All insights
AnalyticsPublished Updated

The UPC at Three Years: Outcome Base Rates Every Litigant Should Know

Three years in, what is a normal UPC win rate? Preliminary injunctions granted ~54% at first instance, infringement found in ~53% of merits rulings, and only about 4 in 10 challenged patents upheld as granted.

By

The Unified Patent Court turned three on 1 June 2026, and the single question every client still asks first — "What's the UPC win rate?" — finally has enough decisions behind it to answer with numbers instead of impressions. Across the UPC decisions our platform has classified, the base rates that drive settlement value are these: at first instance, preliminary injunctions are granted about 54% of the time, infringement is found in about 53% of merits rulings, and when validity is put in issue on the merits, the patent comes out exactly as granted only about 4 times in 10. None of those is the number a single firm headline gives you, because each headline picks a different denominator. This piece sets out what is actually normal.

That is the point of a base rate. A UPC case is rarely fought to judgment — most resolve by settlement or withdrawal — and the settlement price is set by what each side believes the court would do. Get the base rate wrong and you misprice the case. So here is the definitive "what's normal at the UPC" reference, built on the outcomes our platform has classified rather than on any single firm's headline.

Validity is a real risk, but not the whole story

The validity row is where careless reading does the most damage. When validity is put in issue on the merits, the patent is upheld exactly as granted in only about 4 in 10 contested-validity outcomes. In the rest it is amended, partially revoked or fully revoked: about 59% of outcomes leave the patent narrower or gone.

That 59% counts an amended-but-surviving patent as a loss of ground, because the claims narrowed. Strip amendments out and the patent is revoked in full or in part in about a third of contested-validity outcomes. In other words, a challenged patent survives in some form in roughly two cases out of three, but rarely untouched. That is the number a patentee facing a real prior-art case should price against, and it is not the same thing as the court-wide statistics that get quoted in headlines. It covers only the cases where validity was actually litigated to an outcome — a self-selected pool where someone already thought the patent vulnerable.

This is a different number from the court-wide statistic, and it measures a different thing. The UPC's own framing is of infringement actions as a whole — just over 50% end valid and infringed, about 25% not infringed, and about 25% invalid — but most infringement actions raise no serious validity challenge, so that picture is dominated by cases where validity was never really in play. It describes the whole infringement docket, not the risk to a patentee in a genuinely contested case. Bardehle's published breakdown of the merits is starker still: in infringement main proceedings, patents were maintained as granted in 52% of revocation counterclaims, amended in 16%, fully revoked in 31%; in standalone revocations, amended (38%) outran maintained-as-granted (24%). The lesson: if you hold the patent, price validity as a serious risk, and treat amendment as a likely way through.

The preliminary-injunction base rate, and the divisional spread beneath it

The preliminary-injunction grant rate is the most-quoted base rate, and our first-instance number is about 54%. As a separate court-wide metric, the UPC reports preliminary injunctions granted "just under 65% at first instance," and Bardehle's first-instance review puts it at 60%. The gap is mostly definitional: what counts as a decided application, and whether partial grants count as grants.

The average also hides a wide divisional spread: the German local divisions combined grant about 64%, ahead of the first-instance average. Bardehle adds the useful complement on why preliminary injunctions fail — the leading reasons for denial were insufficient proof of infringement (42%) then validity judged more likely invalid than not (32%), the same two fronts as the merits. The full divisional table is in a companion piece; the headline is that a UPC preliminary injunction is more likely than not to be granted, but the odds swing 20-plus points on forum and on infringement-evidence strength.

Most cases never reach these numbers — the settlement base rate

Every rate above is computed on the minority of cases that run to a ruling — because most disputes exit before judgment. In our corpus, the cases that settle, withdraw, or are dismissed outnumber the merits and preliminary-injunction rulings that produce the win rates. The UPC's own caseload framing is consistent — it reports that just over half of all infringement actions end in settlement, closure, or withdrawal — with withdrawals increasingly used to force a cross-border settlement and walk away.

This closes the loop on why base rates matter. A settlement is a bet against the judgment that would otherwise come, priced on the grant rate, the infringement rate, and above all the validity-survival rate. A patentee who anchors to a whole-docket "valid and infringed in just over half" figure holds out for more than one who understands that a contested patent comes out exactly as granted only about 4 times in 10. The number you settle on is only as good as the base rate you anchored it to.

Where the base rates bend — division and technology

Two slices move the averages enough to price in — read the segment-level rates below as directional rather than settled.

By technology, the rates differ too, but the samples per field are still small enough that a single campaign can swing them, so read any field-level rate as directional. Telecom and pharma dominate the docket by volume, and pharma is where validity attacks are fought hardest.

By division, infringement-finding rates differ — Düsseldorf ~69%, Mannheim ~58% — but the samples per division are small. The signal across our cuts is consistent: forum choice moves the preliminary-injunction odds and the speed more reliably than the infringement-finding rate.

What to watch in year four

Three things will tell us whether these base rates are settling or shifting:

  1. Does the validity-survival rate hold as the docket matures? The validity decisions behind these figures will keep accumulating, and the two Court-of-Appeal landmarks of 25 November 2025 — on inventive step and on sufficiency of disclosure — will frame how year-4 validity challenges are argued. If the CoA's inventive-step approach tightens, expect the revocation share to firm up, not fall.
  2. Does the CoA keep overturning a meaningful share of merits decisions? A notable share of the first merits appeals were reversed. As the appellate pool grows, that reversal rate is the key number for anyone deciding whether a first-instance UPC win is bankable or merely round one.
  3. Does the divisional preliminary-injunction spread narrow? With a 20-plus-point gap between divisions today, year 4 will show whether the Court of Appeal's harmonisation of the urgency and proportionality standards pulls the divisions closer together.

We will refresh every figure above as the year-4 decisions land. The base rates are now stable enough to plan around — not yet so settled that they will not move.

Want the live base rate for your division, technology, or opponent before you price a settlement? That is what UPClytics is built for. Explore the data.

Sources

Base RatesInjunctions

Related insights

See the data behind the analysis.

Every article here is built on UPClytics — cross-case UPC analytics on divisions, judges, firms, and outcomes.