In an antitrust trial held earlier this year in Philadelphia, IBM was acused of several illegal business practices by AMI (CI Nos 725, 727). AMI is a tiny company that makes its living refurbishing and reconfiguring IBM mainframes; its customers are mainly computer leasing companies. In the prior parts of this series, AMI presented two aspects of its case. AMI said that a surcharge called IWSC, placed on certain mainframes shipped internationally, unfairly restrained trade. AMI also accused IBM of pressuring an independent parts distributor into withholding certain components AMI needed to perform its work.
Juggled its parts
The heart of AMI’s antitrust case against IBM, and the issue with the greatest complexity, surrounds IBM’s pricing of upgrades for its 308X series. AMI has tried to show that IBM inflated its charges for the parts used in mainframe upgrades, and reduced its charges for the associated labour. This prevented AMI from competing, for AMI’s upgrade business was based on using IBM’s parts and its own technicians. AMI also argued that IBM juggled its parts prices, and the resultant costs for certain components available only from IBM made it nearly impossible for the refurb house to do business. AMI’s revenues dropped from more than $15m in fiscal 1982 to about $5m in fiscal 1985. AMI attributes this loss of two thirds of its trade directly IBM’s pricing practices, which it alleges were deliberate and illegal. In court, AMI was obligated to demonstrate that IBM caused it more than inconvenience by its pricing; that in itself is not illegal. AMI’s challenge was to show that IBM distorted its prices to cut AMI out of business, and that this was done by means that violate the law. One tactic that can be illegal under antitrust law is tying. Tying is a way of pricing goods and services so that one cannot be purchased at a reasonable price without the other. For example, if IBM charged an arm and a leg for upgrade parts, but pretty much gave away the labour, it might have tied the parts to the labour in violation of law. Also, if IBM’s pricing was designed to knock out legal competition, that pricing might be considered predatory. Under US antitrust law, IBM has wide discretion in pricing its goods and services. IBM’s defence against AMI’s claims was in part an assertion that AMI’s problems in the upgrade business were the result of changing technology, not IBM’s predatory pricing, or IBM’s tying of one good or service to the purchase of another. IBM has argued this in prior antitrust cases. As this aspect of the case unfolded in the courtroom, IBM and AMI gave the judge quite an education in the workings of the computer business. However, only a fraction of the lawyers’ lectures will enlighten users, or the general public. Much of the material the lawyers dug up was sealed from view, even though excerpts were included in testimony. This censorship came about because one party or the other believed that potential injury could result from the disclosure of certain proprietary information. The lawyers argued that the threat to their clients’ privacy outweighed any public interest. Even so, the documents that were not sealed, but which instead were kept in the public record, shed new light on IBM and the way its competitors in leasing and the refurb businesses conduct their affairs. Users are only vaguely aware that IBM bundles bits and pieces of computers into kits for its technicians to install.
IBM calls such a kit an MES, which literally means Miscellaneous Equipment Specification. This is a misleading term. An MES is a good deal more than a mere technical description; it is a complete set of parts and related documentation. IBM sells an MES with or without labour charges, but the prices are nearly identical. IBM’s installation fee for a 308X upgrade is only 2% of the hardware charge. This wasn’t always the case. In prior computer generations, the labour content in an upgrade was a much larger part of the cost. With the 308X line IBM introduced what it calls thermal conduction modules, or TCM
s. A TCM is a metal can about six inches square and two inches deep. It is crammed with logic pressure bonded to a multi-layer ceramic substrate. The circuits are in a chamber filled with inert gas, and this chamber is cooled by an adjacent chamber, called a cold plate, through which chilled water is passed. In a 308X mainframe, TCMs are plugged into holders called Clark boards that each have nine 1,800-pin sockets. A whole 3083 CPU can be built of a fewer than a dozen TCMs; a dyadic 3081 requires only 20 to 30, depending on what features it sports. If a part in a 308X fails, or a chunk of a computer must be replaced during an upgrade, only a small amount of effort is required. It’s much easier for a technician to change a TCM than to swap an equivalent amount of IBM’s prior generation of circuitry. IBM is very proud of its TCM, and its struggle to mass produce it. According to Gerry Granito and G Kenneth Williams of IBM’s Poughkeepsie Laboratory, the company began to develop the TCM in 1967. It took IBM 14 years – until 1981 – and over a billion dollars, to get it out the door in a commercial product. The 308X computer didn’t come out on time, IBM’s lawyer Evan Chesler said on the first day of the trial. It was so difficult to make them [TCMs] in large volumes and get enough of them out at the end of the line that worked… that it took two years longer than IBM had thought to get it into the field. Meanwhile, the 3033 computer was… by computer industry standards, an old machine. With the 308X way behind schedule, IBM had to come up with a scheme that would keep 3033 orders rolling in. It had to promote the 1977 vintage 3033 as a living system rather than the dying one it actually was. So IBM announced new versions of its flagship 3033-U. These were actually downgraded versions of the faster machine, called the 3033-N and 3033-S.
Sales reps flogged
Simultaneously, IBM added features to all the 3033s that improved performance somewhat. As IBM’s sales reps flogged its renewed 3033s to customers, the manufacturer’s TCM engineers were flailing away at their stubborn problems. IBM’s smaller model 3033 mainframes were distinctly different from the machines they succeeded, the comparably-powered 3032 and 3031. The 3033s embodied the germ of an idea – copied from Amdahl Corp – that would flower on the succeeding 308X products. The 3033 models N and S could be upgraded in the field to versions with more MIPS. Upgrading was the user’s alternative to trading in smaller machines for larger ones. The upgrade potential of the smaller 3033s was pitched as a means to preserve the user’s investment in hardware, much the way the upward compatibility of 360 and 370 family computers was marketed as a way for the user to preserve an investment in software. There was something in it for IBM, too. Competitors now had to cope with sales prospects who didn’t have to replace their computers in order to get more power.The labour involved in upgrading 3033s was considerable, and permitted third parties like AMI to compete successfully with IBM’s service engineers. But these machines would be the last generation of mainframes that afforded independents like AMI such a golden opportunity. The 308X machines, like the late-model 3033s, can be upgraded on the customer’s premises. But most of the price of a 308X upgrade is parts; only about 2% of IBM’s charge is labour that might be performed by AMI. The stage was set for conflict. Copyright (C) 1987 Hesh Wiener