United States v. Microsoft Corp.

The court upheld the District Court’s market definition. Mac OS was not a reasonably interchangeable substitute because switching from Windows to a Macintosh system required consumers to buy different hardware and compatible applications, learn a new system, and bear file-conversion costs. Handheld devices, web portals, and other non-PC products likewise did not yet perform the full range of PC functions and therefore did not constrain Windows pricing in the foreseeable future.

Middleware, including Netscape Navigator and Java, was properly excluded from the operating-system market. Middleware could potentially weaken Windows’s role as the principal applications platform, but it did not then expose enough application programming interfaces, or APIs, to replace the operating system’s platform role. A nascent threat can matter to monopolization analysis without being a present substitute for market-definition purposes.

Microsoft’s share exceeded 95% in the relevant market. That share, combined with the applications barrier to entry, supported an inference of monopoly power. Consumers favored an operating system with a large existing library of applications, while developers favored an operating system with a large installed base of users. This feedback loop made it difficult for a new operating system to attract both developers and consumers.

Microsoft’s low Windows price and substantial research-and-development spending did not disprove monopoly power. A monopolist may invest in innovation, and a price below the short-run profit-maximizing level may still preserve long-run monopoly power. The court also noted evidence that Microsoft could set Windows prices without meaningful reference to rivals’ prices.

Microsoft’s OEM restrictions prevented computer manufacturers from removing IE icons, Start-menu entries, and other visible avenues for using IE. The court accepted the finding that OEMs often avoided preinstalling a rival browser alongside IE because multiple browser icons could confuse users and increase support costs. The restrictions therefore impeded a principal, low-cost channel for distributing Navigator and other rival browsers.

The prohibition on OEM modifications to the initial boot sequence also prevented OEMs from presenting consumers with internet-access offerings that promoted rival browsers. Other desktop restrictions likewise limited OEMs’ ability to feature non-Microsoft browsers or internet-access providers. These provisions reduced rival-browser usage not through better competition on the merits, but by limiting the choices OEMs could present to consumers.

Microsoft could not justify the restrictions merely by invoking copyright. Intellectual-property rights do not create immunity from antitrust law. Nor did Microsoft show that ordinary OEM desktop customization undermined Windows’s stability or consistency; the meaningful effect of the restrictions was to prevent promotion of browser rivals.

One restriction survived scrutiny: Microsoft could prohibit OEMs from automatically launching a substitute user interface that prevented users from seeing the Windows desktop. That was a substantial alteration of Microsoft’s copyrighted work, and the court concluded that the marginal anticompetitive effect did not outweigh Microsoft’s legitimate interest in preserving the Windows interface.

The court stressed that product-design claims require caution because antitrust law should not punish innovation simply because a dominant firm improves or changes its product. But product design is not categorically immune from § 2 scrutiny when it reduces competition without a valid procompetitive justification.

Microsoft’s removal of IE from Windows 98’s Add/Remove Programs utility discouraged OEMs from preinstalling a rival browser alone. Microsoft did not identify an integration-related benefit from denying users and OEMs that removal option. The court therefore treated the design change as exclusionary conduct that helped preserve the Windows monopoly.

The record supported the finding that Microsoft placed browser-specific code in files that also contained operating-system functions. Removing those files would disable Windows, making IE effectively irremovable and discouraging OEMs from installing rival browsers. Microsoft supplied no substantiated procompetitive justification for that commingling.

Microsoft did provide a technical explanation for overriding a user’s default-browser selection in a small number of Windows functions, including functions dependent on ActiveX controls or other Microsoft technologies. Because the plaintiffs did not rebut that explanation or prove that the anticompetitive harm outweighed the asserted benefits, this particular design choice could not support § 2 liability.

Microsoft could lawfully offer IE and the Internet Explorer Access Kit free of charge, or even at a negative price, absent a viable predatory-pricing theory. Likewise, merely developing the IE Access Kit was lawful product development. Antitrust law does not condemn a monopolist simply for offering an attractive product or price.

Microsoft’s exclusive arrangements with major internet access providers were unlawful because those providers were one of the two most important browser-distribution channels. The agreements secured preferred desktop placement for the providers while requiring them to promote IE, limit distribution of Navigator, or make IE the default browser. By covering most major access providers, the agreements significantly impeded Navigator’s distribution and protected Microsoft’s operating-system monopoly.

Microsoft’s First Wave agreements with major ISVs also had a substantial exclusionary effect. In exchange for valuable technical support and early access to Windows information, ISVs agreed to make IE the default browser for certain applications and to use Microsoft technologies. These agreements reached millions of users, reinforced Microsoft’s earlier foreclosure of primary distribution channels, and lacked a procompetitive justification.

Microsoft used its leverage over the Office suite to obtain an agreement requiring Apple to bundle IE, make IE the default browser, and limit promotion of Navigator on Mac computers. The agreement substantially restricted a cross-platform browser rival’s distribution, and Microsoft offered no legitimate competitive justification for the exclusivity.

The court reversed liability based on Microsoft’s arrangements with internet content providers. The District Court had expressly found insufficient evidence that the promotional restrictions in those agreements had a substantial harmful effect on Navigator’s usage share.

Microsoft did not violate § 2 simply by developing and promoting a Windows-optimized Java Virtual Machine that was incompatible with Sun’s JVM. Microsoft’s JVM allowed Java applications to run faster on Windows, and incompatibility alone does not establish an anticompetitive effect that outweighs a product’s procompetitive benefit.

Microsoft’s First Wave Java agreements with important ISVs were unlawful to the extent they made Microsoft’s JVM the practical default and restricted distribution of Sun-compliant JVMs. Those agreements foreclosed a substantial portion of JVM distribution and reduced Java’s ability to become a cross-platform middleware threat to Windows.

Microsoft also deceived Java developers by distributing tools containing Windows-specific keywords and directives while representing its Java implementation as cross-platform. Developers who relied on those representations could unwittingly create applications that ran only on Windows. Internal Microsoft documents showed that the goal was to fragment Java and prevent it from weakening Windows’s platform position.

Microsoft unlawfully pressured Intel to stop assisting Sun’s cross-platform Java efforts. Its demands were backed by threats that Microsoft would withhold support for Intel technologies or assist Intel rival AMD. This coercion impeded the development of a fast, cross-platform Java implementation and lacked a procompetitive explanation.

A court may consider conduct in its overall economic context rather than artificially isolating each act. But the District Court did not identify a distinct collection of individually minor acts whose combined effect independently harmed competition.

The District Court’s course-of-conduct discussion was largely conclusory and relied in part on Microsoft’s expenditures to promote IE. Spending aggressively to promote a product is not, standing alone, unlawful. Because the court identified no additional actionable conduct supporting an independent theory, there was no separate basis for liability.

The court rejected Microsoft’s claim that the plaintiffs had to reconstruct the hypothetical market that would have existed absent Microsoft’s conduct and prove that Navigator or Java would certainly have displaced Windows. Such a requirement would make antitrust enforcement especially difficult where a monopolist suppresses emerging technologies before their competitive potential can be tested.

Exclusion of nascent threats can materially preserve monopoly power, particularly in technologically dynamic markets. The District Court found that Navigator and Java had the potential to become cross-platform middleware and erode the applications barrier to entry. That potential supported an inference that Microsoft’s conduct helped maintain its monopoly.

The uncertainty about whether Netscape or Java would actually have succeeded mattered more to remedy than to liability. A structural remedy designed to eliminate a monopoly requires a clearer causal connection than an injunction directed at the proven exclusionary conduct.

Attempted monopolization requires anticompetitive conduct, specific intent to monopolize, and a dangerous probability of obtaining monopoly power. The dangerous-probability inquiry requires an independent analysis of the alleged browser market; Microsoft’s monopoly in operating systems could not substitute for proof that the browser market was susceptible to monopolization.

The District Court did not adequately define what counted as a browser or identify the relevant substitutes. Its references to a browser market, browsing technology, and platform-level browsing software were too imprecise to supply the market definition required by Spectrum Sports.

The plaintiffs also did not prove substantial barriers to entry into a properly defined browser market. General references to possible network effects and users’ reluctance to switch browsers were speculative and did not show that Microsoft could profitably exercise browser-market power. Because the plaintiffs had not developed this proof at trial or on appeal, the court reversed outright rather than remanding for a second attempt.

Traditional per se tying doctrine usually asks whether the tying and tied items are separate products, whether the defendant has power in the tying market, whether consumers were forced to take the tied product, and whether a substantial volume of commerce was foreclosed. The court concluded that this framework fit poorly when a software platform is integrated with complementary software functionality.

The usual separate-products inquiry relies on consumer demand for separate products and industry practice as proxies for the lack of efficiencies from bundling. Those backward-looking proxies can misfire when a firm integrates new functionality into a rapidly evolving platform product, because demand for a formerly standalone feature does not reveal the efficiencies of making it part of the platform.

Platform software may generate distinct efficiencies from integration. When browser APIs are distributed with an operating system, third-party developers can rely on their presence and need not distribute duplicate code. Uniform distribution can reduce delays, improve interoperability, and simplify software development. These possible benefits meant that the court could not presume that the integration had no redeeming virtue.

The court did not hold Microsoft’s conduct lawful. It held only that the rule of reason, rather than per se condemnation, must govern this tying claim. On remand, the plaintiffs could pursue theories of harm that did not depend on a new, precise browser-market definition or additional proof of independent entry barriers, because they had failed to establish those matters in litigating the attempted-monopolization claim.

The District Court acted within its broad discretion in setting an expedited schedule and using summary witnesses. Microsoft did not identify particular discovery it was unable to obtain, had agreed to the witness limits, and failed to overcome the presumption that a judge in a bench trial disregards inadmissible evidence.

The remedy phase was different. Microsoft made detailed offers of proof disputing the feasibility and likely effects of the proposed breakup, including evidence concerning prices, innovation, consumer welfare, and shareholder value. Predictions about future economic effects are factual matters, and the existence of disputed remedy-specific facts required an evidentiary hearing.

The District Court also did not adequately connect the structural breakup and conduct restrictions to the governing remedial objectives: ending the unlawful monopoly, denying the defendant the fruits of its violation, and preventing recurrence. Its brief explanation focused instead on Microsoft’s lack of contrition, perceived untrustworthiness, and the Government’s entitlement to a remedy of its choice.

The appellate decision eliminated the attempted-monopolization claim, remanded the tying claim under a new standard, and narrowed the monopoly-maintenance theory. The original remedy had been crafted for a broader set of violations, so the court could not assume that the same decree would be appropriate for the remaining liability findings.

On remand, the new district judge was directed to consider whether divestiture was justified. Structural breakup is especially difficult where a company has grown as a unitary enterprise rather than through mergers or acquisitions, and it requires a clearer causal connection between the unlawful conduct and the monopoly than is required for liability alone.

While the case was pending, the District Judge gave secret, embargoed interviews to reporters and later made public comments about Microsoft, Bill Gates, the evidence, legal theories, settlement, and the proper remedy. These statements violated the Code of Conduct’s prohibition on public comment about the merits of pending cases and raised concerns about ex parte communications.

The court held that an informed, reasonable observer would question the judge’s impartiality under 28 U.S.C. § 455(a). The secrecy of the interviews and the judge’s apparent willingness to provide colorful comments for eventual publication undermined public confidence in the integrity and neutrality of the judicial process.

The proper remedy was disqualification retroactive to the remedies stage. The court found no evidence of actual bias infecting the factual findings, reviewed the record carefully under the clearly erroneous standard, and declined to discard the entire liability proceeding. But the judge’s comments about Microsoft’s supposed intransigence and the breakup remedy created a particularly serious appearance of partiality at the remedies stage, requiring that a different judge handle remand.