Patents Are Taking Longer to Grant Again — Especially in Semiconductors

If you filed a patent application in the last couple of years, it's probably taking longer to grant than it would have in 2021 — and if it's a semiconductor patent, a lot longer. Median time from filing to grant across all US utility patents just hit a decade high, and one technology area is driving the swing far more than the rest: semiconductor patents (CPC class H01L) have gone from the fastest-improving category in the late 2010s to nearly doubling in pendency since, now sitting well above the industry-wide average.

We found this by mining our own patent catalog — over 13.6 million USPTO records spanning more than two decades, the same corpus behind prior-art search and Draft Builder's coverage checks. For this analysis we focused on the most recent decade of filing, grant, and status data. Rather than asking what's being invented, we asked something inventors and attorneys don't usually get to see directly: how the examination process itself has been changing underneath them.

A pandemic-era dip, then a steady climb

Across all granted patents, median time-to-grant fell from 32.4 months in 2016 to a low of 28.6 months in 2021 — likely reflecting a run of USPTO efficiency initiatives and pendency-reduction targets through the late 2010s. From there it reversed: 30.9 months in 2022, dipped slightly, then climbed to 32.4 in 2024 and 33.3 in 2025 — a decade high, and worse than where it started in 2016.

Semiconductors got hit hardest

The overall number undersells what's happening in specific technology areas. We broke pendency down by CPC classification and found semiconductor patents — CPC class H01L, covering semiconductor devices — moved far more sharply than the average. Median time-to-grant for H01L patents actually improved through the late 2010s, bottoming out at 22.4 months in 2018. Then it reversed hard: 24.4 months in 2021, 27.5 in 2022, and a jump to 32.2 in 2024 and 37.2 in 2025 — up nearly 66% from its 2018 low, and now well above the all-patent median.

Chemistry patents (CPC section C) show a similar pattern at a smaller scale, with median pendency stretching from 34.3 months in 2016 to 46.5 months in 2025. Electrical and electronics filings generally (CPC section H, which includes H01L) followed the same late-decade reversal, but nowhere as sharply as semiconductors specifically.

The counterintuitive part

The obvious explanation for slower grants is a filing surge overwhelming examiners — more applications in the queue than the USPTO can clear. That's not what's happening here. CPC section H — the whole broad "Electricity" category, covering everything from semiconductors to batteries to wireless comms — held a steady 35.8–37.3% share of each year's published patents from 2016 through 2025. No surge, no disproportionate shrinkage at that broad level.

Semiconductor patents specifically — CPC class H01L, one narrow slice within section H — tell a different story. H01L patents were 7.78% of 2020's published output, falling to 4.51% of 2025's: a real decline in H01L's share of recent patents, even as section H as a whole stayed flat. Fewer semiconductor patents each year, not more, running in parallel with rising pendency rather than explaining it via any kind of backlog.

It's also not a story of prosecution getting harder per case. We checked: for semiconductor patents (H01L) specifically, the rate of applications hitting a Request for Continued Examination fell from 44.2% (2012 filings) to 19.7% (2020 filings), and the rate hitting at least one final rejection fell from 55.4% to 28.1% over the same period — fewer rounds of back-and-forth, not more, even as pendency climbed. We also checked whether examiners simply had more cases to handle: they did, corpus-wide, but the rise was the same in CPC section G (whose pendency didn't spike the way H's did), and H01L's decisions-per-examiner held roughly flat over the same years its pendency kept climbing. So harder prosecution and examiner caseload growth are both ruled out, and the volume story runs the opposite direction from what a backlog theory would predict. What's actually driving the H01L pendency spike is still an open question for us; we'll dig further and follow up.

Why this matters if you're filing

If you're working on a semiconductor or chemistry-heavy invention, budget for a longer runway to grant than the industry-wide average suggests, and factor that into decisions like provisional filing timing, licensing negotiations, or funding milestones tied to an issued patent. It's also a reminder that "average time to grant" numbers — the kind you'll see quoted industry-wide — can hide very different realities depending on what you're actually patenting.

This kind of analysis is possible because our catalog spans the full USPTO corpus, not just a sample — and it's exactly the same underlying prior-art data we use to ground search results and Draft Builder's drafts in what the USPTO actually has on record.