United States v. Microsoft

People and parties

Judges at different stages

Trial1998–2000

On remand2001–2002

Plaintiff

v.

Defendant

Select a participant for details.

The browser fight was about Windows

Netscape’s browser put more than Internet Explorer’s market share at stake. Along with Sun’s Java, Navigator offered developers a way to write software that depended less on Windows. Microsoft’s restrictions on those technologies’ distribution became the core of the monopoly-maintenance case. [1, pp. 53–54, 59–60]

The D.C. Circuit sustained monopoly-maintenance liability while rejecting several individual grounds for it. It also reversed attempted browser monopolization, vacated the tying ruling, and set aside the breakup order. The lasting lesson is to connect a specific practice to the competitive process, then connect the requested remedy to the wrong actually proved. [1, pp. 46, 58–59, 103–07]

Why a browser threatened Windows

Windows benefited from a reinforcing relationship: users wanted an operating system with many useful applications, and developers wanted to write for an operating system with many users. The trial court called the resulting obstacle to rivals the applications barrier to entry. The appellate court sustained the finding of monopoly power in the worldwide licensing market for Intel-compatible PC operating systems. [2, ¶¶ 30–44] [1, pp. 51–58]

Navigator and Java were middleware: software that exposed its own application programming interfaces, or APIs. Those interfaces gave developers functions they could use without writing everything directly for Windows. If enough applications ran through middleware available on several operating systems, users could have more freedom to choose the system underneath. This was a potential route around the entry barrier, whose eventual success remained uncertain. [2, ¶¶ 68–77, 411]

Distribution made that potential commercially important. Computer manufacturers, or original equipment manufacturers (OEMs), could preinstall browsers, and internet access providers could supply them with connection software. The trial findings identified these as particularly efficient paths to users. Limiting those paths could keep developers focused on Windows even when another browser remained available to download. [1, pp. 60–64, 67–71] [2, ¶ 145]

The case from 1998 to 2011

1. 
   May 18, 1998

   The United States and state plaintiffs sue Microsoft over Windows, Internet Explorer, and rival middleware. The cases are consolidated for trial. [1, pp. 47–48]

2. November 5, 1999

   Judge Jackson finds that Microsoft holds monopoly power in Intel-compatible PC operating systems and documents its conduct toward browser and middleware rivals. [2, ¶¶ 30–44, 68–77]

3. April–June 2000

   The trial court finds antitrust liability and orders Microsoft split into operating-system and applications businesses. The appeal later changes the liability rulings and vacates the breakup order. [1, pp. 48, 98–100]

4. June 28, 2001

   The D.C. Circuit upholds monopoly-maintenance liability in part, reverses attempted monopolization, remands tying, and vacates the remedy. It requires reassignment to a different judge. [1, pp. 46, 118–19]

5. September 6, 2001

   DOJ announces that it will no longer seek a breakup or pursue the tying claim on remand, aiming instead for prompt, effective relief. [3]

6. 
   November 12, 2002

   Judge Kollar-Kotelly enters the federal consent judgment, restricting specified Microsoft practices and imposing interoperability obligations with technical oversight. [4, §§ III–V]

7. June 30, 2004

   The D.C. Circuit affirms the litigating states’ remedial decree and upholds the public-interest determination supporting the federal consent decree. [5, slip pp. 3–5, 81–83]

8. May 12, 2011

   The final judgment’s remaining obligations expire after extensions of specified interoperability requirements. Some other provisions had already expired in 2007. [6]

When protecting a monopoly becomes unlawful

Did Microsoft preserve monopoly power through exclusionary conduct, rather than through the merits of its products? Section 2 required both monopoly power and unlawful acquisition or maintenance of that power. A large market share alone did not answer the second question. [1, pp. 50–51, 58] [7]

The court’s focus was conduct that would “harm the competitive process and thereby harm consumers.” Injury to a rival alone did not establish that harm. [1, p. 58]

Its framework placed the burdens in sequence:

1. The plaintiff must show an anticompetitive effect.
2. The defendant may offer a nonpretextual justification, such as improved efficiency or consumer appeal.
3. The plaintiff must rebut that justification or show that the competitive harm outweighs the benefit. Intent can help explain likely effects; it does not replace proof of them. [1, pp. 58–59]

The government’s argument

Microsoft used Windows licensing, distribution agreements, and technical restrictions to keep rival middleware from attracting users and developers. That weakened a route by which competition in operating systems might emerge. [1, pp. 59–78]

Microsoft’s argument

Microsoft argued that rapid technological change constrained Windows and that integration and licensing protected a useful, consistent product. Microsoft also argued that plaintiffs had not proved its conduct caused the continuation of its monopoly. The court rejected a blanket copyright defense, but accepted particular justifications and reversed liability where proof was insufficient. [1, pp. 51–58, 63–67, 79–80]

The conduct that made the difference

Control over distribution

Restrictions on computer manufacturers’ ability to remove visible access to Internet Explorer or promote competing browsers discouraged rival preinstallation. Manufacturers faced support costs when multiple browsers confused users. Microsoft’s answer that Netscape still had other distribution routes did not resolve the loss of the most efficient ones. Most challenged OEM restrictions were unlawful; the restriction on automatically replacing the Windows desktop survived. [1, pp. 60–64]

Certain agreements with internet access providers and Apple likewise limited competing browsers’ reach without an adequate justification. Giving Internet Explorer away, however, was not itself a basis for liability: the government did not pursue predatory pricing on appeal. [1, pp. 67–75]

Apple’s role shows why distribution outside Windows mattered. Microsoft threatened to stop developing Mac Office, then secured an agreement making Internet Explorer the Mac’s default browser and limiting Apple’s promotion of rivals. The court reasoned that Navigator’s reach across operating systems could attract developers to its APIs. Restricting its use on Macs could therefore help protect the Windows monopoly. [1, pp. 73–74]

What the code did—and why

The court sustained liability for excluding Internet Explorer from Windows 98’s Add/Remove Programs utility and commingling browser and operating-system code. The record connected those choices to reduced rival distribution; Microsoft did not substantiate an integrative benefit for those particular practices. [1, pp. 64–67]

A different challenged feature survived. Windows invoked Internet Explorer for certain functions despite a user’s alternative default browser. Microsoft supplied a technical explanation that plaintiffs failed to rebut. The analysis turned on the specific feature and the evidence offered for it. [1, p. 67]

Java: performance, exclusivity, and deception

Microsoft’s faster, incompatible Java virtual machine was not itself unlawful. The court reached a different result for agreements making Microsoft’s implementation effectively exclusive, misleading developers about whether their applications would run beyond Windows, and pressure on Intel to stop supporting cross-platform Java. The distinction separated product improvement from practices that obstructed a competing platform. [1, pp. 74–78]

What survived the appeal

The seven participating judges spoke unanimously in a per curiam opinion, issued in the court’s name. There was no separate concurrence or dissent. The disposition had four distinct parts: [1, caption and pp. 46, 118–19] [3]

• Monopoly maintenance: affirmed in part and reversed in part. The § 2 violation survived on the exclusionary practices the court upheld.
• Attempted browser monopolization: reversed without remand. Plaintiffs had not adequately established the relevant browser market and entry barriers needed to show a dangerous probability of monopolization.
• Tying Windows to Internet Explorer: the per se ruling was vacated. The court required analysis under § 1’s rule of reason, weighing competitive harms and benefits for this platform-software arrangement. That was not a ruling that all software bundling is lawful.
• Remedy: vacated in full, with further proceedings before a different judge. [1, pp. 78–97, 101–07, 118–19]

Proving harm before a rival matures

For this government action seeking equitable relief, the court did not demand proof that Navigator or Java certainly would have displaced Windows. It allowed a causal inference from exclusionary conduct reasonably capable of significantly contributing to the continued monopoly. The uncertainty surrounding an emerging competitor did not insulate its exclusion from scrutiny. [1, pp. 79–80]

That inference did not settle the remedy. A breakup required a stronger connection between the unlawful conduct and the monopoly power the proposed restructuring would address. [1, pp. 105–07]

Why Microsoft was not broken up

The original remedy rested on liability findings that the appeal had narrowed. The trial court also failed to hold an evidentiary hearing to resolve disputed remedy facts and did not adequately explain the relief it selected. Each problem mattered to the decision to vacate the order. [1, pp. 101–07]

Judge Jackson’s contacts with the press and public comments created an appearance of partiality, requiring disqualification and reassignment. The appellate court found no evidence of actual bias and did not erase the factual findings; it continued to review them for clear error. [1, pp. 107–18]

DOJ then elected to seek conduct relief. The 2002 federal decree protected specified manufacturer choices, restricted retaliation and exclusive arrangements, required defined API disclosures and communications-protocol licensing, and established a three-person technical committee. The decree’s provisions had definitions and exceptions; it was not a general command to disclose all Windows source code. [3] [4, §§ III–IV]

One remedial detail matters: § III.H addressed access to middleware through icons, shortcuts, and default settings. It did not require Microsoft to extract every browser component from Windows. In 2004, the D.C. Circuit upheld the choice to reduce the support costs that discouraged rival preinstallation by allowing access to be removed, while avoiding compulsory redesign of the operating system. [4, § III.H] [5, slip pp. 11–19]

The decree’s remaining obligations expired in 2011. [6] The appellate opinion continued to supply a framework for later cases, including the Google search litigation discussed below. [9, printed pp. 134–35]

Using the case in practice

Practice analysis. For counsel, the useful discipline is to keep the theory of harm and the theory of relief separate:

• Identify the market being protected. Browser restrictions mattered because of their effect on operating-system competition.
• Explain the path to competitive harm. Connect distribution, user adoption, developer incentives, and entry barriers with evidence.
• Test each justification. A general claim of innovation does not explain every restrictive design or contract term.
• Build the remedy record. Evidence sufficient for liability may leave important questions about structural relief unanswered. [1, pp. 59–80, 101–07]

Change one fact

Hypothetical: suppose Microsoft had substantiated an efficiency benefit for a challenged integration choice. Under the court’s framework, the plaintiffs would then have needed to rebut that justification or show that the harm outweighed the benefit. That change would affect the required proof, without predetermining the result. [1, pp. 58–59]

Later applications and limits

Novell v. Microsoft · Tenth Circuit · 2013

The Tenth Circuit cited Microsoft but applied the narrower rules governing a refusal to help a rival. Novell’s claim failed for lack of evidence satisfying that framework, including the required profit sacrifice on the facts presented. A general exclusion theory does not by itself establish a duty to share technology. [8, slip pp. 16–28]

United States v. Google · D.D.C. · 2024

The Google search liability opinion expressly adopted Microsoft’s framework for monopoly power, competitive harm, and justifications. It provides a modern district-court application of the same analytical structure to platform distribution agreements. [9, printed pp. 134–35, 203–04]

United States v. Google · D.D.C. · 2025

The remedies opinion returned to Microsoft’s distinction between establishing liability and justifying relief. It examined the strength of the causal connection needed for different remedies and rejected a categorical equation of that requirement with but-for causation. [10, printed pp. 63–73]

Selected later treatment checked September 27, 2026. These decisions operate within their respective jurisdictions. The Google case index lists a December 2025 final judgment, a July 2026 appellate response and cross-appeal brief, and September 2026 compliance filings. This is a targeted source review, not comprehensive citator clearance or a complete account of appellate status. [11]

Sources and citations

1. United States v. Microsoft Corp., 253 F.3d 34

   D.C. Circuit · June 28, 2001 · En banc, per curiam. Inline page references use the Federal Reporter pagination.

2. Findings of fact, 84 F. Supp. 2d 9

   D.D.C. · November 5, 1999 · Especially ¶¶ 30–44, 68–77, 145, 411.

3. DOJ’s plans for proceedings on remand

   Department of Justice · September 6, 2001 · Announcement concerning breakup and tying.

4. Final judgment, No. 98-1232 (CKK)

   D.D.C. · November 12, 2002 · Entered federal consent judgment, §§ III–V · PDF.

5. Massachusetts v. Microsoft Corp., 373 F.3d 1199

   D.C. Circuit · June 30, 2004 · Slip pp. 3–5, 11–19, 81–83 · PDF.

6. Microsoft antitrust final judgment expires May 12

   Department of Justice · May 11, 2011 · Expiration and extensions.

7. Sherman Act, 15 U.S.C. § 2

   U.S. Code · Cornell Legal Information Institute · Current statutory text; this case was a civil enforcement action. Related provision: 15 U.S.C. § 1 (restraints of trade and tying).

8. Novell, Inc. v. Microsoft Corp., 731 F.3d 1064

   Tenth Circuit · September 23, 2013 · Slip pp. 16–28 · Court opinion hosted by Applied Antitrust Law.

9. United States v. Google LLC — liability opinion

   D.D.C. · August 5, 2024 · ECF 1033, printed pp. 134–35, 203–04 (PDF pp. 138–39, 207–08).

10. United States v. Google LLC — remedies opinion

    D.D.C. · September 2, 2025 · ECF 1436, printed pp. 63–73 (PDF pp. 67–77).

11. Google search case — subsequent filings

    Department of Justice case index · Reviewed September 27, 2026.

Participant images are AI-generated illustrations. Visual references: the U.S. flag; Microsoft’s 1987–2012 wordmark; Beverly Rezneck’s court portrait of Thomas Penfield Jackson; and the court photograph of Colleen Kollar-Kotelly, all via Wikimedia Commons. The filing folder and network connector are generated illustrations of the timeline events.