate the need for a costly and bulky separate switching device. The pulses on the wires were multiplexed. That meant more sense information could be transmitted and diagnostics and maintenance were improved. Signals were converted from analog to digital at the tape head (rather than in the control unit), meaning information could be transmitted more reliably (with less noise interference). And, tape drives were attached to the 3803 radially (rather than serially), making maintenance and computer room layout easier. All of these changes required interface changes, and all were undeniably improvements.
Transamerica (which never owned any 2420s anyway) will not be heard to complain that it was somehow injured by an improved product. The Aspen interface was just that. It was adopted by IBM because it was a product improvement, and even if its effect was to injure competitors, the antitrust laws do not contemplate relief in such situations.
The Mandan interface was also changed. It was different from the interface between the predecessor 2401 Model 2 tape drive and its control unit, the 2803 Model 1. But this interface change was truly minimal. Five wires were added, and two wires were removed. The evidence was uncontradicted that: the change was purposeful, to identify the model to the control unit; the change was ordinary, all 2401 models were so identified; and the impact on PCMs was insignificant, the technical information was made public and with it a qualified engineer could adapt a 2401 Model 2 interface in one hour. There was no evidence that any PCMs wished to hook onto the Mandan control unit, but had they so desired, this interface change would certainly not have precluded it.
A more complex problem is presented by the interface between the 2319A and the IFA. The IFA represented a different mode of peripheral attachment than had been utilized on most System/360 CPUs. Instead of a CPU to channel to control unit to peripheral attachment, the IFA resided in the CPU itself and attached directly to the peripheral device. That eliminated the need for control units and channels, saving the user space and money. The IFA, unquestionably a superior product, was an optional feature for sale or lease on the 145 and 135 System/370 CPUs.
From approximately 1969 until mid-1970, IBM’s Endicott laboratory, responsible for the design of the 145, planned to design the IFA so that IBM 2314 — type disks would attach with minor modifications. This was known as the Apricot plan. IBM’s Hursley laboratory in England, responsible for the development of the 135, the sister to the 145 CPU, opposed this plan and, as early as 1968, and again in 1969, had proposed specifications for disk drives that were ultimately adopted by IBM in the summer of 1970 and announced as the 2319, code named Mallard. Under Apricot, a user could have attached IBM 2312s (one spindle), IBM 2318s (two spindles), or IBM 2313s (four spindles) in any combination. IBM 2314-type spindles would attach because the planned Apricot interface duplicated the interface between the 2314 control unit and the drives. PCM 2314-type drives would not have attached without modification to the Apricot IFA, but because the 2314 interface was well known, the PCMs could have quickly and easily duplicated that interface and made their drives available for *1005 attachment to the IFA. Primarily because of this exposure to PCM competition, IBM dropped Apricot and adopted the Mallard (2319A) plan.
With Mallard, one of the spindles was removed from the frame of a 2313. Into that vacant space IBM moved some of the control electronics that would have resided in the CPU under Apricot. The result was that the interface into the IFA was different under Mallard than under Apricot. And because it was different, and unknown to the PCMs, their efforts at copying the interface had to be delayed until they could examine the device IBM had produced.
With Mallard IBM bought itself some time. The interface change meant that PCM copies would not be available as quickly as if a known interface had been used. In a lease oriented market, where product obsolescence limits rental lives, time is money. If the PCMs could be sufficiently delayed, their market might even be limited to the point that the expense and trouble of duplicating the 2319A interface would’ no longer be worth it.
IBM’s predominant intent in adopting the Mallard design was undoubtedly to preclude or delay PCM competition and gain a competitive advantage. However, other considerations also played a role. Mallard alleviated some difficult technical problems. It saved some space in the space-critical 135 CPU; it resulted in the location of a diagnostic device (the CE panel) near the drives where it was easier’to use; it reduced cable lengths and electrical noise interference problems; and it simplified engineering control and record keeping for IBM by assuring that the engineers and the laboratories that were most familiar with the disk-related electronics would be responsible for them. On this record, the Mallard concept was a superior design.
Mallard’s effect should also be considered. It appears the market segment affected by the 2319A was not something that held a great attraction for PCMs. The 2319A was expected to be an interim disk file for the 135 and the 145. Once IBM’s 3330 became available, 2314-type devices would no longer be very attra

[Text truncated at 120,000 characters. The full text is on the page linked above.]

---

Source: Frix Law Library, https://www.frixlaw.com/law-library/cases/8868473. Public record. Not legal advice.