Patents

Microsoft v. AT&T: When Software Crosses the Border but the Patent Statute Does Not

The Supreme Court holds that supplying a master disk of software from the United States, then copying it abroad, does not 'supply' the patented invention's components under §271(f).

February 11, 2025
A master software disk and circuit board on a workbench
Microsoft v. AT&T held that intangible software code, and foreign-made copies of it, fall outside §271(f). Shutterstock
Educational content, not legal advice. This article explains general legal concepts. It does not create an attorney–client relationship. For your specific situation, consult a licensed attorney.

In Microsoft Corp. v. AT&T Corp., No. 05-1056 (U.S. Apr. 30, 2007), the Supreme Court confronted a deceptively narrow question with sweeping consequences for the global software industry: when a U.S. company sends a single master disk of patented software abroad, where foreign workers copy it onto computers assembled and sold outside the United States, has the company “supplie[d]” the “components” of the patented invention “from the United States” within the meaning of 35 U.S.C. §271(f)? Argued February 21, 2007, and decided April 30, 2007, the case produced a 7-1 decision reversing the Federal Circuit. Justice Ginsburg wrote for the Court, joined by Justices Scalia, Kennedy, and Souter; Justice Alito concurred in part and in the judgment, joined by Justices Thomas and Breyer; Justice Stevens dissented alone. Chief Justice Roberts took no part in the case.

At a glance

  • Case: Microsoft Corp. v. AT&T Corp., No. 05-1056
  • Court: Supreme Court of the United States
  • Argued / Decided: February 21, 2007 / April 30, 2007
  • Vote: 7-1 (reverse and remand)
  • Majority: Ginsburg, J., joined by Scalia, Kennedy, and Souter, JJ.
  • Separate opinions: Alito, J., concurring in part and in the judgment, joined by Thomas and Breyer, JJ.; Stevens, J., dissenting
  • Not participating: Roberts, C.J.
  • Holding: Software in the abstract, untethered from a physical medium, is not a “component” under §271(f); and even when embodied on a disk, the components actually combined into the foreign-assembled computers were copies made abroad, not the article “supplie[d] … from the United States.”

AT&T held a patent on a computer programmed to encode and compress recorded speech. Microsoft’s Windows operating system, when installed, enabled a computer to perform the patented speech-processing functions. Microsoft conceded domestic infringement for U.S.-installed copies. The dispute was about computers built and sold abroad. Microsoft shipped a master version of Windows—on a disk or by electronic transmission—to foreign manufacturers, who then made copies and installed them on machines sold to foreign customers. AT&T argued that Microsoft thereby supplied components of the patented invention from the United States for combination abroad, triggering §271(f). The Federal Circuit agreed. The Supreme Court did not.

Section 271(f) and the Deepsouth gap

Section 271(f) was Congress’s answer to Deepsouth Packing Co. v. Laitram Corp., 406 U.S. 518 (1972), which had held that exporting the unassembled parts of a patented machine for assembly abroad escaped liability because no one “ma[de]” the patented combination inside the United States. Congress closed that gap in 1984 by making it an act of infringement to “suppl[y] … from the United States” the “components of a patented invention” for “combination” abroad in a manner that would infringe if it occurred domestically. The statute is, by its terms, an express and deliberate extension of U.S. patent law to a defined slice of foreign conduct—but a slice, the Court stressed, that Congress drew with care.

The interpretive task in Microsoft was to map a 1984 statute written with tangible machine parts in mind onto the realities of mass-replicated software. Two questions had to be answered in sequence: first, is software a “component” at all; and second, even if so, was the relevant component “supplie[d] … from the United States”?

Software as an “abstract idea” versus a “component”

The Court drew a sharp line between software in the abstract and software embodied in a usable medium. Windows in the abstract—code conceived as a set of instructions, like the notes of a musical composition or the blueprint of a machine—is not a “component” capable of being “combined” into a computer. A component, the majority reasoned, is a physical, combinable part. Until the instructions are encoded onto a medium from which a computer can read and install them, there is nothing to “combine.” Justice Ginsburg likened abstract code to a detailed set of plans: a blueprint may tell you exactly how to build a patented device, but the blueprint is not itself a component of it.

That framing did real work. It meant the electronic transmission of code, and the design information on a master disk, could not by themselves be the supplied “components.” Only a copy installed on a particular machine is a usable, combinable component of that machine. The locus of the analysis therefore shifted to those installed copies—and to where they were made.

”Supplied from the United States”: the copies were made abroad

Here the majority delivered the decisive move. The computers sold abroad were not loaded with the master sent from Redmond; they were loaded with copies generated overseas. Section 271(f) attaches liability to components “supplie[d] … from the United States,” and the components actually combined into the foreign machines—the installed copies—were made outside the country. The master disk was never installed on any of the foreign-sold computers; it was a template. Copying, the Court held, is not “supplying,” even when the copying is foreseeable, easy, and the very purpose of sending the master.

AT&T urged that for software, copying is so trivial that the distinction between a master and its copies is an exaltation of form over substance. The majority answered that the distinction is exactly the kind of line Congress drew, and that any anomaly was Congress’s to fix, not the Court’s. The Court reinforced that conclusion with the presumption against extraterritoriality: §271(f) is an exception to the general rule that U.S. patent law stops at the water’s edge, and exceptions to that rule are read narrowly, with foreign conduct left to foreign law and any expansion left to Congress. Where the statute was ambiguous as applied to a technology Congress did not anticipate, the tie went to the territorial default.

Justice Alito’s concurrence would have gone further, reasoning that no physical object that traveled from the United States ever became part of the foreign computers—because even a disk is removed after installation, leaving only locally generated magnetic or electronic states inside the machine—so the “supplied from the United States” requirement failed regardless of the abstract-versus-embodied debate. Justice Stevens dissented, arguing that software is functionally the active ingredient of the patented computer and should count as a supplied component, with the master serving as the effective source of every copy.

Open questions

  • Does the “copies made abroad” rule survive new distribution models? Cloud delivery, streaming installation, and over-the-air updates blur the master/copy line the Court relied on; whether a server-side U.S. origin changes the analysis is unresolved.
  • What is the boundary between an abstract “design” and an embodied “component” for other intangibles? Firmware, encrypted keys, and AI model weights all sit uneasily on the line Ginsburg drew.
  • How far does the extraterritoriality presumption push §271(f) generally? Microsoft invoked the presumption to resolve ambiguity, foreshadowing the narrowing the Court later applied to component-counting in Life Technologies v. Promega.
  • Has Congress responded? The decision openly invited a legislative fix for software, but §271(f)‘s text remains as enacted in 1984, leaving the gap the majority identified.

Implications

  • U.S. software companies can, consistent with Microsoft, ship a master abroad and license foreign replication without §271(f) exposure for the foreign-made copies—provided the copying genuinely occurs outside the United States.
  • Patent drafters seeking global protection cannot rely on U.S. patents to police foreign manufacturing of software-implemented inventions; parallel foreign patents remain the primary tool.
  • The “component must be physical and combinable” holding constrains §271(f) far beyond software, affecting any case built on transmitted information or instructions.
  • The decision is a cornerstone of the modern presumption against extraterritoriality in patent law, later echoed in §271(f) component-counting and in the damages analysis of WesternGeco.
  • Litigants framing infringement theories around intangibles must locate a tangible, U.S.-origin article that is actually combined abroad—an exacting requirement.

Frequently asked questions

Did Microsoft escape all liability? No. Microsoft conceded infringement for copies of Windows made and installed on computers in the United States. The decision concerned only the computers manufactured and sold abroad.

Why did the master disk not count as a supplied component? Because the master was never installed in the foreign-sold computers. The components actually combined into those machines were copies generated overseas, and §271(f) covers only components “supplie[d] … from the United States.”

Does this mean software can never be a §271(f) component? Not quite. The Court held that abstract code is not a component, and that foreign-made copies are not “supplied from the United States.” A copy actually made in the U.S. and shipped abroad for combination could present a different question.

Authorities and sources