The UPC in 2026: A Faster Court, Now Mostly in English
UPC 2026 in data: the median time to a first merits decision falls to 13.7 months across three consecutive half-years, English reaches 69% of new main proceedings, and most disputes still end without a judgment.
Three years into the Unified Patent Court's life, the first half of 2026 sharpened two things the data had only been hinting at: the court is getting faster to a decision on the merits, and it is increasingly an English-language court.
This is our half-year read, built from 2,870 case records and the roughly 1,900 decisions and orders we have classified, current through the end of June 2026.
The court keeps getting faster
Speed is the promise the UPC was built on, and the data shows it holding under a growing load. We measure the median time from filing to the first decision on the merits in first-instance main proceedings, with provisional-measures cases excluded so that fast-by-design urgency proceedings do not flatter the average. That median has fallen for three consecutive half-years:
| Half-year of decision | Median months to first merits decision | Decisions |
|---|---|---|
| H1 2025 | 15.2 | 15 |
| H2 2025 | 14.7 | 28 |
| H1 2026 | 13.7 | 32 |
And the court managed it while deciding more merits cases than in any previous half. A party filing an infringement action today should plan around a first substantive ruling in roughly 13 to 14 months, with the Court of Appeal adding a median of about 11 months for those who go the distance.
Where the work sits
The docket remains concentrated in a handful of divisions, and their profiles are by now distinct. The table counts all main proceedings since the court opened, with the median time to a first merits decision shown where at least five such decisions exist.
| Division | Main proceedings | Median months to merits decision | Infringement / revocation / provisional |
|---|---|---|---|
| Munich LD | 386 | 14.5 | 191 / 107 / 31 |
| Düsseldorf LD | 221 | 16.1 | 116 / 50 / 33 |
| Mannheim LD | 192 | 17.0 | 95 / 55 / 5 |
| Paris CD | 120 | 14.2 | 6 / 80 / 0 |
| Hamburg LD | 98 | too few decided | 50 / 20 / 18 |
| The Hague LD | 85 | 15.2 | 54 / 13 / 14 |
Two things stand out. First, the big divisions sit closer together on speed than reputation suggests: the spread between Munich and Mannheim is about two and a half months, not a categorical difference. Forum choice at the UPC is better understood as a choice of panel and case mix than a choice of calendar. Second, the Court of Appeal has quietly become one of the busiest venues in European patent law in its own right, with 489 main appellate proceedings on our count.
English becomes the court's default language
The language shift that began in 2024 kept running through the first half of 2026. Of the new main proceedings visible for H1 2026, 69% were filed in English, up from 63% a year earlier. German fell from 30% to 24% over the same period; French, Italian and Dutch together account for the rest. These are shares of the filings visible so far and, like the raw filing counts below, they will firm up as late filings surface; but the English-language trend has run in one direction for two years. A court whose largest divisions sit in Germany now conducts more than two thirds of its new main proceedings in English.
How proceedings actually end
A judgment on the merits is not how most UPC disputes end. Between January and June, alongside the 37 substantive merits outcomes we classified, we recorded 34 withdrawals and 12 settlements. In other words, for every proceeding the court decided on the merits this half, roughly one and a quarter ended without a judgment at all.
That balance is the quiet argument for knowing the base rates before filing. Most UPC disputes are priced and resolved in the shadow of the numbers, not in the courtroom, which is exactly why the numbers are worth having.
Security for costs grows into underwriting data
Security for costs kept developing from a procedural footnote into a genuine underwriting question. Our corpus now tracks more than 170 proceedings with security-for-costs activity, with ordered amounts ranging from five figures to seven. The Court of Appeal's ruling in Syntorr v Arthrex in February 2026, accepting litigation insurance in place of a cash deposit, has begun to reshape what respondents ask for and what courts accept as collateral.
For funders and insurers, this is now a dataset rather than an anecdote: who gets ordered to post, in which divisions, for how much, and what happens to the money.
A note on filing counts
We deliberately read this court through decisions rather than filing volume. New filings surface in the public registry with a delay that can run from days to months, so counts for the most recent window are floors, not finals. On what is visible as of the end of June, we hold 263 case records with H1 2026 filing dates, 226 of them main proceedings, against 413 and 428 records for the two halves of 2025. Those 2025 figures themselves grew for months after the periods closed, and we expect the H1 2026 count to do the same. We do not read a trend into a lagging count.
Method, in brief
The numbers here come from classifying each decision the court publishes, in the language it publishes it, then deduplicating to one outcome per case. Counts use main proceedings (infringement, revocation, provisional measures, appeals) and exclude derivative procedural applications, so a scheduling order is never mistaken for a substantive ruling. Time-to-decision medians run from filing to the first decision on the merits, exclude provisional-measures proceedings, and are shown with the number of decisions behind them so their weight is visible. Figures reflect our data as of 6 July 2026.
If you write about the court and want a number we have not published here, ask: we answer journalists' data questions free, with attribution.