DOJ v. Microsoft: winning the case, losing the fix

In May 1998, the U.S. Department of Justice filed one of the most meticulously documented antitrust complaints in American legal history against a company whose operating-system software ran on more than 90 percent of the world's personal computers. 1 Three years later, the judge who had ordered that company split in two was removed from the case for talking too much to journalists. The DOJ walked away with a behavioral consent decree that nine states considered too weak to sign. Microsoft remained intact, posted revenue of $950 billion over the following decade and a half, and Internet Explorer hit 95 percent browser market share two years after the settlement. 2

The arc of United States v. Microsoft Corporation is a precise case study in the gap between legal victory on substance and strategic failure on remedy. The DOJ built an airtight liability case. It then lost the remedy through a sequence of judge-management failures, administration changes, and negotiation miscalculations that practitioners have been dissecting for twenty-five years.

The positions each side actually held

Before the first brief was filed, the parties' objectives, leverage, and constraints were as asymmetric as the companies themselves.

Dimension	DOJ (and 20 states)	Microsoft
Primary objective	Break up or behaviorally constrain Windows monopoly; open browser/middleware market	Preserve integrated product; avoid divestiture at all costs
Secondary objective	Establish legal precedent for antitrust in tech markets	Delay resolution until market conditions changed
Best alternative (BATNA)	Win breakup order on appeal	Win full reversal at D.C. Circuit
Worst-case scenario	Behavioral decree too weak to enforce	Structural split into two separate companies
Core leverage	95%+ OS market share documented in internal Microsoft emails	Complexity of tech markets; pace of innovation as complicating factor
Core vulnerability	Trial judge prone to extrajudicial commentary	Internal documents showing predatory intent in Gates's own words
Time pressure	Political: Clinton DOJ needed resolution before 2001	Financial: litigation distracted leadership; stock fell ~40% during proceedings 3

The asymmetry that ultimately decided the case was not about market power — Microsoft had that in abundance — but about institutional behavior. The DOJ had the facts. Microsoft had the clock.

Act 1: The threat nobody took seriously (1994–1998)

In May 1995, a 27-year-old programmer named Marc Andreessen was running Netscape Communications, which had just completed the most successful technology IPO in American history. His Navigator browser was installed on 90 percent of internet-connected computers. It was, in the language of the Sherman Act, becoming a genuine alternative platform — a layer of software that developers could build on regardless of which operating system sat beneath it. 1

Bill Gates recognized the structural threat before almost anyone else inside Microsoft. In a May 1995 memo titled "The Internet Tidal Wave," he wrote: "Now I assign the Internet the highest level of importance. The Internet is a tidal wave. It changes the rules." 4 That same month, Microsoft representatives met with Netscape and proposed dividing the browser market — Microsoft would supply browsers on Windows; Netscape would supply browsers on other operating systems. Netscape's team walked out. 1

What followed was systematic. Microsoft began distributing Internet Explorer at zero cost — Bill Gates was explicit about the calculation: "Our business model works even if all Internet software is free... We are still selling operating systems. What does Netscape's business model look like? Not very good." 1 Group Vice President Paul Maritz was blunter: "We are going to cut off their air supply. Everything they're selling, we're going to give away for free." 1

The mechanism was contracts. Microsoft required original equipment manufacturers (OEMs) to ship Internet Explorer as a condition of their Windows license, restricted their ability to modify the Windows boot sequence or desktop, and signed exclusive distribution agreements with internet access providers — paying some of them to carry IE while prohibiting them from promoting Netscape. Netscape's share of browser installations fell from roughly 90 percent in early 1996 to approximately 50 percent by late 1997. In Q4 1997 Netscape posted a loss exceeding $88 million. 4 AOL acquired the company for $4.2 billion in November 1998 — seven months after the DOJ complaint was filed. 5

Act 2: Building the liability case (1998–1999)

On May 18, 1998, the DOJ, led by Assistant Attorney General Joel Klein with trial counsel David Boies, filed its complaint in the U.S. District Court for the District of Columbia (Civil Action No. 98-1232). 1 The document alleged four violations of the Sherman Act: unlawful maintenance of a monopoly in the market for Intel-compatible PC operating systems (§2); attempted monopolization of the browser market (§2); illegal tying of Internet Explorer to Windows (§1); and unlawful exclusive dealing (§1). Twenty states and the District of Columbia co-filed parallel claims.

The bench trial opened on October 19, 1998, before Judge Thomas Penfield Jackson. Over 76 days of testimony — running through June 24, 1999 — both sides called twelve witnesses each, with direct testimony submitted in writing. 6 Microsoft's most self-inflicted damage came from its own evidence. Gates appeared via videotaped deposition and performed so evasively — repeatedly claiming not to recall decisions documented in his own emails — that Judge Jackson reportedly laughed aloud during screening. 7

On November 5, 1999, Judge Jackson issued his Findings of Fact (84 F.Supp.2d 9). In 412 paragraphs, he concluded that Microsoft held a monopoly in the relevant market — Intel-compatible PC operating systems — where its share had exceeded 90 percent for nearly a decade and exceeded 95 percent in the two years preceding the findings. 6 The court identified the "applications barrier to entry" as the structural moat protecting that share: developers write applications for platforms with large user bases; users adopt platforms with large application libraries; no new entrant can break into either side of that loop without already being on the other side.

"Microsoft enjoys so much power in the market for Intel-compatible PC operating systems that if it wished to exercise this power solely in terms of price, it could charge a price for Windows substantially above that which could be charged in a competitive market. Moreover, it could do so for a significant period of time without losing an unacceptable amount of business to competitors."
— Judge Thomas Penfield Jackson, Findings of Fact ¶33 6

Between November 1999 and April 2000, U.S. Court of Appeals Chief Judge Richard A. Posner was brought in as mediator. Talks broke down after nearly four months. 8

On April 3, 2000, Judge Jackson issued his Conclusions of Law (87 F.Supp.2d 30). The rulings: §2 monopoly maintenance — found; §2 attempted monopolization of the browser market — found; §1 tying of IE to Windows — found; §1 exclusive dealing — not found (insufficient evidence). 8 Microsoft's copyright defense — that its intellectual property rights entitled it to bundle products as it chose — was rejected: "Even constitutional privileges confer no immunity when they are abused for anticompetitive purposes." 8

The liability case was nearly airtight. What Jackson did next put the entire remedy at risk.

Act 3: The breakup order and its flawed foundation (June 2000)

On June 7, 2000, Judge Jackson issued his Final Judgment (97 F.Supp.2d 59), ordering Microsoft split into two independent companies: an Operating Systems Business and an Applications Business. The split was to be completed within 16 months of the order becoming effective. 9 The order also imposed interim behavioral constraints: non-discriminatory OEM licensing, API disclosure, and a ban on retaliatory contracts.

Jackson refused Microsoft's request for additional evidentiary hearings on the remedy. His explanation was explicit about his reasoning: "A structural remedy has become imperative: Microsoft as it is presently organized and led is unwilling to accept the notion that it broke the law or accede to an order amending its conduct." 9 He added that Microsoft "has proved untrustworthy in the past." 9

Jackson was right on the facts and wrong on process. Skipping an evidentiary hearing on a remedy of this scale — in a technology market that had already shifted substantially between 1998 and 2000 — created a procedural gap that the appeals court would use to nullify the entire order. Jackson also gave journalists something they would use against him: access.

Between the Findings of Fact and the Final Judgment, Jackson gave Ken Auletta of The New Yorker more than ten hours of on-record interviews for what would become the book World War 3.0. He described Gates as possessing "a Napoleonic concept of himself and his company, an arrogance that derives from power and unalloyed success, with no leavening hard experience, no reverses." 10 He compared Microsoft's executives to street gangs. He made these comments while the case was on appeal — a violation of the judicial conduct rules that govern ex parte communication.

Act 4: The D.C. Circuit dismantles the remedy (June 2001)

The U.S. Supreme Court declined to hear the case directly, returning it to the D.C. Circuit Court of Appeals. On June 28, 2001, a seven-judge en banc panel issued a per curiam opinion (253 F.3d 34) that simultaneously affirmed the core of the liability case and demolished the remedy.

Affirmed: Microsoft's §2 monopoly maintenance. The court reviewed each category of anticompetitive conduct — OEM licensing restrictions, technical bundling of IE, exclusive contracts with internet access providers, arrangements with ISVs and Apple, and suppression of Java — and found that each category had anticompetitive effects that Microsoft could not justify with procompetitive rationale. 11

Reversed: The §2 attempted monopolization of the browser market. The government had not proved that Microsoft had a "dangerous probability of success" in achieving a browser monopoly — a required element that the district court had skipped. 11

Remanded: The §1 tying claim. The court ruled that per se illegality — the shortcut that treats certain practices as automatically unlawful — did not apply to platform software where technical integration might produce genuine consumer benefits. Rule of reason analysis was required instead. This ruling became one of the most-cited passages in subsequent tech antitrust cases, including the 2024 Google search case. 11

Vacated: The entire remedy order. Three grounds: the order rested partly on a claim (attempted browser monopolization) that had been reversed; Jackson had not held an evidentiary hearing on the remedy in a fast-moving technical market; and Jackson's out-of-court commentary constituted an "appearance of partiality" that, while not proving actual bias, "seriously tainted the proceedings." 11 The court noted pointedly: "Six years seems like an eternity in the computer industry. By the time a court can assess liability, firms, products, and the marketplace are likely to have changed dramatically." 11 Jackson was removed. The case was assigned to Judge Colleen Kollar-Kotelly.

Act 5: Settlement — the behavioral decree and its limits (2001–2011)

The political context had shifted. In January 2001, George W. Bush took office. On September 6, 2001 — five days before the attacks that would redirect the entire U.S. government's attention — the new DOJ Antitrust Division, now under Charles James, announced it would no longer seek to break up Microsoft. 12

Negotiations ran through the fall of 2001 under mediator Eric Green. On November 2, 2001, Microsoft and the DOJ signed a consent decree. 13 Bill Gates called it "fair and reasonable" and acknowledged it contained "some very tough rules and restrictions." 13 Judge Kollar-Kotelly approved it under Tunney Act procedures on November 1, 2002. 14

The decree's eight core provisions required Microsoft to: allow OEMs to install competing middleware without retaliation (§III.A); offer uniform Windows licensing terms to its twenty largest OEMs (§III.B); permit OEMs to configure desktops, boot sequences, and default programs freely (§III.C); disclose through MSDN the APIs its own middleware used to communicate with Windows (§III.D); share the server communication protocols needed for third-party products to interoperate with Windows networks (§III.E); refrain from retaliating against software or hardware vendors supporting competitors (§III.F); ban exclusive distribution deals (§III.G); and allow end users and OEMs to remove Microsoft middleware from the visible interface (§III.H). 15 A three-member independent Technical Committee held complete access to Microsoft's systems, records, and source code for five years.

Nine states and the District of Columbia refused to sign, seeking stronger remedies. They lost their separate appeal at the D.C. Circuit in June 2004. 5

What the decree produced — and failed to produce — is the subject of twenty years of debate. On the negative side: the communication-protocol documentation Microsoft was required to deliver under §III.E was still not complete five years later. In January 2008, Judge Kollar-Kotelly extended the entire decree to November 2009 specifically because of that failure. 16 She was careful to note the extension "should not be viewed as a sanction against Microsoft" — a formulation that former FTC Commissioner Dennis Yao described as underscoring the practical problem with behavioral remedies: they require sustained, complex monitoring that courts are not built to provide. 17

On the positive side, the European Union's parallel proceeding — which imposed tougher interoperability requirements than the U.S. decree — produced one clear success. The open-source Samba project used the EU-mandated server protocol disclosures to build software that let non-Windows computers participate fully in Windows networks, a concrete competitive outcome that the American settlement never matched in scope. 2 The EU fined Microsoft a total of more than €1.6 billion across three separate rulings between 2004 and 2012. 18

Steven Sinofsky, who led Windows teams during the consent-decree years and was directly responsible for compliance, later wrote that the most lasting damage was not operational but reputational: every employee was required, in social settings, to explain and defend the company's conduct. "Those side effects of litigation," he wrote, "were more difficult than the specific structural and regulatory remedies." 19 David Schmittlein, then Wharton's vice dean, offered the most economical version of the same observation: "Top management time and attention is a very precious commodity, and the litigation has been draining for Microsoft in this respect." 17

The consent decree expired in 2011. Brad Smith, who became Microsoft's General Counsel in 2002 and made improving government relations his primary mandate, gave a retrospective lecture at Harvard's Berkman Center in 2008 that remains the clearest statement of what the company learned: "Lessons that are learned but not applied, that don't really lead to any change, are of some academic interest but little practical import." He also offered the line that has since circulated widely: "Using the political discourse of the day, if you're from Microsoft, there's no way to put any lipstick on that trial." 20

Frameworks you can use

The case distills into four frameworks that translate directly from antitrust litigation into boardroom and deal-room practice.

Framework 1: Internal documents as the most dangerous counterparty

Every negotiation and dispute carries discovery risk — the possibility that internal communications, if subpoenaed or leaked, will directly contradict the public position being argued. The Microsoft case is the definitive illustration of what happens when that risk is ignored at the highest level.

Paul Maritz's "cut off their air supply" email, Bill Gates's "business model works even if all Internet software is free" observation, and James Allchin's note that Windows was "the one thing they don't have" were all written by senior executives discussing competitive strategy in plain language. 1 They were not written for litigation; they were written to inform internal decision-making. The DOJ's complaint cited them as direct evidence of predatory intent, and Judge Jackson's Findings of Fact built the monopoly-maintenance conclusion on top of them.

The manager's takeaway is not that internal communication should be sanitized — that creates its own legal risks and destroys decision-making quality. It is that strategy documents should be written to withstand a future reading in which the author's interests are adverse to those of the reader. If a competitive strategy can only be articulated in language that would look predatory in a courtroom, that is a signal about the strategy itself, not just about the language.

Framework 2: The judge-management problem — every dispute has a process owner whose conduct can overturn the outcome

Judge Jackson's media engagement cost the DOJ its entire remedy. The Findings of Fact and Conclusions of Law were, by the D.C. Circuit's own account, largely correct. The breakup order — the strategic outcome the DOJ had been building toward for three years — was vacated not because it was wrong on the merits but because the official who issued it behaved in a way that created an appearance of partiality.

This pattern generalizes far beyond antitrust. Any complex negotiation, dispute, or regulatory process runs through individuals whose authority to decide the outcome is inseparable from their credibility to do so without bias. When those individuals compromise that credibility — through off-the-record commentary, public posturing, or conduct that gives the losing party grounds to challenge the process — the substantive outcome can be reversed regardless of its merits.

The practical implication for deal-makers: when you are the party that benefits from a process running cleanly to conclusion, proactively protect that process. The party with the weaker substantive case has a strong incentive to find procedural grounds for reversal. Don't hand them the grounds.

Framework 3: BATNA asymmetry across political cycles

David Beier, who served as a senior domestic policy advisor in the Clinton White House, put the government's position directly: "They never really expected to secure a breakup remedy, and they never tried an attempted monopoly case. So, in point of fact, the government won the maintenance case." 17 This was true as a legal matter. It was false as a strategic matter, because the DOJ's ability to convert that legal win into a meaningful remedy depended on political continuity that did not survive the 2000 election.

The Bush administration DOJ's decision to abandon the breakup in September 2001 — and to settle quickly — reflects something that every negotiator in a multi-year, government-involved transaction needs to model explicitly: the other side's BATNA shifts with every election, every leadership change, and every external event that redirects attention. The settlement Microsoft obtained in November 2001 was substantially better than anything the Clinton DOJ would have accepted in 2000. What changed was not the facts of the case. What changed was the person with signing authority.

The manager's takeaway: when your transaction involves a government counterparty, build your timeline analysis around political cycles, not just legal schedules. A position that is untenable today may be entirely acceptable after the next election. A concession that looks generous today may look like the floor after a regulatory transition. Model both sides' BATNAs under multiple political scenarios, not just the current one.

Framework 4: Structural vs. behavioral remedies — the enforcement cost problem

The DOJ's original preference for structural divestiture — splitting Microsoft in two — was not arbitrary. Former FTC Commissioner Dennis Yao explained the logic: behavioral remedies require ongoing monitoring, ongoing enforcement, and ongoing technical judgment that courts and regulators are structurally ill-equipped to provide. The communications protocol documentation that was still not complete five years after Judge Kollar-Kotelly approved the decree is a clean example of what Yao was describing. 17

Structural remedies — split the company, sell the division, divest the asset — are self-executing. Once completed, they require no continued supervision. They are also politically harder to impose and legally harder to survive on appeal. Behavioral remedies are politically easier to accept and harder to enforce in practice.

The trade-off maps directly to corporate settlements and consent agreements outside antitrust. When a negotiation produces a behavioral commitment rather than a structural change — "we agree to do X going forward" rather than "we will sell or separate Y" — the value of that commitment depends entirely on the enforcement mechanism and the incentive structure of the monitoring body. In the Microsoft consent decree, the three-member Technical Committee had full access to Microsoft's systems but no power to impose sanctions; it could only report to the court. The court, in turn, could hold Microsoft in contempt — an expensive, slow, and politically costly process that was never invoked. The compliance teeth were always softer than the compliance language.

For managers negotiating any settlement or consent agreement, the central question is not "what does the other side agree to?" but "what happens to them if they don't do it?" If the answer is "they report a violation to a committee, which reports to a judge, who may hold a hearing" — that is a different agreement than one where the answer is "they forfeit a $500 million escrow and we get the rights back."

What to remember

Internal documents are permanent testimony. Every competitive strategy memo, pricing analysis, and internal forecast written by a senior executive is a potential exhibit in a future proceeding. The standard for what you write should be: how does this read if I'm on the wrong side of it in five years?
Process integrity determines remedy survival. The DOJ proved its liability case and still lost its remedy, because the judge who issued it compromised the proceedings by talking to reporters. In any dispute with a formal adjudicator, protecting the integrity of that adjudicator's process is a core strategic interest for the party that benefits from the outcome.
Model your counterparty's BATNA across political cycles, not just legal ones. The settlement Microsoft obtained in November 2001 was structurally better than anything available in 2000 — the underlying facts had not changed, but the political ownership of the case had. Government-involved transactions should always include a scenario in which the regulator's principals change before the deal closes.
Behavioral remedies are worth approximately what their enforcement mechanism is worth. The Microsoft consent decree contained detailed, specific obligations; the documentation underlying one of its key provisions was still incomplete five years after approval. Before accepting behavioral commitments in any settlement, map the enforcement chain from obligation to consequence — and be honest about how many steps that chain has and who controls each one.

Cover image generated for editorial use.