Indirect and Divided Patent Infringement: Inducement, Contributory Liability, and the Akamai Problem

Summary. Most modern patent cases accuse someone who never performed the whole invention: a software vendor whose customers run the claimed method, a component supplier whose part only infringes once a downstream buyer installs it, or a drug maker whose label tells physicians what to do. This article explains the three doctrines that reach those defendants, inducement under section 271(b), contributory infringement under section 271(c), and attribution of divided performance under section 271(a) after Akamai, together with the export and import provisions of sections 271(f) and 271(g). It walks through the knowledge and intent requirements set by Global-Tech and Commil, the substantial noninfringing use test, the direction-or-control and joint-enterprise standards, the pleading rules after Twombly, the pre-suit versus post-suit knowledge split in the district courts, and the skinny-label pharmaceutical cases. It then covers damages, willfulness, customer suits, notice letters, discovery, expert proof, summary judgment, jury instructions, and the full menu of defenses. Two worked examples, a SaaS vendor whose customers perform claim steps and a chipmaker shipping a component abroad, show how the doctrines combine in practice.

Suppose Juniper Labs sells a cloud analytics platform. Its customers upload data, click a button that triggers Juniper's servers to run a scoring model, and then review the output on their own laptops. A competitor holds a method patent whose claim recites all three steps: receiving the data, scoring it, and displaying the result. Nobody performs the whole method. Juniper's servers do the middle step; the customer does the first and last. Is anyone liable? The honest answer is that it depends on facts that neither Juniper nor the patentee has looked at yet: what Juniper's terms of service require, what its onboarding documentation says, what Juniper knew about the patent and when, and whether the scoring engine has any real use other than the claimed one.

That is the terrain of indirect and divided infringement. It is where most software, medical-device, and pharmaceutical patent cases are actually won and lost, because the defendant who makes the money is rarely the one who performs every step.

The short answer

A defendant who does not itself make, use, sell, offer to sell, or import the complete claimed invention can still be liable in four ways:

1. Inducement under 35 U.S.C. § 271(b). The defendant knew of the patent, knew (or was willfully blind to the fact) that the acts it was encouraging would infringe, took active steps to bring those acts about, and someone actually directly infringed as a result. Global-Tech Appliances, Inc. v. SEB S.A., 563 U.S. 754 (2011); Commil USA, LLC v. Cisco Systems, Inc., 575 U.S. 632 (2015).
2. Contributory infringement under § 271(c). The defendant sold or imported a component that is a material part of the invention, that it knew was especially made or adapted for an infringing use, and that has no substantial noninfringing use, and someone directly infringed with it. Aro Manufacturing Co. v. Convertible Top Replacement Co., 377 U.S. 476 (1964) (Aro II).
3. Divided (joint) infringement under § 271(a). The defendant performed some steps of a method claim and the remaining steps are attributable to it because it directed or controlled the other actor, or because the actors formed a joint enterprise. Akamai Technologies, Inc. v. Limelight Networks, Inc., 797 F.3d 1020 (Fed. Cir. 2015) (en banc).
4. Export or import liability under § 271(f) and § 271(g). The defendant supplied components from the United States for combination abroad, or imported a product made abroad by a patented process. Life Technologies Corp. v. Promega Corp., 580 U.S. 140 (2017); Microsoft Corp. v. AT&T Corp., 550 U.S. 437 (2007).

Two rules sit underneath all of them. First, there is no indirect infringement without a direct infringement by someone. Limelight Networks, Inc. v. Akamai Technologies, Inc., 572 U.S. 915 (2014). Second, indirect infringement, unlike direct infringement, is a knowledge-and-intent tort: the patentee must prove what the defendant knew and meant, which turns ordinary patent cases into document cases about marketing, instructions, engineering choices, and legal advice.

For the elements of direct infringement itself, start with What Constitutes Patent Infringement: Claims and Defenses: A Practical Guide. This article assumes that background and concentrates on the secondary theories.

The statutory map: section 271 from (a) to (g)

Everything in this area is built on 35 U.S.C. § 271. The subsections divide the labor as follows.

The commentators treat § 271(b) and (c) together as "indirect" or "secondary" liability, and the chapter in 5 Donald S. Chisum, Chisum on Patents ch. 17 remains the most complete map. The Federal Circuit Bar Association's Model Patent Jury Instructions (the current edition, sections on induced and contributory infringement) are the practical reference most trial judges begin from, and this article follows their element structure.

The 1952 Act codified doctrines that had been judge-made for seventy years. Contributory infringement traces to Wallace v. Holmes, 29 F. Cas. 74 (C.C.D. Conn. 1871), where a defendant sold every part of a patented lamp except the chimney, confident that the buyer would supply it. Section 271(d), which immunizes certain patentee conduct from misuse, was added in the same codification to settle a decade of fights between the contributory-infringement and misuse doctrines. The Supreme Court explained that history at length in Dawson Chemical Co. v. Rohm & Haas Co., 448 U.S. 176 (1980), which held that a patentee who sells an unpatented nonstaple chemical for use in its patented process, and refuses to license competitors to do the same, does not misuse the patent. Section 271(f) came later, in 1984, as Congress's response to Deepsouth Packing Co. v. Laitram Corp., 406 U.S. 518 (1972), which had let a shrimp-deveining machine maker export unassembled parts for assembly abroad. Section 271(g) arrived in 1988 with the Process Patent Amendments Act.

Direct infringement comes first, always

Every indirect theory borrows its wrong from a direct infringement that someone else committed. The Supreme Court said so in Aro I, Aro Manufacturing Co. v. Convertible Top Replacement Co., 365 U.S. 336 (1961): if the car owners who replaced worn convertible-top fabric were only repairing (which is permitted), the supplier of replacement fabric could not be a contributory infringer, because there was nothing to contribute to. It said so again in Limelight, 572 U.S. 915, when it held that a defendant cannot induce infringement of a method claim unless some single actor has performed, or is responsible for, all of the steps.

The Federal Circuit enforces this at trial. In Power Integrations, Inc. v. Fairchild Semiconductor International, Inc., 843 F.3d 1315 (Fed. Cir. 2016), the district court told the jury that inducement did not require the defendant's encouragement to have actually caused anyone to infringe; it was enough that Fairchild "took steps to encourage" infringement whether or not the encouragement succeeded. The court vacated the verdict. Inducement means successfully inducing: the plaintiff must prove that the defendant's conduct led to a third party's direct infringement.

Proving that customers directly infringe when you cannot find the customers

Patentees rarely have a signed confession from an end user. The Federal Circuit permits circumstantial proof. In Lucent Technologies, Inc. v. Gateway, Inc., 580 F.3d 1301 (Fed. Cir. 2009), the claim covered a method of entering dates with an on-screen "date picker." No witness testified to using it in the claimed way, but the jury could infer from the sheer volume of Outlook sales, the design of the feature, and Microsoft's own instructions that at least one customer had done so. That was enough for liability (the damages award was vacated on other grounds, which is the usual fate of indirect-infringement verdicts, as discussed below).

The inference has limits. Where the accused product is merely capable of infringing use and the instructions teach a noninfringing mode, courts have refused to presume that anyone infringed. The Federal Circuit has set aside verdicts, or affirmed judgments as a matter of law, where the patentee proved that a product could be used in the claimed way but not that anyone actually used it that way. ACCO Brands, Inc. v. ABA Locks Manufacturer Co., 501 F.3d 1307 (Fed. Cir. 2007), reversed an inducement verdict over a computer lock that could be operated in either an infringing or a noninfringing way; the patentee's expert testifying that he had used it the infringing way was not evidence that customers did. Mirror Worlds, LLC v. Apple Inc., 692 F.3d 1351 (Fed. Cir. 2012), put the point bluntly: excerpts from user manuals, standing alone, do not prove the predicate direct infringement for a product that can be used in a noninfringing manner. In Fujitsu Ltd. v. Netgear Inc., 620 F.3d 1321 (Fed. Cir. 2010), the court held that compliance with an industry standard can prove infringement when the claims cover every standard-compliant device, but where the relevant portion of the standard (packet fragmentation in 802.11) is optional, the patentee must prove that someone actually used it. Summary judgment of no contributory and no induced infringement was affirmed for every product except the four for which the patentee had actual evidence of infringing use.

The practical lesson for plaintiffs: build the direct-infringement record from the defendant's own side of the relationship. Default settings, mandatory workflows, customer support tickets, usage logs, training videos, and standards that mandate rather than permit the claimed behavior are all better than a hired expert reporting that he personally performed the method in his office.

"Use" of a system claim is a different question

Divided-performance problems arise with method claims. System and apparatus claims have their own rule. Under Centillion Data Systems, LLC v. Qwest Communications International, Inc., 631 F.3d 1279 (Fed. Cir. 2011), a party "uses" a claimed system when it puts the system as a whole into service, controlling it and obtaining benefit from it, even if some elements sit on a vendor's servers. The customer who ran Qwest's billing software and pulled reports used the whole system; Qwest, which did not control the customer's personal computer, did not. The vendor's exposure in that posture is inducement of its customers' use, not direct infringement, and that routes the case straight back into § 271(b)'s knowledge and intent requirements. Claim drafters who want to avoid that detour write system claims from the vendor's side of the network, a point developed below.

Inducement under section 271(b): knowledge, intent, and active steps

Section 271(b) is eleven words long: "Whoever actively induces infringement of a patent shall be liable as an infringer." The case law supplies four elements:

1. an act of direct infringement by a third party;
2. the defendant's knowledge of the patent;
3. the defendant's knowledge that the induced acts constitute infringement, which means specific intent to cause infringing conduct, not merely intent to cause the conduct; and
4. active steps (instructions, advertising, design, contractual requirements, technical support) that actually brought the infringement about.

DSU Medical and the specific-intent requirement

The Federal Circuit settled a long internal disagreement in the en banc portion of DSU Medical Corp. v. JMS Co., 471 F.3d 1293 (Fed. Cir. 2006). Some panels had required only that the defendant intend the acts that happen to infringe; others required intent to cause infringement itself. The en banc court chose the latter: the inducer must have "knew or should have known his actions would induce actual infringements," which presupposes knowledge of the patent. On the facts, the Australian supplier ITL had obtained opinions of counsel that its needle guard did not infringe and believed them, and the jury's verdict of no inducement stood.

Global-Tech and willful blindness

The Supreme Court finished the job in Global-Tech, 563 U.S. 754. Pentalpha bought an SEB deep fryer in Hong Kong, copied everything but the cosmetics, and hired a lawyer to write a right-to-use opinion without telling him about the copying. The fryer carried no U.S. patent markings because it was an overseas model. The Court held that § 271(b) requires knowledge that the induced acts constitute patent infringement, which carries with it knowledge of the patent, and that willful blindness satisfies the requirement. Willful blindness has two parts: the defendant subjectively believed there was a high probability that the fact (here, a patent covering the product) existed, and the defendant took deliberate actions to avoid learning it. The Court rejected the Federal Circuit's looser "deliberate indifference to a known risk" formulation but affirmed because the record met the stricter test. Recklessness and negligence do not suffice.

In practice, willful blindness is proved from a pattern: copying a competitor's product, a decision not to run a freedom-to-operate search when the company runs them for everything else, an instruction to engineers not to read patents, or an opinion of counsel procured on incomplete facts. See Conducting Freedom-to-Operate Analysis for New Products for how the search side works.

Commil: belief in invalidity is not a defense, belief in noninfringement is

Commil, 575 U.S. 632, closed a door the Federal Circuit had opened. Cisco argued that it could not have intended to induce infringement because it believed in good faith that Commil's patent was invalid. The Court disagreed: infringement and validity are separate questions under the Patent Act, a patent is presumed valid under 35 U.S.C. § 282, and an accused inducer who thinks the patent is invalid has the usual remedies (declaratory judgment, inter partes review, an invalidity defense) but may not use that belief to negate intent. The Court reaffirmed Global-Tech's rule that the inducer must know the induced acts infringe, and it explained why: the contrary rule would make a defendant liable "even if the defendant reads the patent's claims differently from the plaintiff, and that reading is reasonable." A good-faith, reasonable belief in noninfringement therefore still negates the intent element. The Federal Circuit had applied that principle long before Commil; DSU itself affirmed a no-inducement verdict resting on the defendant's noninfringement opinions.

The asymmetry is deliberate and it drives defense strategy. An accused inducer's noninfringement opinion is a shield. Its invalidity opinion is not, at least on the inducement question, though it remains relevant to willfulness and enhancement.

Commil also contained a paragraph that patentees dislike and defendants quote: the Court observed that frivolous infringement suits can be met with sanctions under Fed. R. Civ. P. 11 and fee awards under 35 U.S.C. § 285.

What counts as an active step: instructions, design, and marketing

The Federal Circuit's instruction cases draw a line that is easy to state and hard to apply. Instructions that merely describe how a product can be used in an infringing way do not prove intent; instructions that recommend or direct the infringing use do.

• In Vita-Mix Corp. v. Basic Holding, Inc., 581 F.3d 1317 (Fed. Cir. 2009), the blender instructions taught a stirring technique that did not infringe. The fact that some users might ignore the instructions and infringe did not show that the maker intended infringement. Summary judgment of no inducement, and of no contributory infringement because the blender had substantial noninfringing uses, was affirmed. The opinion also supplies the standard phrasing of the noninfringing-use test: a use is substantial unless it is "unusual, far-fetched, illusory, impractical, occasional, aberrant, or experimental."
• In Toshiba Corp. v. Imation Corp., 681 F.3d 1358 (Fed. Cir. 2012), the accused DVD media could be used without "finalizing" the disc, and unfinalized discs did not infringe. Toshiba offered no survey or usage evidence showing that unfinalized use was insubstantial, so summary judgment of no contributory infringement was affirmed. But the same defendants' instructions told users that finalization was necessary for playback on other devices, and the court vacated summary judgment of no inducement: a substantial noninfringing use defeats § 271(c), but it does not defeat § 271(b) where the defendant's own materials steer users toward the infringing mode.
• In Ricoh Co. v. Quanta Computer Inc., 550 F.3d 1325 (Fed. Cir. 2008), the accused optical disc drives performed many noninfringing functions, but the court held that a defendant cannot escape contributory liability by embedding a component that has no noninfringing use (here, the firmware and circuitry that executed the patented write methods) inside a larger product that does. It also reversed summary judgment on inducement because the evidence of Quanta's design choices and knowledge raised a triable intent question.
• In Lifetime Industries, Inc. v. Trim-Lok, Inc., 869 F.3d 1372 (Fed. Cir. 2017), the complaint alleged that Trim-Lok hired two of Lifetime's former employees who knew the patent, then sent its own representative to a Forest River plant to install the accused seal on an RV. That was enough to plead knowledge, intent, and an act of inducement (and, for that matter, direct infringement by making the claimed combination).

Where instructions push in both directions, the record is for a jury. Where the defendant's marketing touts the infringing feature by name, the plaintiff usually survives summary judgment; where the feature is buried in a manual as one option among many and the default is noninfringing, the defendant usually wins.

Pre-suit versus post-suit knowledge, and the opinion-of-counsel question

Because knowledge is an element, when the defendant learned of the patent determines how much of the accused conduct is actionable as inducement. Three sources of knowledge recur: a notice letter, a prior lawsuit or licensing discussion, and the complaint itself.

The district court split on whether the complaint supplies knowledge

Courts agree that a defendant who keeps instructing customers after being served with a complaint identifying the patent and the accused conduct has knowledge from that point. They disagree about whether the complaint can be the only source of knowledge pleaded in the same action.

One camp holds that it cannot. The leading statement is Chief Judge Connolly's opinion in ZapFraud, Inc. v. Barracuda Networks, Inc., 528 F. Supp. 3d 247 (D. Del. 2021), which adopted the rule that an operative complaint fails to state an indirect-infringement claim where the defendant's alleged knowledge of the patent rests solely on that complaint or an earlier version of it filed in the same suit. The reasoning is that a claim built on knowledge the complaint itself supplies pleads what might happen, not what has happened, and that the rule rewards patentees who skip the pre-suit letter. Delaware judges have applied it since, for example in WSOU Investments, LLC v. Xilinx, Inc., 2021 WL 2328014 (D. Del. 2021), and similar reasoning appears in Ravgen, Inc. v. Quest Diagnostics Inc., 2022 WL 2047613 (C.D. Cal. 2022), and in Judge Alsup's ruling in Sonos, Inc. v. Google LLC, 2022 WL 799367 (N.D. Cal. 2022), which held that a copy of the complaint delivered one day before filing gave Google no meaningful opportunity to stop or take a license (the Federal Circuit declined to hear an interlocutory appeal).

The other camp holds that post-suit knowledge is enough for post-filing conduct, reasoning that a contrary rule would give a defendant license to keep inducing infringement after it has been told exactly what the patent is and why its conduct infringes. Representative decisions include RightQuestion, LLC v. Samsung Electronics Co., 2022 WL 507487 (E.D. Tex. 2022); Seven Networks, LLC v. Motorola Mobility LLC, 2022 WL 426589 (N.D. Tex. 2022) (dismissing pre-suit inducement but keeping the post-suit claim); Merrill Manufacturing Co. v. Simmons Manufacturing Co., 2021 WL 3493565 (N.D. Ga. 2021); Scramoge Technology Ltd. v. mophie Inc., 2022 WL 2203674 (C.D. Cal. 2022); and Teradata US, Inc. v. SAP SE, 2021 WL 326930 (N.D. Cal. 2021). The Western District of Texas has typically dismissed pre-suit indirect claims that rest only on the complaint while allowing post-suit claims and permitting the patentee to replead pre-suit knowledge once fact discovery opens; see Castlemorton Wireless, LLC v. Bose Corp., 2020 WL 6578418 (W.D. Tex. 2020), and USC IP Partnership, L.P. v. Facebook, Inc., 2021 WL 3134260 (W.D. Tex. 2021). Judges within the same district disagree, the Federal Circuit has not resolved the question in a precedential opinion, and the case collections in 6 Robert A. Matthews, Jr., Annotated Patent Digest §§ 39:7.25 and 39:7.50 are the quickest way to find the assigned judge's own rulings before you plead.

The safe course for a patentee is obvious: send a letter before suing, or at least plead facts (prior litigation on the same patent, citations to the patent in the defendant's own prosecution history, licensing negotiations, trade press) that show pre-suit knowledge independent of the complaint. The safe course for a defendant who receives such a letter is less obvious, because the letter is also what creates declaratory judgment jurisdiction; see the section on notice letters below.

Opinions of counsel and 35 U.S.C. § 298

Before 2011, an accused inducer faced a dilemma. Relying on an opinion of counsel to show good faith waived privilege over the opinion and, often, over trial counsel's related communications; not relying on it invited an adverse inference. The America Invents Act added 35 U.S.C. § 298, which provides that the failure to obtain advice of counsel, or the failure to present such advice to the court or jury, may not be used to prove that the accused infringer willfully infringed or intended to induce infringement. The section applies to civil actions commenced on or after January 14, 2013.

Section 298 removes the adverse inference. It does not make opinions worthless. A competent, timely noninfringement opinion remains the single best piece of evidence that an accused inducer lacked the intent Commil requires, and it is usually the centerpiece of a defendant's summary judgment motion on inducement. Opinions limited to invalidity do less work after Commil. An opinion obtained on incomplete facts, as in Global-Tech, can affirmatively hurt.

The pharmaceutical cases: labels, skinny labels, and physicians

Method-of-treatment patents are the purest inducement cases. The generic manufacturer never treats anyone; physicians and patients do. The generic's label is the instruction manual, and under Hatch-Waxman the generic is required to copy the brand's label except for patented indications it "carves out" under 21 U.S.C. § 355(j)(2)(A)(viii), producing a so-called skinny label. For the regulatory framework, see Hatch-Waxman Litigation: The Orange Book, Paragraph IV Certifications, and the 30-Month Stay.

• Eli Lilly & Co. v. Teva Parenteral Medicines, Inc., 845 F.3d 1357 (Fed. Cir. 2017), involved a method of administering pemetrexed after pretreatment with folic acid and vitamin B12. The patient took the folic acid; the physician gave the drug. The court held that physicians direct or control the patients' step under Akamai: they condition pemetrexed treatment on the pretreatment and establish its manner and timing by prescribing dose and schedule. Because a single actor (the physician) was responsible for every step, Teva's label, which instructed exactly that regimen, induced infringement.
• Sanofi v. Watson Laboratories Inc., 875 F.3d 636 (Fed. Cir. 2017), held that a label whose indication section described the claimed patient population and clinical benefit encouraged the infringing use, and that the existence of substantial noninfringing uses of the drug was no defense to inducement (it would be a defense only to contributory infringement).
• GlaxoSmithKline LLC v. Teva Pharmaceuticals USA, Inc., 7 F.4th 1320 (Fed. Cir. 2021), reinstated a jury verdict that Teva induced infringement of a method of treating heart failure with carvedilol even during the years Teva sold under a skinny label. The majority relied on the content of the partial label (which still described a use that the jury could find met the claim), Teva's press releases describing its product as the AB-rated generic equivalent of Coreg, and its later full label. Judge Prost dissented at length, and the decision drew heavy criticism from the generic industry for the proposition that a carve-out does not reliably avoid inducement liability. The Supreme Court denied review in 2023.
• Amarin Pharma, Inc. v. Hikma Pharmaceuticals USA Inc., 104 F.4th 1370 (Fed. Cir. 2024), reversed a Rule 12(b)(6) dismissal. Hikma's skinny label omitted the patented cardiovascular-risk indication, but its press releases called the product the "generic version" of Vascepa and quoted total sales figures that were mostly attributable to the carved-out use, and its website listed the drug under a therapeutic category matching the patented use. Taken together, the court said, those allegations plausibly pleaded inducement. The panel stressed the pleading posture and did not hold that the conduct would survive summary judgment.

The defense side of the ledger matters as much. In H. Lundbeck A/S v. Lupin Ltd., 87 F.4th 1361 (Fed. Cir. 2023), the patentee relied solely on the generic's label, which was the label the FDA required for an unpatented indication and which predated the patent; with no promotional statements to point to, the court affirmed a judgment of noninfringement and expressly distinguished GSK. Earlier cases set the baseline that mere knowledge that some physicians will prescribe off-label is not inducement, Warner-Lambert Co. v. Apotex Corp., 316 F.3d 1348 (Fed. Cir. 2003), that "vague label language cannot be combined with speculation about how physicians may act to find inducement," Takeda Pharmaceuticals U.S.A., Inc. v. West-Ward Pharmaceutical Corp., 785 F.3d 625 (Fed. Cir. 2015), and that a label which merely permits the claimed step, without encouraging it, does not induce, HZNP Medicines LLC v. Actavis Laboratories UT, Inc., 940 F.3d 680 (Fed. Cir. 2019). On the other side, a label that recommends the claimed practice to the prescriber will carry an inducement finding, as in Vanda Pharmaceuticals Inc. v. West-Ward Pharmaceuticals International Ltd., 887 F.3d 1117 (Fed. Cir. 2018), and AstraZeneca LP v. Apotex, Inc., 633 F.3d 1042 (Fed. Cir. 2010). Skinny-label law has moved quickly since 2021, so check the subsequent history of Amarin and its progeny before relying on them.

The pattern across these cases: the label is necessary but rarely sufficient on either side. Generics that carve out cleanly and then market the product as a full substitute invite a GSK or Amarin problem. Brand companies that rely on the label alone, without evidence that physicians actually read and follow it in the claimed way, invite a Fujitsu-style failure of proof on direct infringement.

Contributory infringement under section 271(c)

Section 271(c) reaches whoever sells, offers to sell, or imports "a component of a patented machine, manufacture, combination or composition, or a material or apparatus for use in practicing a patented process, constituting a material part of the invention, knowing the same to be especially made or especially adapted for use in an infringement of such patent, and not a staple article or commodity of commerce suitable for substantial noninfringing use." Five elements follow:

1. a direct infringement by someone;
2. the defendant sold, offered, or imported a component or material;
3. the component is a material part of the invention;
4. the defendant knew the component was especially made or adapted for a use that infringes the patent (which, after Aro II and Global-Tech, includes knowledge of the patent); and
5. the component is not a staple article suitable for substantial noninfringing use.

Aro II and the knowledge requirement

Aro II, 377 U.S. 476, is the case that put knowledge into § 271(c). On the second trip to the Supreme Court, the facts had changed: the convertible tops at issue were on Ford cars, and Ford was unlicensed, so the owners' replacement of fabric was infringing (there is no right to repair an infringing article). The fabric supplier could be a contributory infringer, but only from the date it received a letter from the patentee identifying the patent and the infringing combination. The Court read "knowing" to require knowledge that the combination for which the component was especially designed was both patented and infringing. That is why the notice letter has been the opening move in contributory-infringement cases for sixty years.

Material part, staple article, and the "component" that matters

The three remaining elements are usually litigated together. The two Federal Circuit cases every brief cites are:

• i4i Ltd. Partnership v. Microsoft Corp., 598 F.3d 831 (Fed. Cir. 2010), which affirmed a verdict of contributory infringement against Microsoft for the custom XML editor in Word. Microsoft argued that Word had countless noninfringing uses. The court held that the relevant "component" was the XML editor, a separable feature, not the whole program, and that the editor had no substantial noninfringing use. The knowledge evidence was circumstantial: Word developers had attended an i4i presentation about software practicing the patent, and internal emails showed Microsoft employees receiving i4i marketing materials that cited the patent number and predicting that Word's editor would make i4i's product "obsolete." (The Supreme Court later affirmed on the unrelated question of the clear-and-convincing standard for invalidity, Microsoft Corp. v. i4i Ltd. Partnership, 564 U.S. 91 (2011).)
• Fujitsu, 620 F.3d 1321, applied the same logic to the hardware and software in a router that performed 802.11 fragmentation: that functionality, not the router, was the component, and when used it always infringed.

The staple-article inquiry is therefore about the accused feature, not the box it ships in. Defendants answer by showing that the feature itself has real, non-trivial uses that do not meet the claim (as in Vita-Mix and Toshiba), and the plaintiff bears the burden of proving the absence of such uses. A use the defendant's own documents recommend against can still be substantial. A use that is theoretically possible but that no customer actually adopts usually is not.

"Material part" is rarely the decisive element, but it has teeth where the accused component is a commodity (a generic resistor, a standard chemical reagent) that happens to be used in the patented combination. The Supreme Court's discussion in Dawson Chemical of propanil, a nonstaple with no use other than the patented herbicidal method, remains the paradigm of a material, especially adapted, nonstaple component.

Divided infringement: the Akamai problem

A method claim is directly infringed only when every step is performed. When two or more independent actors each perform some steps and nobody performs them all, there is no direct infringement unless the law attributes one actor's steps to another. For a decade the Federal Circuit made attribution very hard, and the Supreme Court then made it harder before the en banc court made it easier.

BMC Resources and Muniauction: the single-entity rule

BMC Resources, Inc. v. Paymentech, L.P., 498 F.3d 1373 (Fed. Cir. 2007), involved a PIN-less debit transaction method whose steps were spread across a merchant, a payment processor, a debit network, and a bank. The court held that direct infringement requires a single party to perform every step or to "direct or control" those who do, and that arm's-length cooperation among independent businesses is not direction or control. It added a line that claim drafters have quoted ever since: the problem "can usually be offset by proper claim drafting," because a patentee can write the claim to cover the acts of a single actor.

Muniauction, Inc. v. Thomson Corp., 532 F.3d 1318 (Fed. Cir. 2008), applied the rule to an online bond auction in which bidders performed the "inputting" step and Thomson performed the rest. Thomson controlled access to its system and told bidders how to use it, but that, the court said, is not control of the bidders' performance. The verdict was reversed. The "mastermind" vocabulary of this period required something close to agency or a contractual obligation.

Limelight (2014): no inducement without a single direct infringer

Akamai's patent claimed a content-delivery method in which content was "tagged" so that it would be served from a distributed network. Limelight performed every step except tagging, which its customers did. A panel had held that there was no direct infringement under Muniauction and therefore no inducement, Akamai Technologies, Inc. v. Limelight Networks, Inc., 629 F.3d 1311 (Fed. Cir. 2010). The en banc Federal Circuit tried a different route in Akamai Technologies, Inc. v. Limelight Networks, Inc., 692 F.3d 1301 (Fed. Cir. 2012) (en banc) (per curiam): it held that a defendant could induce infringement by performing some steps itself and inducing others to perform the rest, even though no single entity directly infringed. The Supreme Court reversed unanimously in Limelight, 572 U.S. 915. Section 271(b) borrows its definition of infringement from § 271(a); if no one has committed direct infringement under § 271(a), there is nothing to induce. The Court expressly declined to review the Muniauction standard itself and invited the Federal Circuit to reconsider it on remand.

Akamai (Fed. Cir. en banc 2015): direction or control, broadened

The Federal Circuit took the invitation. Its per curiam en banc opinion, Akamai, 797 F.3d 1020, holds that an entity is responsible for others' performance of method steps in two circumstances: where it directs or controls their performance, and where the actors form a joint enterprise. Direction or control includes the traditional agency and contract cases, and also a new category: where the alleged infringer (1) conditions participation in an activity or receipt of a benefit on the performance of a step or steps of the patented method, and (2) establishes the manner or timing of that performance. Whether that test is met is a question of fact.

A joint enterprise requires four things borrowed from the Restatement (Second) of Torts § 491: an express or implied agreement among the members, a common purpose, a community of pecuniary interest in that purpose, and an equal right to a voice in the direction of the enterprise. Each member is then charged with the acts of the others.

On the facts, the court reinstated the jury's verdict. Limelight's standard contract required customers to perform the tagging and serving steps if they wanted to use the network; Limelight's welcome letter and installation guidelines told them exactly how to do it, and Limelight engineers helped. That was conditioning plus manner-or-timing, and it was enough.

Travel Sentry and the reach of the new test

The new test was not confined to Akamai's facts. In Travel Sentry, Inc. v. Tropp, 877 F.3d 1370 (Fed. Cir. 2017), the claimed method of luggage screening had steps performed by Travel Sentry (supplying dual-access locks and master keys) and steps performed by the Transportation Security Administration (identifying the locks and opening them with the keys). There was no contract compelling the TSA to do anything, only a three-page memorandum of understanding under which Travel Sentry supplied passkeys at no cost and the TSA promised good-faith efforts to use them "whenever practicable." The court vacated summary judgment of noninfringement. The "benefit" the TSA received was the ability to open and relock certified luggage without breaking the locks, and a jury could find that receipt of that benefit was conditioned on using the passkeys in the manner Travel Sentry specified. The panel held that the district court had defined the relevant activity and benefit too narrowly, that intangible benefits count, and that conditioning does not require a legal obligation or a technological prerequisite, while repeating Eli Lilly's caution that "mere guidance or instruction" is not enough. The question is whether the defendant made the step a precondition of something the other actor wanted.

Eli Lilly v. Teva Parenteral, 845 F.3d 1357, discussed above, is the medical version: physicians condition treatment on the patient's pretreatment step and set the dose and schedule.

What the post-Akamai cases actually turn on

Three facts decide most divided-infringement disputes today:

1. Is the other actor's step a precondition of something the defendant controls access to? A terms-of-service clause that says "you must configure X before the service will function" is strong. A help article that says "you may configure X" is weak.
2. Does the defendant dictate how or when the step is done? Required data formats, mandatory API calls, prescribed sequences, and dosing schedules are manner-or-timing evidence. "Here are some options" is not.
3. Does the defendant benefit from the step being performed? The joint-enterprise prong rarely succeeds because independent commercial parties almost never share an equal right of control, but the "receipt of a benefit" language in the conditioning test has done considerable work.

How claim drafting avoids the problem

The cleanest solution is to write the claim so that one actor performs every step. Prosecutors do this by:

• drafting method steps from the perspective of the party you expect to sue ("receiving, at a server, a request..." rather than "a user sending a request...");
• replacing steps performed by others with steps that respond to or process what the others did ("in response to receiving data tagged by a client..." makes the server the only actor);
• pairing every method claim with a system or computer-readable-medium claim whose elements sit on the vendor's side, so that Centillion-style "use" analysis is unnecessary; and
• including at least one claim set drafted from the customer's side as well, so that the patentee can prove direct infringement by customers and plead inducement against the vendor in the alternative.

For how claims get construed once litigation starts, see Claim Construction and Markman Hearings: How Courts Decide What a Patent Actually Covers. A single construction of "user" or "server" can convert a single-actor claim into a divided one.

Exports and imports: sections 271(f) and 271(g)

Indirect liability does not stop at the border, but the border changes the rules. Three settings recur: a foreign company whose conduct abroad induces infringement in the United States, a supplier that ships components out of the United States for assembly overseas, and an importer of goods made abroad by a patented process.

Inducing U.S. infringement from abroad

The direct infringement must happen here; the inducing conduct need not. The jury instruction approved in DSU Medical told jurors that inducement "does not require any activity by the indirect infringer in this country, as long as the direct infringement occurs here," and the Federal Circuit later put it as a holding: liability under § 271(b) "can be imposed based on extraterritorial acts, provided that the patentee proves the defendant possessed the requisite knowledge and specific intent to induce direct infringement in the United States." Enplas Display Device Corp. v. Seoul Semiconductor Co., 909 F.3d 398 (Fed. Cir. 2018). In the 2016 Power Integrations decision, Fairchild sold its controller chips overseas into a worldwide distribution system and argued that its sales were market-agnostic. The court found enough for a jury anyway: Fairchild shipped demonstration boards to U.S. customers and promoted its chips as compliant with U.S. energy-efficiency standards, which allowed a finding that it induced its customers, as a class, to bring infringing products into the country. In MEMC Electronic Materials, Inc. v. Mitsubishi Materials Silicon Corp., 420 F.3d 1369 (Fed. Cir. 2005), a Japanese wafer maker that sold only to its customer's Japanese affiliate escaped direct infringement, but the court reversed summary judgment on inducement because the supplier built wafers to its customer's specifications and gave technical support to the customer's fab in Austin.

The damages side is where foreign sellers usually win. In Power Integrations, Inc. v. Fairchild Semiconductor International, Inc., 711 F.3d 1348 (Fed. Cir. 2013), the court refused damages measured by Fairchild's worldwide sales, reasoning that entirely foreign production, use, or sale ordinarily breaks the causal chain from domestic infringement. The Supreme Court's later decision in WesternGeco LLC v. ION Geophysical Corp., 585 U.S. 407 (2018), allowed lost foreign profits for § 271(f)(2) infringement, and in Brumfield v. IBG LLC, 97 F.4th 854 (Fed. Cir. 2024), the Federal Circuit said that extraterritorial damages questions should now be analyzed under the WesternGeco framework; a patentee who wants a royalty enlarged by foreign conduct that is not itself infringing must at least show how that conduct increases the value of the domestic infringement. Three other decisions shape the royalty base:

• Enplas vacated a damages award because the expert's royalty base included lenses that never ended up in products imported into the United States.
• Carnegie Mellon University v. Marvell Technology Group, Ltd., 807 F.3d 1283 (Fed. Cir. 2015), affirmed a royalty on chips made and delivered abroad that were later imported, but ordered a new trial on whether chips that never entered the country were "sold" here.
• Halo Electronics, Inc. v. Pulse Electronics, Inc., 831 F.3d 1369 (Fed. Cir. 2016), held that components manufactured, shipped, and delivered abroad were neither sold nor offered for sale in the United States, even though the supplier negotiated prices with its U.S. customer here, met with the customer's U.S. engineers, and sent samples for approval.

Section 271(f): components shipped out for assembly abroad

Section 271(f) is a deliberate exception to territoriality. Subsection (f)(1) reaches whoever supplies "all or a substantial portion of the components" of a patented invention from the United States, uncombined, in a manner that actively induces their combination abroad. Subsection (f)(2) reaches whoever supplies even a single component that is especially made or adapted for use in the invention and is not a staple article, knowing it is so made and intending that it be combined abroad. Three Supreme Court decisions define the edges:

1. Microsoft v. AT&T, 550 U.S. 437, held that software in the abstract is not a "component," and that Microsoft did not "supply" from the United States the copies of Windows that foreign manufacturers made from a master disk shipped abroad. Only the physical copies actually installed count, so golden masters copied overseas fall outside the statute.
2. Life Technologies, 580 U.S. 140, held that "substantial portion" in (f)(1) is quantitative and that a single component (there, the Taq polymerase in a genetic testing kit) can never be a substantial portion. A single especially made component is an (f)(2) problem, not an (f)(1) problem.
3. WesternGeco, 585 U.S. 407, held that a patentee who proves (f)(2) infringement can recover lost profits from foreign uses of the combined invention, because the infringing conduct the statute focuses on, the domestic act of supplying the components, occurred here.

Two Federal Circuit holdings matter in practice. Section 271(f) does not apply to method claims, because the components of a method are steps, which cannot be "supplied." Cardiac Pacemakers, Inc. v. St. Jude Medical, Inc., 576 F.3d 1348 (Fed. Cir. 2009) (en banc in relevant part). And (f)(2) does not require proof that the combination actually happened abroad; the defendant's intent that it be combined is the offense. Waymark Corp. v. Porta Systems Corp., 245 F.3d 1364 (Fed. Cir. 2001).

Section 271(g): products made abroad by a patented process

Section 271(g) makes it an infringement to import, sell, offer to sell, or use in the United States a product made by a process patented here. It closes the gap that otherwise lets a manufacturer practice a U.S. process patent in a foreign plant and ship the output home. Two statutory exceptions do most of the work: the product is not covered if it is "materially changed by subsequent processes" or becomes "a trivial and nonessential component of another product." Eli Lilly & Co. v. American Cyanamid Co., 82 F.3d 1568 (Fed. Cir. 1996), applied the material-change exception to a chemical intermediate that went through several further reactions before becoming the imported antibiotic. The courts have also read "made by" narrowly. Section 271(g) covers physical products, not information generated by a patented screening method, Bayer AG v. Housey Pharmaceuticals, Inc., 340 F.3d 1367 (Fed. Cir. 2003), and it covers the actual making of a product, not quality-control testing that confirms the product was made correctly, Momenta Pharmaceuticals, Inc. v. Teva Pharmaceuticals USA Inc., 809 F.3d 610 (Fed. Cir. 2015). Remedies against innocent downstream sellers are limited by 35 U.S.C. § 287(b), which is one reason process-patent owners send notice letters to importers early.

Pleading and proving indirect infringement: from the complaint to the verdict

The doctrine above decides cases only if it can be pleaded and proved. This section summarizes how the elements play out at each stage; the companion guide, Pleading and Proving Induced and Contributory Patent Infringement: A Practical Guide, walks through the same sequence step by step.

Pleading after Twombly and Iqbal

Fed. R. Civ. P. 8(a)(2) requires a short and plain statement showing entitlement to relief, and Bell Atlantic Corp. v. Twombly, 550 U.S. 544 (2007), and Ashcroft v. Iqbal, 556 U.S. 662 (2009), require factual content that makes the claim plausible rather than merely conceivable. For a few years the bare-bones patent complaint of old Form 18 shielded direct-infringement claims, but even then the Federal Circuit held that the form did not govern indirect infringement, and the form was abrogated in December 2015. Plausibility now governs every patent claim.

The leading decision remains In re Bill of Lading Transmission & Processing System Patent Litigation, 681 F.3d 1323 (Fed. Cir. 2012). The court reinstated inducement claims against vendors of in-cab trucking technology because the complaints, built on quotations from the defendants' own websites, advertising, and industry publications, supported a reasonable inference that the vendors knew of the patent and intended their customers to use the products in the claimed way. The patentee did not have to name a specific direct infringer or prove its case at the pleading stage. The same court affirmed dismissal of the contributory claims because the complaints themselves described noninfringing uses of the products (tracking trucks, monitoring conditions, speeding up billing), which pleaded the patentee out of § 271(c). Superior Industries, LLC v. Thor Global Enterprises Ltd., 700 F.3d 1287 (Fed. Cir. 2012), shows the other outcome: an inducement count with no facts about knowledge or intent was properly dismissed.

Divided-infringement theories need their own facts. In Lyda v. CBS Corp., 838 F.3d 1331 (Fed. Cir. 2016), a complaint alleging that a television network directed or controlled the independent contractors who performed steps of a method for collecting real-time audience responses was dismissed because it pleaded no facts about that direction or control. In Nalco Co. v. Chem-Mod, LLC, 883 F.3d 1337 (Fed. Cir. 2018), the court agreed that selling a coal additive to power plants in pursuit of tax credits did not plead a joint enterprise, but reinstated the case because the defendants' own pilot-scale and full-scale testing, as alleged, put all the steps under their direction. Nalco also held that disputes that boil down to claim construction are not suitable for resolution on a Rule 12(b)(6) motion. Finally, Bot M8 LLC v. Sony Corp. of America, 4 F.4th 1342 (Fed. Cir. 2021), reminds plaintiffs that the level of detail required rises with the complexity of the technology and the materiality of the disputed element, and that a complaint can plead facts that defeat its own theory. For a simple mechanical product, photographs and a comparison to the claims may be enough, as in Disc Disease Solutions Inc. v. VGH Solutions, Inc., 888 F.3d 1256 (Fed. Cir. 2018).

A complaint that survives these cases typically pleads: the direct infringers by class (customers, physicians, distributors) and what they do; the specific instructions, marketing, design choices, or contract terms that encourage the infringing acts, quoted or screenshotted; facts showing knowledge of the patent before suit (letter, prior litigation, citation of the patent during the defendant's own prosecution, licensing talks, hiring of people who knew of it); for contributory claims, the specific feature that is the "component" and why it has no substantial use outside the claims; and for divided claims, the conditioning and manner-or-timing facts. For the motion practice, see Motions to Dismiss Under Rule 12: A Practical Guide for Plaintiffs and Defendants, and for the defendant's responsive pleading, Answering a Federal Complaint: Denials, Affirmative Defenses, Counterclaims, and Judgment on the Pleadings.

Discovery that finds (or refutes) intent

Indirect-infringement discovery is document discovery about what the defendant told its customers and what it knew. Requests under Fed. R. Civ. P. 34 should reach user manuals and every version of the online help center, training materials and certification courses, onboarding and implementation guides, sales decks and demo scripts, customer-support tickets and knowledge-base articles, default-setting histories, product requirement documents, competitive analyses mentioning the patentee, and usage telemetry. A Rule 30(b)(6) deposition should cover the date and circumstances of first knowledge of the patent, design decisions about the accused feature, the content and distribution of instructions, and what the company knows about how customers actually use the feature. Interrogatories under Rule 33 pin down the earliest-knowledge date and the factual basis for any good-faith belief, and third-party subpoenas under Rule 45 to a few cooperative customers often produce the cleanest proof of direct infringement in the case. Everything is bounded by the proportionality requirement of Rule 26(b)(1).

The opinion-of-counsel decision drives privilege fights. A defendant that relies on an opinion waives privilege over communications on the same subject matter, including work product communicated to the client, In re EchoStar Communications Corp., 448 F.3d 1294 (Fed. Cir. 2006), although the waiver ordinarily does not reach trial counsel, In re Seagate Technology, LLC, 497 F.3d 1360 (Fed. Cir. 2007) (en banc), whose willfulness standard was later displaced but whose privilege holding survives. Fed. R. Evid. 502 orders can limit collateral damage. The broader privilege framework is covered in Attorney-Client Privilege and Work Product for Businesses: Upjohn, In-House Counsel, and the Common Interest.

Expert proof and summary judgment

Experts carry two loads in these cases. The technical expert proves that someone actually performed the claim, ideally with usage data, logs, standards that mandate rather than permit the claimed behavior, and customer testimony, not with the expert's own demonstration. The same expert (or a survey expert) must quantify noninfringing uses, because the patentee bears the burden of proving their absence, Golden Blount, Inc. v. Robert H. Peterson Co., 438 F.3d 1354 (Fed. Cir. 2006), and Toshiba lost its contributory claim for want of any survey, expert, or usage evidence. Fed. R. Evid. 702, as amended in 2023, requires the proponent to show by a preponderance that each opinion rests on sufficient facts and a reliable application of method. Courts routinely exclude expert testimony about what the defendant intended or knew; Rule 704 permits opinions on ultimate issues, but state of mind is for the jury, and an expert who reads a defendant's emails aloud adds only the expert's credentials. Experts also may not argue claim construction to the jury; it is "improper for juries to hear conflicting expert testimony on the correctness of a claim construction." Warsaw Orthopedic, Inc. v. NuVasive, Inc., 824 F.3d 1344 (Fed. Cir. 2016). The expert-discovery side is covered in Expert Depositions: Taking Apart the Opinion, Protecting Your Own Expert, and Building the Daubert Record.

At summary judgment under Fed. R. Civ. P. 56, defendants win most often on the predicate: no proof that anyone directly infringed (ACCO, Mirror Worlds, Fujitsu), a component with proven noninfringing uses (Toshiba, Vita-Mix), or no evidence of conditioning in a divided case, as in Medgraph, Inc. v. Medtronic, Inc., 843 F.3d 942 (Fed. Cir. 2016), where Medtronic did not condition use of its remote-monitoring system on patients' performance of the claimed steps. Intent itself rarely goes away on summary judgment, as Ricoh, Toshiba, and MEMC all show.

Jury instructions, verdict forms, and judgment as a matter of law on intent

Most courts start from the Federal Circuit Bar Association's model instructions on active inducement and contributory infringement, and the model language is safer than a creative variation: Power Integrations lost an inducement verdict because the court adopted a plaintiff-drafted instruction saying that infringement "need not have been actually caused" by the defendant's encouragement. Object on the record before the jury retires, as Fairchild did, or the error is usually forfeited. Ask for special interrogatories under Fed. R. Civ. P. 49 separating inducement from contributory infringement and pre-notice from post-notice periods; a general verdict that lumps theories together, as in Lucent, makes every damages error harder to cure.

Judgment as a matter of law under Fed. R. Civ. P. 50 is where the predicate bites again. ACCO reversed an inducement verdict because the evidence of direct infringement was only the expert's own use. Commil itself ended this way: after winning the intent question in the Supreme Court, the patentee lost on remand because substantial evidence did not show that Cisco's devices, when used, performed the claimed "running" step. Commil USA, LLC v. Cisco Systems, Inc., 813 F.3d 994 (Fed. Cir. 2015). Intent verdicts, by contrast, usually survive. On remand after Commil, the court in Warsaw Orthopedic again affirmed a judgment that Medtronic's spinal business induced infringement of one of NuVasive's patents, because the jury could find that it knew, or was willfully blind to the fact, that surgeons using its nerve-monitoring system infringed, and Enplas affirmed denial of judgment as a matter of law on inducement while calling it a close question. Post-trial practice is covered in Post-Trial Motions in Federal Court: Judgment as a Matter of Law, New Trial, Remittitur, and Relief from Judgment.

Damages and willfulness when the defendant is an indirect infringer

35 U.S.C. § 284 applies the same measure to every infringer: damages adequate to compensate, and in no event less than a reasonable royalty. Three features make the numbers different for indirect infringers.

The clock starts with knowledge. Inducement and contributory infringement are not actionable before the defendant knew of the patent (or was willfully blind to it). Aro II limited Aro's liability to sales after the patentee's January 1954 letter, absent proof of earlier knowledge. A divided-infringement theory under § 271(a), by contrast, is strict liability, so a patentee who can attribute the customers' steps to the vendor under Akamai may reach back the full six years allowed by 35 U.S.C. § 286. That difference alone often decides which theory a patentee leads with.

Marking can cut off the early years. Under 35 U.S.C. § 287(a), a patentee (or licensee) that sells unmarked products practicing an apparatus claim recovers only from actual notice, and actual notice requires an affirmative communication of a specific charge of infringement by a specific accused product. Amsted Industries Inc. v. Buckeye Steel Castings Co., 24 F.3d 178 (Fed. Cir. 1994). Marking does not apply when only method claims are asserted, Crown Packaging Technology, Inc. v. Rexam Beverage Can Co., 559 F.3d 1308 (Fed. Cir. 2009), but it does apply where apparatus and method claims are asserted together and there is a tangible product to mark, American Medical Systems, Inc. v. Medical Engineering Corp., 6 F.3d 1523 (Fed. Cir. 1993). Once the accused infringer identifies unmarked products it believes practice the claims, the patentee bears the burden of proving compliance. Arctic Cat Inc. v. Bombardier Recreational Products Inc., 876 F.3d 1350 (Fed. Cir. 2017). For the mechanics, see Understanding Patent Marking Requirements: A Practical Guide for Practitioners and Entrepreneurs.

The royalty base follows the direct infringement. The award must be tied to the infringing acts the defendant induced or contributed to. Enplas vacated an award whose base included products that never entered the United States, and Lucent vacated a $358 million lump sum because the evidence did not connect the payment to the value of the small date-picker feature that infringed. A patentee also cannot collect twice for the same infringing units. In Aro II, the Supreme Court held that once a patentee has recovered full compensation from the direct infringer, it cannot collect further damages from someone liable only for contributing to the same infringement. Settlements with customers therefore need careful release language on both sides; see Patent Damages: Reasonable Royalties, Lost Profits, Apportionment, and Enhancement.

Individuals are exposed too. Corporate officers who actively assist their company's infringement can be personally liable for inducement without any veil piercing, Orthokinetics, Inc. v. Safety Travel Chairs, Inc., 806 F.2d 1565 (Fed. Cir. 1986); Wordtech Systems, Inc. v. Integrated Networks Solutions, Inc., 609 F.3d 1308 (Fed. Cir. 2010), although the same knowledge and intent requirements apply to them.

Willfulness and enhancement

Halo Electronics, Inc. v. Pulse Electronics, Inc., 579 U.S. 93 (2016), replaced the old two-part Seagate test with a discretionary inquiry: the district court may enhance damages up to three times in egregious cases of culpable infringement, and the facts are proved by a preponderance. The jury's willfulness question is narrower: whether the infringement was deliberate or intentional. Eko Brands, LLC v. Adrian Rivera Maynez Enterprises, Inc., 946 F.3d 1367 (Fed. Cir. 2020). Knowledge of the patent plus infringement is necessary but not sufficient. Bayer HealthCare LLC v. Baxalta Inc., 989 F.3d 964 (Fed. Cir. 2021). Enhancement itself is guided by the factors in Read Corp. v. Portec, Inc., 970 F.2d 816 (Fed. Cir. 1992), which include deliberate copying, investigation of the patent and formation of a good-faith belief, litigation conduct, and remedial action.

Willfulness and inducement draw on the same evidence, which is why defendants should think about them together. Before the AIA, the Federal Circuit had barred an adverse inference from a failure to obtain an opinion for willfulness purposes, Knorr-Bremse Systeme Fuer Nutzfahrzeuge GmbH v. Dana Corp., 383 F.3d 1337 (Fed. Cir. 2004) (en banc), but it held in Broadcom Corp. v. Qualcomm Inc., 543 F.3d 683 (Fed. Cir. 2008), that the failure to procure an opinion could be probative of intent to induce. Section 298 now forbids using either the failure to obtain advice or the failure to present it to prove willfulness or intent to induce. What remains admissible is the affirmative story: an opinion obtained and followed, a design change after notice, a decision to stop recommending the accused configuration. Fee shifting is a separate question under 35 U.S.C. § 285 and Octane Fitness, LLC v. ICON Health & Fitness, Inc., 572 U.S. 545 (2014), and injunctions follow the four-factor test of eBay Inc. v. MercExchange, L.L.C., 547 U.S. 388 (2006); against an inducer, an injunction often targets the instructions, labels, and marketing rather than the product itself.

Customers, suppliers, notice letters, and indemnity demands

Indirect-infringement disputes are three-cornered. The patentee, the supplier, and the supplier's customers each have different exposure, and most of the early moves in a case are about which corner gets sued first.

Why patentees sue customers, and how suppliers respond

Customers are direct infringers, and direct infringement needs no knowledge or intent. They are also numerous, often less motivated to fight, and sometimes sueable in districts where the supplier is not, since 28 U.S.C. § 1400(b) limits venue against a domestic corporation to its state of incorporation or a district where it has committed infringement and has a regular and established place of business. TC Heartland LLC v. Kraft Foods Group Brands LLC, 581 U.S. 258 (2017). The AIA's joinder provision, 35 U.S.C. § 299, prevents a patentee from joining unrelated defendants in one case merely because they are accused of infringing the same patent, so customer campaigns usually mean many separate suits.

Courts have tools for this. Under the customer-suit exception, litigation against the manufacturer generally takes precedence over suits against its customers, because the manufacturer is "the true defendant." Katz v. Lear Siegler, Inc., 909 F.2d 1459 (Fed. Cir. 1990); see also Codex Corp. v. Milgo Electronic Corp., 553 F.2d 735 (1st Cir. 1977). In In re Nintendo Co., 756 F.3d 1363 (Fed. Cir. 2014), the Federal Circuit ordered a district court to sever and stay the claims against retailers and transfer the claims against Nintendo itself. The exception is narrow; it applies only where resolving the manufacturer's case will resolve the major issues in the customer cases. Spread Spectrum Screening LLC v. Eastman Kodak Co., 657 F.3d 1349 (Fed. Cir. 2011). Forum strategy is covered in Forum Selection, Transfer, and Parallel Proceedings: Choosing and Keeping the Court You Want.

A supplier can also go on offense by filing a declaratory judgment action under 28 U.S.C. § 2201. It has standing if it is obligated to indemnify its customers, or if the patentee's claims against customers implicitly accuse the supplier of inducing or contributing to their infringement. Arris Group, Inc. v. British Telecommunications PLC, 639 F.3d 1368 (Fed. Cir. 2011). In Microsoft Corp. v. DataTern, Inc., 755 F.3d 899 (Fed. Cir. 2014), claim charts sent to customers that relied on the suppliers' own user guides and documentation created a controversy about inducement, but "simply selling a product capable of being used in an infringing manner" was not enough. A supplier of a material with no noninfringing use in the relevant application may even sue over future inducement before launch. Arkema Inc. v. Honeywell International, Inc., 706 F.3d 1351 (Fed. Cir. 2013).

Indemnity is the money question. Absent contrary agreement, a merchant seller warrants that goods will be delivered free of rightful infringement claims under UCC § 2-312(3), and most software and SaaS agreements replace that default with an express IP indemnity. The carve-outs in those clauses (combinations with items the vendor did not supply, customer modifications, compliance with customer specifications) line up almost exactly with indirect-infringement theories, which is why tender letters and coverage fights follow every customer suit. See Indemnification and Limitation of Liability: The Risk Allocation Engine of Every Contract and Cloud and SaaS Agreements: Service Levels, Data Rights, Security, and Exit.

Notice letters: creating knowledge without inviting a lawsuit

A notice letter does three things at once. It supplies the knowledge that § 271(b) and (c) require and starts the willfulness clock. If it identifies the patent and charges a specific product, it satisfies the actual-notice requirement of § 287(a), provided it comes from the patentee or someone acting for it. Lans v. Digital Equipment Corp., 252 F.3d 1320 (Fed. Cir. 2001). And it may hand the recipient a declaratory judgment action in a forum of its choosing. After MedImmune, Inc. v. Genentech, Inc., 549 U.S. 118 (2007), the question is whether all the circumstances show a substantial, immediate controversy. A licensing presentation that walked through an infringement analysis was enough in SanDisk Corp. v. STMicroelectronics, Inc., 480 F.3d 1372 (Fed. Cir. 2007). Two short letters from a patent-holding company that identified a patent "relating to" HP's blade servers, imposed two-week deadlines, and asked HP not to sue were enough in Hewlett-Packard Co. v. Acceleron LLC, 587 F.3d 1358 (Fed. Cir. 2009). Letters can also supply personal jurisdiction over the patentee in the recipient's home forum when the communications are sufficiently numerous and pointed. Trimble Inc. v. PerDiemCo LLC, 997 F.3d 1147 (Fed. Cir. 2021).

Patentees also face limits on what letters may say. Many states have bad-faith patent assertion statutes (Vermont's, 9 V.S.A. §§ 4195 to 4199, was the first), though federal law protects good-faith infringement allegations, and a patentee's communications cannot be enjoined absent a showing that its claims are objectively baseless. Myco Industries, Inc. v. BlephEx, LLC, 955 F.3d 1 (Fed. Cir. 2020). The recipient's choices (respond, ask for claim charts, obtain an opinion, design around, file for inter partes review, or sue) are sequencing questions. Filing a civil action challenging validity before petitioning for inter partes review bars the petition under 35 U.S.C. § 315(a)(1), and the petition must be filed within one year of service of an infringement complaint under § 315(b). For the patentee's side of the decision, see Do You Have to Enforce Your Patent? Duty, Demand Letters, and the Cost of Doing Nothing, and for the recipient's, Declaratory Judgment Actions: Jurisdiction, Discretion, and the Race to the Courthouse and Inter Partes Review Strategy: Discretionary Denial, Estoppel, and the Parallel District Court Case.

Defenses that work, and three that usually do not

Accused indirect infringers have more defenses than direct infringers, because every element the patentee must prove is also an opening. In rough order of how often they decide cases:

1. No direct infringement. If no one practiced every claim element (or every step, attributed to one actor), there is nothing to induce or contribute to. Aro I, Limelight, ACCO, and the Commil remand all ended this way.
2. No knowledge, or not yet. Liability begins with knowledge of the patent and of infringement. A defendant that first learned of the patent from the complaint may have no pre-suit exposure at all, and in some courts no viable indirect claim in that complaint.
3. Good-faith belief in noninfringement. A reasonable, honestly held claim reading negates the intent element (Commil, DSU). Its strength depends on when it was formed, whether it was based on complete facts, and whether the company acted consistently with it.
4. Substantial noninfringing use (contributory claims only). The defendant should frame the "component" at the level the cases use (Ricoh, i4i, Lucent: the separable feature, not the product) and then prove real, measured noninfringing uses (Vita-Mix, Toshiba). The staple-article rule is old enough that the Supreme Court borrowed it for copyright in Sony Corp. of America v. Universal City Studios, Inc., 464 U.S. 417 (1984), and when the Court later imported patent law's inducement rule into copyright in Metro-Goldwyn-Mayer Studios Inc. v. Grokster, Ltd., 545 U.S. 913 (2005), it explained why the rule does not protect inducers: purposeful encouragement overcomes the protection a staple product otherwise enjoys.
5. Exhaustion and implied license. An authorized sale exhausts the patent in the item sold, including method claims the item substantially embodies, Quanta Computer, Inc. v. LG Electronics, Inc., 553 U.S. 617 (2008), whether the sale occurred here or abroad, Impression Products, Inc. v. Lexmark International, Inc., 581 U.S. 360 (2017). Because Keurig's sale of its brewers exhausted its method claims for brewing, a seller of compatible cartridges could not be liable for inducing or contributing to their use, Keurig, Inc. v. Sturm Foods, Inc., 732 F.3d 1370 (Fed. Cir. 2013), and even meters given away for free exhausted a method patent in LifeScan Scotland, Ltd. v. Shasta Technologies, LLC, 734 F.3d 1361 (Fed. Cir. 2013). Exhaustion has limits: it does not reach third parties practicing separate claims, Helferich Patent Licensing, L.L.C. v. New York Times Co., 778 F.3d 1293 (Fed. Cir. 2015), or the making of new copies of a patented seed, Bowman v. Monsanto Co., 569 U.S. 278 (2013). A patentee that sells a nonstaple device whose only use is to practice its method may also grant buyers an implied license, which leaves no direct infringement for anyone to induce. Met-Coil Systems Corp. v. Korners Unlimited, Inc., 803 F.2d 684 (Fed. Cir. 1986). See Patent Licensing and Exhaustion: Field-of-Use Limits, Covenants Not to Sue, and What the First Sale Ends.
6. Repair, not reconstruction. Owners of a lawfully sold article may repair it, and suppliers of repair parts are not contributory infringers (Aro I). Adapting a machine to a related use was "kin to repair" in Wilbur-Ellis Co. v. Kuther, 377 U.S. 422 (1964); replacing an injection mold and carrier plate was repair in Husky Injection Molding Systems Ltd. v. R & D Tool & Engineering Co., 291 F.3d 780 (Fed. Cir. 2002); refurbishing single-use cameras first sold in the United States was repair in Jazz Photo Corp. v. International Trade Commission, 264 F.3d 1094 (Fed. Cir. 2001). But retipping a spent drill was reconstruction in Sandvik Aktiebolag v. E.J. Co., 121 F.3d 669 (Fed. Cir. 1997). The policy debate is summarized in The Right to Repair Movement: IP Implications for Manufacturers and Consumers.
7. Invalidity. Belief in invalidity does not negate intent, but invalidity proved by clear and convincing evidence under 35 U.S.C. § 282 and Microsoft v. i4i ends the case, and inter partes review applies a preponderance standard. Prior-art strategy is covered in Prior Art Under the America Invents Act: Section 102, the On-Sale Bar, Public Use, and the One-Year Grace Period.
8. Statutory carve-outs. The § 271(e)(1) safe harbor protects uses reasonably related to FDA submissions, including for medical devices, Eli Lilly & Co. v. Medtronic, Inc., 496 U.S. 661 (1990); claims arising from work for the federal government run against the United States under 28 U.S.C. § 1498; and § 271(d) lists conduct (such as collecting revenue from contributory infringers or refusing to license) that cannot be called misuse. Laches no longer bars damages within the six-year window. SCA Hygiene Products Aktiebolag v. First Quality Baby Products, LLC, 580 U.S. 328 (2017).

Three defenses usually fail. "We just sell the product; what customers do is their business" fails when the instructions recommend the infringing use. "Our product has a noninfringing mode" defeats contributory infringement but not inducement (Toshiba, Sanofi). "We believed the patent was invalid" is no answer to intent after Commil. A fourth, "our engineers never read patents," cuts both ways: one district court held that a no-read policy, standing alone, did not plead willful blindness, USC IP Partnership, L.P. v. Facebook, Inc., 2021 WL 3134260 (W.D. Tex. 2021), but combined with copying or a known competitor's product it is exactly the evidence Global-Tech found sufficient.

Two worked examples

A SaaS vendor whose customers perform some of the steps

Return to Juniper Labs. Stillwater Data LLC owns a patent with two relevant claims. Claim 1 is a method: (a) receiving, from a client device, a dataset formatted according to a defined schema; (b) generating, at a server, a risk score by applying a trained model to the dataset; and (c) displaying the risk score with a confidence band on the client device. Claim 12 claims a system of a server and a client device configured to do the same things. Juniper's customer agreement says the platform scores only data submitted through Juniper's API in Juniper's schema. Its onboarding guide walks customers through formatting and upload, and its web app renders the score and confidence band in the customer's browser using code Juniper serves. Stillwater sends a letter with claim charts on March 4, 2024. Juniper's counsel delivers a noninfringement opinion in April 2024, reasoning that "displaying ... on the client device" requires software installed by the client. Stillwater sues in September 2024.

Theory one: Juniper as direct infringer of claim 1 through attribution. Juniper performs step (b) itself. Customers perform step (a), and either Juniper or the customer performs step (c), depending on claim construction. Under Akamai, the customers' steps are Juniper's if Juniper conditions use of the platform on performing them and dictates their manner or timing. The API-and-schema requirement looks like Limelight's tagging requirement, and the onboarding guide looks like Limelight's welcome letter. Because this theory is direct infringement, knowledge is irrelevant, and with only a method claim asserted there is no marking limit, so damages can run back six years from the complaint. Juniper's best responses are claim construction of "displaying" and invalidity, not lack of intent.

Theory two: inducing customers' use of the claim 12 system. Under Centillion and Intellectual Ventures I LLC v. Motorola Mobility LLC, 870 F.3d 1320 (Fed. Cir. 2017), a customer "uses" the claimed system if it puts the system into service, controlling (even indirectly) and benefiting from each claimed component. A customer that uploads data and receives a score plausibly does. Juniper's instructions supply the active steps, and the March 2024 letter supplies knowledge, so inducement damages start then. Juniper's April opinion is its intent defense; if its claim reading was reasonable, it may defeat inducement even if Juniper loses claim construction later, and § 298 keeps the jury from drawing an inference if Juniper chooses not to waive privilege and rely on it.

Theory three: contributory infringement. Weak. Section 271(c) requires a sale, offer, or import of a component or of a material or apparatus for use in practicing a process, and a hosted service fits awkwardly. Even if Juniper's browser code counted, it does many things besides displaying scores.

What each side should do. Stillwater leads with theory one for the pre-notice years and keeps theory two as a fallback, after checking whether it or any licensee sells unmarked products practicing claim 12 (asserting that claim alongside claim 1 would then invite a marking defense under American Medical Systems), and it subpoenas two or three customers for usage records. Juniper changes its onboarding guide and API so that customers may submit unstructured data and choose their own display tools, which weakens conditioning going forward (and reads well on willfulness), files an IPR petition within one year of service, and answers its customers' indemnity tenders under its master agreement. Customers sued separately ask for a stay under the customer-suit exception.

A chipmaker selling a component abroad

Ostara Microdevices, a Taiwanese company, designs power-management chips. A U.S. foundry fabricates the wafers in Oregon; they are shipped to Malaysia for packaging; Ostara sells the packaged chips to contract manufacturers in China, who build laptop chargers for U.S. and European brands. Cobalt Power Systems owns a U.S. patent with apparatus claims to a charger circuit (controller chip, transformer, switch, feedback loop) and method claims to regulating its output. Ostara's application notes include a reference design that is the claimed circuit, its datasheets advertise compliance with U.S. Department of Energy efficiency levels, and its field engineers in San Jose support U.S. brand owners' design teams.

• Direct infringement by Ostara under § 271(a)? Probably not for chips delivered abroad: like the transformers in Halo, they were neither sold nor offered for sale in the United States, even with U.S. price talks and design meetings. Where contracts are negotiated and signed here, Carnegie Mellon makes the location of the sale a fact question.
• Inducement under § 271(b)? This is Cobalt's main theory. The brand owners import and sell infringing chargers here; Ostara's reference design, U.S.-standard marketing, and San Jose support are the Power Integrations (2016) and MEMC fact pattern; and Enplas confirms that the inducing acts may occur abroad. Cobalt needs knowledge, so it sends Ostara a letter identifying the patent and the reference design.
• Contributory infringement under § 271(c)? Ostara does not sell or import in the United States, so § 271(c) mostly falls away, except for evaluation boards its U.S. subsidiary sells here.
• Section 271(f)(2)? The Oregon wafers are arguably components "supplied ... from the United States" (Ostara will answer that dicing and packaging abroad make the packaged chip, not the wafer, the relevant component). If the chip is especially made for the claimed circuit and has no substantial noninfringing use, and Ostara knows it and intends the combination abroad, (f)(2) reaches chargers assembled and sold overseas, and WesternGeco allows foreign lost profits. Ostara's firmware is different: under Microsoft v. AT&T, a master image sent abroad and copied there is not a supplied component.
• Damages. For inducement, the royalty base is chips in chargers imported into the United States (Enplas), which requires tracing data from the brands; for (f)(2), it can include foreign units.

Cobalt might also bring a Section 337 case at the International Trade Commission against the imported chargers; in Suprema, Inc. v. International Trade Commission, 796 F.3d 1338 (Fed. Cir. 2015) (en banc), the Federal Circuit held that the Commission may exclude articles whose infringement is induced after importation under 19 U.S.C. § 1337. See Section 337 Investigations at the ITC: Exclusion Orders, Domestic Industry, and Speed. Ostara's defensive moves are equally concrete: move fabrication to a non-U.S. foundry (which ends future (f)(2) exposure but not past liability), obtain a noninfringement opinion on the reference design, remove the claimed configuration from its application notes after notice, and demand tracing discovery to shrink the U.S. royalty base.

Common mistakes on both sides

Patentees

• Pleading inducement with no pre-suit knowledge facts in a district that follows ZapFraud, and losing the claim (or the pre-suit damages) at the pleading stage.
• Proving that the product can infringe instead of proving that someone did. Capability evidence and the expert's own testing lost ACCO and Mirror Worlds.
• Defining the contributory "component" as the whole product, which invites the noninfringing-use defense, instead of the separable feature (Ricoh, i4i).
• Asserting apparatus claims alongside method claims without auditing marking, and forfeiting the pre-notice years.
• Building a royalty base on worldwide sales of a component when only some of it ended up in U.S. products (Enplas).
• Proposing a creative inducement instruction instead of the model language (Power Integrations).

Accused infringers

• Treating a belief in invalidity as a defense to intent (Commil), or obtaining an opinion only on invalidity.
• Commissioning an opinion on incomplete facts, which can look like willful blindness (Global-Tech).
• Ignoring a notice letter and continuing to publish instructions that walk customers through the accused configuration.
• Filing a declaratory judgment action that challenges validity before deciding whether to petition for inter partes review, which forfeits the petition under § 315(a)(1).
• Assuming a noninfringing mode defeats inducement (Toshiba, Sanofi).
• Letting customers fight alone: a supplier that does not engage early loses control of claim construction and inherits the indemnity bill anyway.

Frequently asked questions

Can a company be liable for inducement if it never knew about the patent? Not for the period before it knew, unless it was willfully blind (it believed there was a high probability a patent covered the product and took deliberate steps not to find out). After it learns of the patent (from a letter or, in many courts, from the complaint), continued encouragement of infringing use can be inducement going forward.

Is an opinion of counsel still worth getting after section 298? Yes. Section 298 stops the patentee from arguing that the absence of an opinion proves intent, but a timely, competent noninfringement opinion is still the best affirmative evidence that the defendant lacked intent. Relying on it at trial waives privilege over the same subject matter, so the decision should be made with trial counsel early.

If my product has a noninfringing mode, am I safe? From contributory infringement, often yes, if the noninfringing use is real and substantial and you can prove it. From inducement, no: if your instructions or marketing steer customers into the infringing mode, the existence of another mode does not help.

Are my customers liable too? Yes. Customers who perform the claimed method or use the claimed system are direct infringers whether or not they knew of the patent. That is why patentees sometimes sue customers and why suppliers respond with indemnity, declaratory judgment actions, and motions to stay under the customer-suit exception.

Two companies each perform half of a method. Can both escape? Often, yes. Arm's-length collaboration is not enough, as the strategic partners in Golden Hour Data Systems, Inc. v. emsCharts, Inc., 614 F.3d 1367 (Fed. Cir. 2010), discovered to the patentee's cost. Liability requires that one actor direct or control the other (including by conditioning a benefit on performance and setting its manner or timing) or that they form a true joint enterprise with an equal right of control.

Can selling a part overseas create U.S. liability? Yes, in three ways: by inducing customers to import or sell infringing products here, by shipping components out of the United States for assembly abroad under § 271(f), or through sales whose legally relevant location turns out to be the United States. Damages for purely foreign sales remain difficult to recover except under § 271(f)(2).

What should I do when a notice letter arrives? Calendar it, preserve documents, and get counsel to read the patent and the claim charts before anyone replies. Decide early whether you want an opinion, a design change, a license discussion, an inter partes review petition, or a declaratory judgment action, because those options interact.

Can executives be sued personally? Yes. Officers who actively assist their company's infringement can be personally liable for inducement under Orthokinetics and Wordtech, subject to the same knowledge and intent requirements.

Practical takeaways

• Indirect liability always rides on someone's direct infringement. Prove or attack that first.
• Knowledge dates control damages. Patentees should create a clean knowledge record before suit; defendants should document what they knew and when.
• Instructions are the battlefield. A help article that describes an option is different from onboarding materials that require it.
• Pick the right theory. Attribution under Akamai is strict liability and reaches pre-notice years; inducement requires intent and starts at knowledge.
• Frame the component at the feature level for § 271(c), and prove the noninfringing uses with data, not adjectives.
• Watch the border. Foreign conduct can induce U.S. infringement, but foreign sales rarely produce U.S. damages outside § 271(f)(2).
• Draft claims for a single actor, and draft contracts (indemnities, terms of service, onboarding requirements) knowing they will become exhibits.

This document is general information, not legal advice for any particular situation; the law changes, and the facts matter.

Related Documents

Related Articles

• What Constitutes Patent Infringement: Claims and Defenses: A Practical Guide: the elements of direct infringement that every indirect theory borrows.
• Claim Construction and Markman Hearings: How Courts Decide What a Patent Actually Covers: how one construction can turn a single-actor claim into a divided one.
• Patent Damages: Reasonable Royalties, Lost Profits, Apportionment, and Enhancement: the damages framework applied to indirect infringers.
• Patent Licensing and Exhaustion: Field-of-Use Limits, Covenants Not to Sue, and What the First Sale Ends: exhaustion and implied license as defenses to indirect claims.
• Hatch-Waxman Litigation: The Orange Book, Paragraph IV Certifications, and the 30-Month Stay: the regulatory setting for the skinny-label cases.
• Declaratory Judgment Actions: Jurisdiction, Discretion, and the Race to the Courthouse: what a notice letter can set in motion.
• Trademark Liability Online: Keyword Advertising, Marketplaces, and Contributory Infringement After Tiffany v. eBay: how secondary liability works for marks, a useful comparison.
• Prior Art Under the America Invents Act: Section 102, the On-Sale Bar, Public Use, and the One-Year Grace Period: building the invalidity defense that intent cannot supply.
• Post-Trial Motions in Federal Court: Judgment as a Matter of Law, New Trial, Remittitur, and Relief from Judgment: where intent and direct-infringement verdicts are tested.

Related Guides

• Pleading and Proving Induced and Contributory Patent Infringement: A Practical Guide: the step-by-step companion to this article.
• Proving and Attacking Patent Damages: A Practical Guide: building and challenging the royalty case.
• Motions to Dismiss Under Rule 12: A Practical Guide for Plaintiffs and Defendants: the motion that tests indirect-infringement pleadings.

Related Checklists

• Indirect Patent Infringement Claim Checklist: A Practical Checklist: element-by-element review for either side.
• Freedom-to-Operate Search: A Practical Checklist: the clearance work that shapes knowledge and good faith.

Related Toolkits

• Indirect Infringement Toolkit: Knowledge Evidence, Direction or Control Proof, and Substantial Noninfringing Use Analyses: templates for the evidence this article describes.
• Patent Litigation Toolkit: A Roadmap and Research Guide: the broader litigation workflow.
• Patent Damages Toolkit: Royalty Models, Expert Reports, and Daubert Motions: models for the royalty base and apportionment.