A Patent Dispute Almost Never Ends in Court

The first time a friend forwarded me a patent demand letter, I did what most engineers would do — I pictured the courtroom. Jury box. Stern judge. Two rows of lawyers in matching suits. Months of trial. A verdict announced with cameras outside.

That picture turned out to be almost entirely wrong.

Roughly 95–97% of US patent infringement suits settle before they ever reach a verdict. The cases people read about in the news — the ones with billion-dollar awards and dramatic juries — are the rare 5% that didn't. Almost everything else gets resolved somewhere most people don't think to look: in a demand letter, at the USPTO's review board, at the International Trade Commission, or in a hotel conference room three weeks before trial.

Across the first four posts in this series I've walked through how the patent system grants rights — what a patent is, when to file, how to read one, and how the international layer works. This post is about the part nobody explains until you're already inside it: how those rights actually get fought over, and where the fights actually get decided.

The thesis I want to argue is this: a patent dispute is a negotiation system with the courthouse as a backstop. The forums that decide outcomes — demand letters, PTAB review, ITC exclusion orders, settlement on the courthouse steps — run on different timelines, costs, and burdens of proof than a jury trial. Choosing the right forum is usually more consequential than choosing whether to enforce at all.

If that's true, then the question worth answering isn't "how do I win at trial." It's "how do I never get there in the first place."

Almost every patent fight starts with a letter, not a lawsuit

The first move in a patent dispute is rarely a complaint filed in federal court. It's a letter — sometimes called a cease-and-desist or a notice of infringement — that says, in slightly more legalistic language, "You're using our patent. Either pay us or stop."

What makes this letter so consequential is the cost asymmetry behind it. According to the AIPLA's 2023 Economic Survey, here's what it actually costs to defend a patent suit through trial and appeal:

So a recipient doing simple math sees that even a small case — under a million dollars at stake — costs six hundred thousand dollars just to defend. That's before counting management distraction, discovery, expert witnesses, and the risk of losing. For most companies, even a strong legal position is worth less than the cost of proving it.

That math is what the sender of the letter is counting on. The letter isn't a prelude to litigation. It's an offer to settle the litigation that hasn't happened yet. The asymmetry is the leverage.

The cleanest evidence that the system runs on this logic is the settlement rate itself. The Marcum 2024 Patent Litigation Study and Lex Machina's tracking both put it at 95–97% — patent cases that file, then settle, then disappear from the public record. The other 3–5% are the ones that argue all the way to a verdict, which is why those are the ones that make headlines.

I think this is the part that took me longest to internalize. The demand letter isn't aggressive lawyering before things get serious. It is the serious part. What happens in the first 30 days after that letter lands — whether the recipient ignores it, replies with prior art, files a preemptive PTAB challenge, or just opens a settlement conversation — usually decides the case.

The four real forums (and what each one optimizes for)

When people picture "patent dispute" they picture federal district court. That's one of four real forums, and for most modern software disputes it's not even the most important one.

Each forum optimizes for something different, and the choice of where to fight is often more decisive than the underlying merits. A quick tour:

Federal district court is the version everyone pictures. A plaintiff sues for infringement, asks for damages and sometimes an injunction, and the case moves through discovery, motions, claim construction, summary judgment, and eventually trial. It's slow (2–3 years to trial), expensive ($2–5M to defend a mid-sized case), and the only place that hands out the headline numbers like Apple v. Samsung's billion-dollar verdicts. It's also the only forum where a jury actually decides the facts.

The Patent Trial and Appeal Board (PTAB), inside the USPTO, runs a procedure called Inter Partes Review — IPR. It's not a court case. It's a re-examination of whether the patent should have been granted in the first place. A challenger files a petition arguing the claims are invalid based on prior art. The PTAB decides whether to institute the review, and if it does, the patent owner has to defend the patent in front of three administrative patent judges who do patent law all day. The whole thing wraps in about 18 months, total cost is $350K to $600K, and in FY2024 the PTAB instituted about 68% of petitions and cancelled at least some claims in roughly 35–45% of final decisions, per the PTAB Litigation Blog's FY2024 roundup.

The PTAB is the most consequential forum most engineers have never heard of. It's the place where weak patents go to die.

The International Trade Commission (ITC), under Section 337, is the third forum and the one I find structurally most interesting. It can't award money damages. What it can do is issue an exclusion order — an import ban that customs enforces at the border. For any product manufactured outside the US (which is most of them), an ITC exclusion order is closer to a kill switch than a fine. The ITC averages 15–18 months from complaint to final determination, much faster than district court, and according to Sterne Kessler's 2025 year-in-review the agency issues 50–70 new Section 337 investigations a year.

Private arbitration and settlement, finally, is where the actual outcomes get written. Either as part of a court-supervised mediation, or as a private deal negotiated under the implicit threat of one of the other three forums.

The mental model I now use: the first three forums aren't really where disputes get resolved. They're leverage instruments that change the terms of the eventual settlement. Filing a PTAB petition lowers the patent owner's expected value. Filing at the ITC adds the threat of an import ban. Filing in district court adds the threat of a billion-dollar verdict. The settlement that follows reflects whichever instrument got pulled.

Why venue changes outcomes more than evidence does

If you decide the dispute belongs in federal district court, the next question is which district. That choice turns out to be surprisingly consequential.

Per the Lex Machina 2025 Patent Litigation Report, the Eastern District of Texas reclaimed its position as the leading US patent venue in 2024, with more than 1,000 new patent lawsuits — more than double the next-closest district (the Northern District of Illinois). The District of Delaware sits in steady second among the traditional patent venues. The Western District of Texas was the surprise newcomer during the Judge Alan Albright era — at his peak Judge Albright single-handedly handled roughly 25% of all US patent cases, until a July 2022 reassignment order randomized W.D. Texas cases across all judges in the district.

Two Supreme Court and administrative decisions reshaped this picture in the last decade. TC Heartland (2017) tightened where patent suits can be filed against corporate defendants — they basically must be sued where they're incorporated or where they have a regular place of business. That decision sent a flood of cases to Delaware (where most US companies are incorporated) and W.D. Texas (which covers Austin and a lot of tech offices). The 2022 W.D. Texas randomization, then, was a course correction that pulled cases back toward E.D. Texas and Delaware.

Why does venue matter so much? Three reasons that compound:

• Time to trial. E.D. Texas and W.D. Texas both target around 2.4 years. Delaware runs closer to 2.5–3. N.D. California can run longer. Time is leverage — a faster trial means a defendant pays settlement money sooner to avoid the trial expense.
• Jury composition. Different districts pull from different communities, which affects how juries interpret technical evidence and what they're willing to award. Plaintiff lawyers know this intimately.
• Judge specialization. Delaware has a deep bench of judges who handle patent cases routinely. E.D. Texas judges have decades of patent experience. A judge who's seen 200 patent cases will rule on technical motions differently than one who's seen 5.

This is the part most engineers find frustrating about the system — and reasonably so. The merits of a patent case should matter more than the zip code where it's filed. In practice the zip code usually matters as much.

A useful way to think about it: forum choice is the first substantive decision in a patent case, and it's made before any evidence is presented. The plaintiff picks the district. The defendant can try to move it. By the time discovery starts, half the case has already been shaped.

How the process actually unfolds

Once a case is in federal district court, the procedural arc is roughly the same everywhere, even if the timeline differs by district. Knowing the shape helps because each phase is also a settlement window.

A few phases deserve a closer look because they bend outcomes more than people realize.

The Markman hearing — named after the 1996 Supreme Court case that put it on the map — is where the judge decides what the words in the patent's claims actually mean. The patent owner argues for a broad interpretation that catches the defendant's product. The defendant argues for a narrow interpretation that misses it. Whoever wins the Markman often wins the case, because the claim construction propagates through every later motion and through the jury instructions. Many cases settle within weeks of the Markman ruling because both sides suddenly have much better information about the verdict probability.

Discovery is the longest and most expensive phase. The defendant turns over source code, internal emails, design docs, financial data — anything relevant to infringement or damages. The plaintiff turns over its licensing history and any prior assertions. A typical software case generates millions of documents and weeks of depositions. This is where most of the AIPLA cost numbers above get spent.

Summary judgment is where one side asks the judge to rule without a trial because the undisputed facts only support one outcome. Patent cases get decided on summary judgment more often than people think, especially for software where the question of whether a particular feature reads on a particular claim is often resolvable on paper.

And then there's the loop that runs alongside all of this: PTAB review. Defendants routinely file an IPR petition in parallel with the district court case, then ask the judge to stay the court case until the PTAB rules. If the PTAB cancels the claims, the court case is effectively over. If the PTAB upholds them, the court case continues — but the patent owner now has a stronger position. This is why PTAB filings have become as common as they have: they're a relatively cheap insurance policy that can collapse the entire dispute.

VirnetX v. Apple is the canonical example of how messy this parallel track can get. VirnetX, a non-practicing entity, secured multiple large damages awards against Apple in E.D. Texas. Apple fought the cases for more than a decade and used the PTAB to invalidate some of VirnetX's claims along the way, even while losing rounds in district court. The whole saga is the cleanest illustration of how district court wins and PTAB invalidations can run side by side, sometimes for years, on the same patents.

The thing that kept getting clearer to me as I read through how cases actually move is that the 95–97% settlement rate isn't because lawyers are uninterested in trial. It's because every phase of the process produces new information — about the strength of the claims, the strength of the defenses, the appetite of the judge, the budget of the other side — and as that information accumulates, the range of plausible settlement values narrows until both sides can agree on a number.

There's a single statistic that makes this concrete. Per the Lex Machina 2025 Patent Litigation Report, 68% of patent settlements happen in the window between the Markman hearing and summary judgment. That's the exact moment in the case when both sides have enough information to predict the trial outcome but haven't yet committed to the full trial expense. The 68% isn't lawyers giving up — it's lawyers cashing in their option once the information is in.

A patent case isn't a binary win-or-lose event. It's a 24-month-long price discovery process for a settlement.

What "winning" actually looks like (and why injunctions are rare now)

Even on the rare path that does reach a verdict, what the winner gets has shifted significantly in the last twenty years. The Supreme Court's decision in eBay v. MercExchange (2006) is the reason.

Before eBay, a patent owner who won at trial almost always got a permanent injunction — a court order forcing the defendant to stop selling the infringing product. After eBay, courts have to apply a four-factor test:

The four factors sound abstract, but in practice they map directly onto whether the patent owner practices the patent. A company that actually makes the product can usually argue it suffers irreparable injury when a competitor copies it. A non-practicing entity — a company that owns patents but doesn't sell anything — usually can't, because money damages can fully compensate someone who isn't losing market share.

The result, per Patently-O's empirical review of post-eBay grants, is a stark divergence: practicing entities still get permanent injunctions in roughly 70–80% of wins. NPEs get them in under 10–15%. And the absolute number of permanent injunctions in patent cases has fallen sharply — from about 80 between 2008–2012 to 36 between 2018–2022, per the Marcum 2024 Patent Litigation Study, working out to roughly 7 a year recently.

Read that last number again. Across the entire US patent system, fewer than ten permanent injunctions per year.

What plaintiffs get instead is damages. The Marcum 2024 study puts the overall median damages award (2003–2022) at $3.7 million, rising to $5.6 million if you exclude default judgments. NPEs win awards two to three times larger than practicing entities on average — which makes sense once you understand the injunction divergence: an NPE that can't get an injunction has to extract its entire value from the damages number, while a practicing entity that gets an injunction can extract value from the future market exclusion.

The reframe that helped me here: an injunction is a fence the court builds around the defendant's behavior. A damages award is a toll the court charges. After eBay, software patent litigation is mostly about tolls, not fences. The defendant pays and keeps building.

Even the headline tolls can dissolve on appeal. Centripetal Networks v. Cisco produced a $2.75 billion verdict in 2020 — one of the largest patent infringement awards ever — and was then reversed in 2022 after it came out that the trial judge's wife owned Cisco stock. Per Reuters' August 2023 reporting on the reversal, the case got sent back for a new trial in front of a different judge. The lesson worth carrying out of that one: even a trial verdict is provisional until the Federal Circuit signs off. The actual settlement number sits behind that uncertainty, not in front of it.

The ITC is the cross-border surprise

For any product that's manufactured outside the US — which is basically every consumer electronics product, most software-hardware bundles, most medical devices — there's a forum that punches above its weight: the International Trade Commission.

The ITC's Section 337 jurisdiction is narrow but powerful: it can investigate whether imported products infringe US patents, and if they do, it can issue an exclusion order that bars those products from entering the country. Customs enforces the order at the port.

The cleanest recent example is Apple v. Masimo. Masimo, a medical-device company, sued Apple at the ITC over the blood-oxygen sensor in the Apple Watch. The ITC found infringement and issued an exclusion order. Apple's response — per Reuters' April 2026 reporting on the follow-up case — was to disable the blood-oxygen feature via a software update on US-bound watches, keeping the rest of the product on shelves. A money-damages judgment would have been a check Apple could write without thinking. The exclusion order forced a product change.

What makes the ITC structurally interesting is the parallel-filing strategy. Plaintiffs often file at the ITC and in district court simultaneously. The ITC moves faster and threatens the import ban; the district court action threatens money damages. The combined leverage is much greater than either alone. District court cases routinely get stayed pending the ITC decision, because the ITC's claim construction often carries forward.

For a Vietnamese founder building a hardware product that ships to the US, this is the threat to take seriously, and it isn't hypothetical. In April 2024 the ITC instituted a Section 337 investigation against VinFast Auto — Vietnam's flagship EV maker — on a complaint by ArcelorMittal alleging that the high-strength aluminum-coated steel in the VF 3, VF 6, VF 7, VF 8, and VF 9 infringes ArcelorMittal patents. ArcelorMittal filed a parallel district-court action in C.D. Cal. (2:24-cv-03128) — the exact textbook ITC + district court combo. ArcelorMittal had reached out to VinFast about the alleged infringement in November 2022; no resolution came, so the ITC complaint followed. The cars are designed and assembled in Vietnam. The fight is in Washington and Los Angeles. That gap — where the product is built versus where the patent is fought over — is the entire reason §337 matters for Vietnamese builders. A US patent owner with a Section 337 complaint can stop a shipment at Long Beach. There is no jury trial to wait for. There is no settlement check that makes the import resume. The order is the order.

Vietnam: where disputes live almost entirely outside the courtroom

If the US system is roughly "courthouse is the backstop, settlement is the norm," the Vietnamese system runs almost entirely in the administrative lane. The courtroom barely exists for patents.

The numbers tell the story. According to DNLAW's February 2026 review of Vietnam's IP litigation system, from 2022 through May 2025 Vietnamese courts handled only 17 IP-related cases total across the IP Court and the Commercial Courts in Hanoi and HCMC. Of those, patent infringement cases were 0–2 per year. The Inspectorate of the Ministry of Science and Technology (Thanh tra Khoa học và Công nghệ), meanwhile, handled 3,049 IP enforcement cases in 2024 alone.

It's worth contrasting that with how much patent activity is happening. Per the IP Vietnam 2024 Annual Report, Vietnam received 9,904 invention applications in 2024 — the highest ever — and granted 4,430 patents, a 21% year-over-year increase. So the system is filing patents at record speed, while enforcing them through courts at near-zero. The gap between those two numbers is the entire story of §7.

So for a builder in Vietnam, the practical question isn't "which Vietnamese court should I sue in?" It's "should I file with Thanh tra?" The administrative path is faster, cheaper, and supported by a body that handles thousands of these cases a year. The maximum fine for patent infringement runs to 250 million VND for an individual or 500 million VND for an organization — roughly $10,000 to $20,000 USD — which is small by US standards but meaningful in context.

Why is patent litigation so rare in Vietnam? Per DNLAW and per Tilleke & Gibbins' commentary, four reasons stack:

• Complexity and cost. There are no specialized patent judges. Technical complexity makes outcomes unpredictable.
• Administrative efficiency. Thanh tra is faster and cheaper, and most right-holders prefer it.
• Cultural preference for settlement. Civil disputes generally favor negotiation over public court battles.
• Burden of proof. The technical evidentiary standards are high, and courts aren't yet equipped for deep patent analysis.

The 2022 amendments to Vietnam's Law on Intellectual Property tried to address some of this. They strengthened pre-trial mediation, clarified provisional measures (easier preliminary injunctions), broadened the definition of infringement, and aligned pharmaceutical patent rules with the CPTPP and EVFTA trade agreements. Thomas Treutler at Tilleke & Gibbins has described the post-2022 trajectory as "slowly professionalizing" the civil court pathway — a gradual shift, not a flip.

The cleanest counter-example is Merck Sharp & Dohme v. Dat Vi Phu Pharmaceutical (2021–2024), a rare four-year patent infringement battle in Vietnamese courts over an anti-diabetic drug. The case stands out precisely because it's rare. It signals that multinational patent owners are increasingly willing to use the Vietnamese court system for high-value patents — but the system isn't yet built for volume.

For a Vietnamese founder thinking through enforcement strategy, the practical playbook is roughly:

• Default to administrative enforcement through Thanh tra. It's the dominant mechanism for a reason.
• Reserve the courts for high-stakes cases where the administrative fine cap is far below the actual damages.
• Use customs enforcement sparingly for patents — it's more of a trademark tool — but it can work for obvious cases.
• For cross-border products, the leverage usually isn't in Vietnamese courts at all. It's in the US ITC, EU national courts, or wherever the product is actually sold.

The asymmetry is worth naming directly: a Vietnamese patent is a useful signal for fundraising, licensing, and acquisition value. It is not, today, a particularly sharp litigation weapon. That changes what the patent is for.

The asymmetric play for small operators is defensive, not offensive

Most of this post has talked about the offensive side — what happens when you enforce a patent. The reality for most software builders is that they're more likely to be on the receiving end of a demand letter than the sending end. The interesting strategic question, then, is: what cheap defensive moves are worth making before you ever get one?

A few stand out.

LOT Network is a defensive alliance with a clever structural design. Members agree that if any of their patents are ever sold to a Patent Assertion Entity, every other member automatically gets a free license to that patent. The effect is to make members' patents permanently useless to NPEs, which removes the incentive for NPEs to buy them in the first place. Per the LOT Network 2025 Report, membership has grown to over 6,000 companies (up from 4,500 in late 2023), including Google, Meta, Amazon, Microsoft, and a long tail of startups. Membership is often free for small startups and scales with company size for larger ones.

Unified Patents operates on a deterrence model: it files IPR petitions against low-quality patents that are most likely to be asserted against its members, often invalidating them before they can become a problem. Membership ranges from free for small startups to roughly $10K–$50K for mid-sized companies, depending on revenue.

RPX plays a different role — a clearinghouse that buys patents off the market before NPEs can acquire them, providing members with a defensive shield. Membership runs more like insurance premiums.

Cloudflare's Project Jengo is the public counterexample for what an aggressive defense can look like. When patent troll Blackbird Technologies sued Cloudflare in 2017, Cloudflare — under then-GC Doug Kramer, who's written the whole multi-year saga publicly on the Cloudflare blog — ran a crowdsourced prior-art bounty program, paying out hundreds of thousands of dollars to people who found references that could invalidate Blackbird's patents. CEO Matthew Prince framed the stakes plainly in an Ars Technica interview at the time:

"Patent trolls create what we consider to be an unfair, unjust, and inefficient system that throttles innovation and threatens emerging companies."

— Matthew Prince, CEO of Cloudflare

The October 2024 Sable Networks update is the most recent chapter: a February 2024 Waco jury found Cloudflare did not infringe, and Sable subsequently dedicated all its patents to the public domain and paid Cloudflare to end the litigation. Two NPEs, two decisive defensive wins, both driven by aggressive prior-art collection rather than just settling. Blackbird Technologies, per Cloudflare's own Winning the Blackbird Battle post, went out of business shortly after.

The pattern across all four options is the same: defensive moves are cheaper than offensive ones, and the system rewards builders who think about them before they need them. By the time the demand letter arrives, the price of any defensive option has gone up.

What kept getting clearer to me as I read through how Cloudflare handled Blackbird and then Sable is that the system isn't fully captured by NPEs. Operating companies can push back. Most won't have Cloudflare's budget — but a defensive aggregator membership plus a basic prior-art habit will catch most of the low-quality assertions that fund the NPE business model.

Counterargument: "If everything settles, isn't the system mostly theater?"

A reasonable reader at this point might push back. If 95–97% of cases settle, if injunctions are rare, if forum shopping decides cases before evidence does, if the same patents get litigated multiple times — isn't this whole apparatus mostly performance? Wouldn't it be cheaper to skip the courthouse entirely and just let everyone negotiate?

I think this objection gets one thing right and one thing wrong.

What it gets right is that the headline picture of patent litigation — dramatic verdicts, righteous inventors, decisive trials — is mostly fiction. The system is a negotiation system. Trials are exception cases. Most disputes are resolved by lawyers exchanging numbers via email under the implicit threat of one of the four forums I described.

What the objection misses is that the negotiation only works because the trial backstop exists. Settlement values are negotiated in the shadow of the verdict that would have happened. Remove the backstop and the negotiation collapses, because neither side has a credible bottom line. The 95–97% settlement rate is evidence of the system working as designed, not evidence of failure. It's the equivalent of asking "if most contracts are honored without lawsuits, why have contract law?" The lawsuit-rare equilibrium is the design.

This isn't a new argument. The canonical academic version is George Priest and Benjamin Klein's 1984 paper The Selection of Disputes for Litigation — the third-most-cited paper in the history of the Journal of Legal Studies. Priest and Klein argue that the cases that reach trial are precisely the ones where both sides genuinely disagree about the outcome, because every case where the outcome is predictable settles instead. High settlement rates are a sign the system is functioning, not failing. Marc Galanter's 2004 "Vanishing Trial" study makes the same point across all of US civil litigation: federal civil cases resolved by trial fell from 11.5% in 1962 to 1.8% by 2002. Patent disputes are part of a much larger pattern.

The USPTO itself takes a similar line. In a September 2024 statement announcing the PTAB Director Review final rule, Director Kathi Vidal wrote that the agency was "dedicated to continuing to refine PTAB procedures to make them as fair, transparent, and efficient as possible, for the benefit of all innovators." You can read that as bureaucrat-speak, or you can read it as the agency's view that the system is working — just imperfectly enough to keep tuning.

The more useful critique isn't "the system is theater." It's that the system is structurally tilted in ways that produce predictable outcomes — NPEs win bigger awards because they can't be enjoined, large defendants pay settlements smaller than litigation cost regardless of merit, and venue choice decides cases before evidence is presented. Those are real problems with the system's calibration. They're not problems with the idea of using settlement as the default resolution path.

The Federal Circuit, the court that hears nearly all patent appeals, gives one more data point on calibration. Per Sterne Kessler's 2025 review of 2024 Federal Circuit IP appeals, the Federal Circuit affirmed roughly 73–77% of district court decisions, fully reversed only 7–10%, and vacated or remanded another 10–14%. The system corrects itself most of the time, but it corrects itself slowly and at the appellate level, which doesn't help any individual defendant in the moment.

So the honest synthesis is something like: the system works the way it does because that's the cheapest equilibrium, the equilibrium has real costs, and the people who pay those costs disproportionately are small operators who can't absorb the litigation expense even when they win.

Pulling this together: where disputes actually get decided

The five forums I've walked through — demand letter, district court, PTAB, ITC, settlement — don't sit in a hierarchy. They're parallel rails, and a sophisticated patent strategy uses the threat of multiple rails to shape the outcome on whichever one the case finally lands on.

Looking back across the series, the same lesson keeps showing up in different costumes:

• The primer argued that a patent is a 20-year bargain. The dispute system is how that bargain actually gets priced.
• The mousetrap post argued that a patent is a fence around a mechanism. The dispute system is what determines how high the fence really is.
• The patent-reading post argued that the USPTO database is the most underused competitive-intelligence source available. The PTAB is the most underused defensive lever, for the same reason.
• The international post argued that international patent strategy is a time game. The dispute system is the moment those time decisions cash out — or don't.

The connecting thesis: the patent system has a way it's taught (legal compliance, courtrooms, verdicts) and a way it actually works (negotiation, leverage, forum choice, settlement). Building durable products in this world doesn't require becoming a patent lawyer. It requires knowing which forum applies when, and what each one will cost in time and dollars before a single document is filed.

A few things I'm taking away

• The vast majority of US patent disputes — 95–97% — settle before a verdict; trials are the rare exception, not the design
• A patent dispute begins with a demand letter, and the recipient's first 30 days usually determines the eventual settlement
• The cost asymmetry — six hundred thousand dollars to defend a sub-million-dollar case — is the leverage behind every demand letter
• There are four real forums (district court, PTAB, ITC, settlement), each optimized for different things, and the choice of where to fight is more consequential than the underlying evidence
• Venue choice inside district court — E.D. Texas, Delaware, W.D. Texas, N.D. California — bends outcomes through time-to-trial, jury composition, and judge specialization
• The Markman hearing usually decides the case; most settlements happen in the weeks after it
• The PTAB is the cheapest, fastest place to kill a weak patent (~18 months, ~$350K–$600K), and IPR petitions in parallel with district court are now standard defensive practice
• After eBay v. MercExchange (2006), permanent injunctions are rare — fewer than ten a year for the whole US patent system, and almost never granted to NPEs
• The ITC's Section 337 power to block imports at the border is the highest-leverage remedy for any product manufactured outside the US, and it works on a 15–18 month timeline
• Vietnam runs patent enforcement almost entirely through Thanh tra Khoa học và Công nghệ administratively — only 0–2 patent court cases per year nationwide, vs over 3,000 administrative cases in 2024
• For small builders, the cheap defensive moves (LOT Network, Unified Patents, RPX, basic prior-art habits) earn their cost long before any demand letter arrives
• Settlement isn't the system failing — it's the system working, with the trial backstop quietly setting the price

The bottom item is the one I want to carry out of this series. A patent isn't a thing you win or lose. It's a thing you negotiate, in the shadow of forums that most people don't think to look at. Once you can see the forums, the negotiation makes sense. Once the negotiation makes sense, the demand letter stops being scary — it's just the opening move in a game whose rules are mostly published, mostly cheap to learn, and mostly resolved without ever stepping into a courtroom.

Where these numbers come from

• Lex Machina 2025 Patent Litigation Report (USPTO PTACTS PDF) — Used for: 2024 district court filing counts, E.D. Texas resurgence, time-to-trial estimates by district, and the 68%-of-settlements-between-Markman-and-summary-judgment stat in §4.
• AIPLA 2023 Economic Survey — Used for: median cost to defend a patent suit through trial broken out by amount at risk.
• Marcum 2024 Patent Litigation Study — Used for: median damages awards, NPE vs practicing-entity damages divergence, permanent-injunction count over time (80 → 36 → ~7/year).
• PTAB Litigation Blog, "PTAB AIA FY2024 Roundup" (January 2025) — Used for: IPR institution rate (~68% by petition / 74% by patent), final-decision outcomes by claim cancellation, IPR vs PGR filing volumes.
• Sterne Kessler, "2025 ITC Section 337 Year in Review" (January 2026) — Used for: ITC 15–18 month timeline, 50–70 new investigations per year, 2024 Final Determination count.
• Reuters, "Apple defeats bid for new Apple Watch import ban at US trade tribunal" (April 2026) — Used for: the Apple v. Masimo exclusion-order example and Apple's software-disable response.
• Patently-O, "Fed. Circ. Patent Decisions In 2024: An Empirical Review" (January 2025) — Used for: post-eBay injunction grant rate divergence between practicing entities (~70–80%) and NPEs (under 10–15%), and the eBay v. MercExchange four-factor test.
• DNLAW, "Vietnam's Trademark Litigation Evolution" (February 2026) — Used for: 17 IP-related cases handled by Vietnamese courts 2022–May 2025, why patent litigation is rare compared to trademark disputes.
• Tilleke & Gibbins, Intellectual Property in Vietnam (2024) — Used for: Thomas Treutler's "slowly professionalizing" framing of the post-2022 civil court trajectory.
• LawFirmElite, "Thủ tục xử lý xâm phạm Sáng Chế mới nhất 2024" — Used for: Vietnam administrative fine caps (250M VND individuals / 500M VND organizations) for patent infringement.
• Thanh tra Khoa học và Công nghệ 2024 enforcement totals — Used for: 3,049 IP enforcement cases in 2024 (administrative pathway dominance).
• LOT Network Report 2025 (November 2025) — Used for: 6,000+ member count, growth from 4,500 in late 2023, notable members including Google/Meta/Amazon/Microsoft, the "vaccine against PAEs" structural design.
• Unified Patents, "About Us" — Used for: the deterrence-via-IPR model and membership cost ranges.
• Cloudflare Blog, "Patent troll Sable pays up, dedicates all its patents to the public" (October 2024) — Used for: the Sable Networks 2024 Waco verdict, public-domain dedication, and Project Jengo bounty program outcomes.
• Sterne Kessler, "2024 Federal Circuit IP Appeals: Summaries of Key Decisions" (February 2025) — Used for: Federal Circuit affirmance/reversal/remand rates on patent appeals.
• The Verge, "Apple and Samsung finally settle their seven-year patent dispute" (June 2018) — Used for: the Apple v. Samsung "Smartphone Wars" settlement framing.
• Law360, "Apple Wins Latest VirnetX Patent Fight At Fed. Circ." (September 2023) — Used for: the VirnetX v. Apple parallel district court / PTAB track as a forum-shopping example.
• Reuters, "Cisco wins reversal of $2.75 bln patent verdict over judge's conflict" (August 2023) — Used for: the Centripetal v. Cisco award reversal and the judicial-conflict cautionary tale.
• Mark Lemley, Stanford Law (2024 papers) — Used for: the "patent system is increasingly a tax on innovation" framing on patent litigation economics.
• George Priest & Benjamin Klein, "The Selection of Disputes for Litigation," Journal of Legal Studies (1984) — Used for: the canonical academic foundation for §9's "high settlement rates are the system working" rebuttal.
• Marc Galanter, "The Vanishing Trial" (Journal of Empirical Legal Studies, 2004) — Used for: the broader civil-litigation context (11.5% → 1.8% trial rate, 1962 → 2002) that makes patent settlements look less unusual.
• USPTO press release on PTAB Director Review final rule (September 2024) — Used for: Director Kathi Vidal's "fair, transparent, and efficient" statement in §9.
• WION News on VinFast v. ArcelorMittal ITC investigation and PACER Monitor docket (2:24-cv-03128 C.D. Cal.) — Used for: the Vietnamese-hardware-at-the-ITC anchor case in §6.
• IP Vietnam 2024 Annual Report (via Lexology summary) — Used for: 9,904 invention applications and 4,430 patent grants in Vietnam in 2024 (+21% YoY) in §7.
• Cloudflare Blog — "The Project Jengo Saga" and "Winning the Blackbird Battle" by Doug Kramer — Used for: §8's Project Jengo narrative and the Matthew Prince Ars Technica quote on patent trolls.