Study the Past to Define the Future: United States v. Microsoft
By Pat Pascarella and Luke Hasskamp, Bona Law • Part II of a four-part series (August 2026)
In 1998, an Apple executive testified that Microsoft was attempting to “kill” Apple’s cross-platform playback software for Windows and replace it with Microsoft’s own multimedia technology. Microsoft’s leverage was its control of the dominant operating system. The district court credited Apple’s account and the D.C. Circuit condemned Microsoft’s use of operating system control to suppress cross-platform threats. Apple would go on to build the iPod, iTunes, and its ultimate siege machine, the iPhone.
Twenty-five years later, Apple occupies Microsoft’s former position. The government and private plaintiffs accuse Apple of using the same playbook its executive once described: contractual restrictions and technical controls that neutralize technologies capable of making the platform matter less. The theory of harm is the same, and Microsoft still supplies the analytic framework.
Part I of our series discusses how courts, not boardrooms, will decide who owns the customer’s first question in the AI era. In Part II, we discuss why that fight is not new.
Microsoft still supplies the framework
Microsoft is not a historical curiosity cited for color. It is the working framework beneath all three of the platform cases now in front of the courts. In United States v. Google LLC, 747 F. Supp. 3d 1 (D.D.C. 2024), the court applied Microsoft directly to Google’s search-distribution agreements. In United States v. Apple, Inc., No. 24-cv-4055 (JXN)(LDW), 2025 WL 1829127 (D.N.J. June 30, 2025), the government recasts Microsoft’s middleware theory for the smartphone. And in X Corp. v. Apple, Inc., No. 4:25-cv-00914-P (N.D. Tex. Nov. 13, 2025), the plaintiffs move the theory one layer higher, to the interface between an AI assistant and the operating system.
Understand Microsoft and you understand the doctrine driving the next decade of platform litigation.
What Microsoft actually held
Windows was protected by what the court called the applications barrier to entry: users preferred the operating system with the most applications, developers wrote for the system with the most users, and each side reinforced the other. Netscape’s browser and Sun’s Java threatened that loop. As middleware, they offered developers a cross-platform layer above Windows, reducing the importance of the underlying operating system. Microsoft held that a monopolist may not use exclusionary means to suppress such a nascent threat merely because no one can prove it would have matured into a full competitor.
Microsoft used both contractual and technical means. Restrictive licenses and exclusive or near-exclusive distribution agreements foreclosed Netscape from efficient routes to users. Microsoft also deployed a Windows-specific Java implementation and developer tools that impaired Java’s cross-platform promise. The combination—contractual restrictions and technical controls aimed at middleware that threatened the platform—is central to Microsoft’s exclusionary-conduct analysis.
Microsoft also supplied the burden-shifting framework employed in later cases. That is: (i) the plaintiff must show anticompetitive effect; (ii)the defendant may then offer a procompetitive justification; and (iii) the plaintiff may rebut or, if it stands, show that the anticompetitive harm outweighs the benefit. United States v. Microsoft Corp., 253 F.3d 34, 58–59 (D.C. Cir. 2001). Google adopted that framework, and Apple plaintiffs are likely to invoke it.
Apple on both sides of the same bargain
The same case contains a sharper irony. In 1997, with Apple in steep decline and 90 percent of Mac users dependent on Mac Office, Microsoft threatened to cancel the product unless Apple made browser concessions. Internally, a Microsoft executive said it could wield Office as “a club.” United States v. Microsoft Corp., 84 F. Supp. 2d 9, 94–97 (D.D.C. 1999) (findings of fact ¶¶ 344–356). Under the August 1997 agreement, Microsoft would continue Mac Office for five years; Apple would make Internet Explorer the Mac’s default browser and stop promoting Navigator. The district court found that Apple increased Internet Explorer distribution because of the threatened loss of Mac Office, and that Microsoft sought the default solely “to protect the applications barrier to entry.” Id. ¶ 355.
The D.C. Circuit did not treat that arrangement as ordinary dealmaking. It held the Apple deal to be anticompetitive exclusive dealing, because reducing the usage share of rival browsers served to protect Microsoft’s monopoly. Microsoft offered no procompetitive justification for it. Microsoft, 253 F.3d at 73–74. Apple thus appears on both sides of the case: through an executive describing Microsoft’s pressure campaign and as the counterparty coerced into a default later held anticompetitive.
Today the positions are reversed. Apple is paid billions each year to keep Google as Safari’s default, the kind of exclusive default arrangement condemned in Google. The company once coerced into a default now collects rent on one.
Seeing through the pattern
In Microsoft, every challenged act came with an engineering or business rationale: browser integration improved Windows; license terms protected the user experience. The Microsoft court did not accept those rationales at face value. It examined how each restriction operated and whether the asserted justification fit the conduct. Context allowed the court to distinguish product design from exclusion.
Today, Apple invokes security and privacy as recurring justifications—what the government describes as an elastic shield that expands or contracts with Apple’s business interests. At the pleading stage, the district court allowed the government to proceed on a course-of-conduct theory rather than assessing each restriction in isolation. That was not a merits ruling, but it preserved the central point: context matters.
Middleware then, AI now
Across eras, the same shape repeats. In 1998, the threats were the browser and Java. In the 2010s, they were cross-platform apps, cloud streaming, and super apps—technologies the government says Apple neutralized one after another. Today, it is the AI assistant. Each threatened to loosen the platform’s grip on the user, and each drew a response scaled to the threat.
AI assistants may be the most consequential threat yet because they compete for the user’s relationship with the device, not merely a feature. An assistant that works equally well across phones can reduce platform differentiation and switching costs, weakening the lock-in on which platform power depends. That is why emerging AI distribution arrangements should be analyzed through Microsoft rather than treated as wholly new.
What Microsoft did not decide
Microsoft had limits. The D.C. Circuit reversed the attempted-monopolization ruling because the plaintiffs had not adequately defined the browser market or established its barriers to entry. 253 F.3d at 80–84. It rejected per se treatment of the tying claim and remanded for rule-ofreason analysis. Id. at 84–97. It also vacated the breakup order and remanded for a new remedy proceeding, in part because of Judge Jackson’s public statements. Id. at 97–118. The eventual remedy was behavioral, not structural.
The limits carry a lesson. By the time relief arrived, Netscape had lost the browser war. A remedy imposed after a nascent threat is extinguished cannot restore the competition that might have developed. That is why the AI cases matter now, while the technology remains viable and before the interaction layer consolidates. The conduct they challenge is the Microsoft playbook, adapted for the AI era. Microsoft’s virtue was not a perfect remedy but an accurate diagnosis. The task now is to apply it sooner.
The pattern and the weapons
Microsoft shows that contractual restrictions and technical controls used to suppress middleware threatening a platform monopoly can constitute exclusionary conduct rather than competition on the merits. That is the pattern. In Part III we turn to the weapons: the causes of action private plaintiffs can use to challenge the walled garden.
Coming in September:
Part III: Storming the Castle: Conventional Weapons to Breach the Walled Garden
Part IV: The Trebuchet: AI