Patent infringement is making, using, offering to sell, selling or importing a patented invention in the United States, without authorisation, while the patent is in force. That is the whole of 35 U.S.C. 271(a), and every additional question about infringement is a question about how those words apply.
Two things surprise people who have not dealt with it before. Direct infringement requires no knowledge of the patent — you can infringe one you have never heard of. And independent development is no defence at all.
What the statute covers
| Section | Conduct | Knowledge required |
|---|---|---|
| 271(a) | Making, using, offering to sell, selling, importing | No |
| 271(b) | Actively inducing another to infringe | Yes — of the patent and the infringement |
| 271(c) | Supplying a component especially adapted for infringing use | Yes |
| 271(f) | Supplying components from the US for assembly abroad | Yes |
| 271(g) | Importing a product made abroad by a patented process | No |
Direct infringement under 271(a) is strict liability. No intent, no knowledge, no copying. A company that developed the same solution independently, in a clean room, without ever seeing the patent, still infringes.
This is the single largest difference from copyright, where independent creation is a complete defence. It is also why freedom-to-operate work exists at all — if independent development protected you, nobody would need to search.
Each act is separately infringing. Making, using and selling are three distinct acts, so a manufacturer, a distributor and an end user can each infringe the same patent through the same product.
Proving infringement: the all-elements rule
An accused product infringes a claim only if it contains every element of that claim. Miss one limitation and there is no infringement of that claim, however similar the product otherwise is.
This is the all-elements rule, and it does more work than any other principle in infringement analysis.
Claims are compared, not products. The comparison is between the language of the claim and the accused product — never between the patentee's product and the accused product. Two competing products that look identical may not infringe if the claim covers something neither of them does.
Independent claims first. If an independent claim is not infringed, the dependent claims that reference it cannot be either, because a dependent claim includes every limitation of its parent plus more.
The output is a claim chart — a table with each claim limitation on one side and the corresponding feature of the accused product on the other. It is the standard work product for infringement analysis and the first thing counsel on either side will ask for.
Literal infringement
Every element of the claim is present in the accused product exactly as claimed. The straightforward case, and the one that produces clean answers.
The doctrine of equivalents
An element is present in substantially the same way even though it does not literally meet the claim language. The test asks whether the difference is insubstantial, often framed as whether the accused element performs substantially the same function, in substantially the same way, to achieve substantially the same result.
It is limited by prosecution history estoppel. Claim language narrowed during prosecution to overcome prior art generally cannot be recaptured through equivalents. This is why the file history matters so much in infringement analysis — see doctrine of equivalents for how the two interact.
Indirect infringement
Somebody else does the infringing and you are liable for it. Both forms require an underlying act of direct infringement by someone, and both require knowledge of the patent.
Induced infringement, 271(b)
Actively encouraging another to infringe, with knowledge of the patent and intent that the acts constitute infringement.
The classic fact pattern is instructions. A supplier sells a component that could be used lawfully or unlawfully, and provides a manual telling customers to use it the infringing way. The sale alone is not inducement; the instruction is.
Willful blindness suffices. After Global-Tech v. SEB, a defendant who deliberately avoids learning about a patent cannot rely on that avoidance. This matters for companies with a policy of not reading competitors' patents — the policy does not create the protection people assume.
Contributory infringement, 271(c)
Supplying a component that is especially made for infringing use, knowing it is so adapted, where the component is not a staple article of commerce suitable for substantial non-infringing use.
The staple article exception does the work. A generic screw sold to someone who builds an infringing device is not contributory infringement. A part that only functions within the patented combination is.
What damages are available
| Type | Basis | When it applies |
|---|---|---|
| Reasonable royalty | 35 U.S.C. 284 | Always — the statutory floor |
| Lost profits | Panduit factors | Where the patentee competes and can prove diversion |
| Enhanced damages | 284, up to 3× | Willful infringement, at the court's discretion |
| Attorney fees | 285 | Exceptional cases |
| Injunction | 283 | At equity, after eBay v. MercExchange |
A reasonable royalty is the floor, not the expectation. It is derived from a hypothetical negotiation between a willing licensor and willing licensee at the time infringement began, informed by the fifteen Georgia-Pacific factors.
Lost profits require the Panduit factors: demand for the patented product, absence of acceptable non-infringing substitutes, capacity to have made the sales, and a quantifiable profit that would have been made. A patentee who does not sell a competing product cannot satisfy these, which is why non-practising entities recover royalties rather than lost profits.
Apportionment constrains both. Where the patent covers a component of a larger product, damages must be apportioned to the patented feature rather than the whole. The smallest saleable patent-practising unit is the usual starting point, and the entire market value rule is available only where the patented feature drives demand for the whole product.
The two limits that surprise patentees
The six-year look-back, 35 U.S.C. 286. No recovery for infringement occurring more than six years before the complaint. This is an absolute cap on damages, not a limitations period — a claim over twenty-year-old conduct is not time-barred, it simply recovers nothing for the first fourteen years.
The marking requirement, 35 U.S.C. 287. A patentee who sells patented articles must mark them with the patent number — or with a public webpage listing it, under virtual marking — to recover for the period before actual notice. Failure to mark generally limits damages to infringement occurring after notice, which is often the date the complaint was served.
Marking failures are expensive and common. A patentee who has sold unmarked product for five years and then sues may find the entire pre-suit period unrecoverable.
A worked example: what the damages actually look like
A patent covering a component of a machine. The infringer sold 8,000 units over four years at $12,000 each. The component represents roughly 8% of the machine's value by any reasonable apportionment.
| Scenario | Basis | Illustrative calculation |
|---|---|---|
| Reasonable royalty | 4% of apportioned base | 8,000 × $12,000 × 8% × 4% = $307,200 |
| Entire market value | 4% of full price — rarely allowed | 8,000 × $12,000 × 4% = $3,840,000 |
| Lost profits | If Panduit satisfied on 2,000 units | 2,000 × patentee's margin |
| Willfulness enhancement | Up to 3× the royalty | Up to $921,600 |
The gap between the first two rows is the entire apportionment fight, and it is where most patent damages disputes actually live. A twelvefold difference in the award turns on whether the patented feature drove demand for the machine or merely contributed to it.
Note what marking would do to this. If the patentee sold unmarked product and gave notice only by filing suit, the four-year period collapses to however long the case takes, and the royalty falls proportionately.
Willful infringement
Deliberate infringement, or reckless disregard of a known patent, exposes the infringer to damages of up to three times the compensatory award.
Since Halo v. Pulse (2016) the standard is discretionary rather than the rigid two-part test that preceded it. Courts look at the conduct as a whole, and enhancement is reserved for behaviour that is egregious rather than merely losing.
An opinion of counsel remains the main rebuttal. A competent, written, pre-conduct opinion that the patent is not infringed or is invalid tends to defeat willfulness — but 35 U.S.C. 298 provides that failure to obtain one cannot be used to prove willfulness. The opinion helps; its absence is not supposed to hurt.
Timing is everything. An opinion obtained after the complaint arrives is worth very little. One obtained when the product was designed is worth a great deal, which is the practical argument for freedom-to-operate work before launch rather than after a letter.
When a demand letter arrives
Do not reply immediately, and do not concede anything. Early correspondence becomes evidence, and a reply written in the first hour is rarely the one you would write in the first week.
The sequence that works:
Get the patent and read the claims. Not the abstract, not the title — the numbered claims at the end. That is what you are accused of infringing.
Check whether it is in force. Compare recorded maintenance fee payments in USPTO Patent Center against the grant date. A meaningful number of demand letters are sent on patents that have lapsed, and the fastest resolution is sometimes the cheapest one.
Have the claims charted against your product. Element by element. If one limitation is genuinely absent, there is no infringement of that claim.
Read the file history. Claim language narrowed during prosecution constrains what the doctrine of equivalents can recapture, and the file history frequently contains the argument that defeats the assertion.
Assess validity. Prior art that reads on the claims may support an inter partes review petition, though the one-year bar under 35 U.S.C. 315(b) starts running from service of a complaint.
Then choose. Non-infringement, invalidity, licence, design-around, or wait. Each has a cost and a timeline, and the choice is much better made with a chart in hand than without one.
Who is asserting
Non-practising entities file a majority of US patent cases. They do not make products, so they cannot claim lost profits and litigate for royalties. Their economics depend on settlement below the cost of defence, which shapes how the early stages of those cases run.
Operating companies assert differently. Competitor suits are more likely to seek injunctions, more likely to involve counterclaims on the defendant's own patents, and more likely to settle in a cross-licence than a payment.
Knowing which you are facing changes the response. A demand from a company that competes with you is a different problem from one seeking a five-figure licence, and the first thing to establish about an assertion is what the asserter actually wants.
Before you respond to an assertion
- Read the asserted claims, not the patent title.
- Check maintenance fee status — the patent may have lapsed.
- Chart the claims against your product element by element.
- Read the file history for narrowing amendments and arguments.
- Diarise the one-year IPR bar from the date of service if a complaint has been filed.
- Establish what the asserter wants — a licence, an injunction, or a settlement below defence cost.
- Get counsel before replying in substance. The letter you send becomes evidence.