According to the United States Patent and Trademark Office, more than 3,420,560 US patents were granted between 2015 and 2025, the window this analysis covers. Research question: Which technology areas attract the most US patent activity between 2015 and 2025, and what does the distribution reveal about the direction of American invention? Every figure below is computed directly from the public USPTO PatentsView dataset by the PlainPatent editorial team; see our methodology for how each number is derived and verified. Last reviewed June 2026.
Which technologies attract the most patents?
According to the United States Patent and Trademark Office, the 3,420,560 US patents granted between 2015 and 2025 are spread unevenly across technology, as our methodology details (reviewed June 2026). Innovation is not spread evenly across technology. The most active class in our dataset, G06F — Electric Digital DATA Processing — carries 580,551 classified patents, comfortably ahead of H04L at 373,517. The classes at the very top are dominated by digital technology: data processing, electrical communication, wireless networks, and semiconductor devices. This reflects the central economic fact of the last decade, namely that software, connectivity, and the chips that run them have been the primary engines of commercial R&D.
The concentration is partly an artifact of how broad these classes are — a general computing class naturally captures inventions from many industries — but it is mostly real. The companies that file most heavily in these areas are the same electronics and semiconductor multinationals that top the overall company ranking, and their portfolios are deliberately stacked in connectivity and chip design where products are most defensible.
A long tail of specialized fields
Below the digital giants sits a long tail of focused technology classes covering everything from additive manufacturing and medical devices to chemistry, materials, and mechanical engineering. These classes have far smaller counts, but they are often where the most contestable opportunities lie. A class with a few thousand patents and no dominant holder is far more open to a new entrant than a class with hundreds of thousands of patents locked up by three incumbents. Browsing the full list of technology classes by activity is the fastest way to find both the crowded arenas and the quieter frontiers.
Class size also shapes how you should read a company's specialization. A firm holding a few hundred patents in a niche class may be a leader in that field even though it barely registers on the overall ranking. This is why the technology pages on PlainPatent show the leading holders for each class: dominance is local, and the interesting competitive stories happen inside individual classes rather than in the aggregate.
Reading the cluster map honestly
Two caveats keep this analysis grounded. First, because a single patent can be assigned multiple CPC codes, the class counts overlap and cannot be summed into a portfolio total — they measure activity within each area, not slices of a pie. Second, CPC classification is assigned by USPTO examiners based on the technical disclosure, and boundary cases between adjacent classes are a matter of expert judgment, so counts near a class boundary carry some noise. Neither caveat changes the headline picture: US patenting concentrates heavily in digital technology, with a wide and varied tail of specialized fields beneath it.
For anyone mapping a market, the practical method is to start from the technology class rather than the company. Identify the CPC classes that cover your product, read their leaderboards to gauge how contested they are, and then trace the leading holders back to their full portfolios. The cluster map tells you where the activity is; the per-class leaderboards tell you whether there is room to compete. Use the technology leaders ranking to see which companies set the pace in each field.
The bottom line
US patenting concentrates heavily in digital technology, led by G06F (Electric Digital DATA Processing), with a long tail of specialized fields trailing the giants. That shape has been stable for years and tracks the broader economy: the most heavily patented areas are the ones where products are software-defined, connected, and chip-dependent, because those are the technologies where a patent most directly protects a commercial advantage. If you want a single sentence summary of where American invention is pointed, it is toward computing and communication, with everything else competing for the remaining attention.
The practical lesson for research is to never stop at the aggregate. The cluster map identifies the crowded centers, but the interesting opportunities — and the clearest competitive stories — live inside individual classes, where a few thousand patents and an open leaderboard can signal a genuinely contestable field. Treat the top of the distribution as the established arenas and the long tail as the frontier, then use the per-class pages to decide which is which for your specific question. The same dataset supports both the bird's-eye view and the close read, which is exactly what good technology-landscape analysis requires.
It is also worth remembering why a broad class can appear to dominate even when the invention inside it is diffuse. The classification system groups technologies hierarchically, and the largest classes are deliberately wide containers that gather many adjacent inventions under one heading. A general computing or data-processing class therefore accumulates filings from finance, healthcare, logistics, entertainment, and a dozen other industries that all touch software, which inflates its count relative to a narrowly drawn mechanical or chemical class. This does not make the concentration false, but it does mean class size and class breadth are entangled, and a careful reader separates the two. The honest way to compare fields is to look not only at how many patents a class holds, but at how many distinct companies actively file there and how evenly the activity is spread among them. A wide class crowded with hundreds of filers behaves very differently from an equally large class controlled by a handful, and only the per-class leaderboard reveals which situation you are looking at.
Methodology
We aggregated granted-utility-patent classifications by CPC subclass from the PatentsView dataset and ranked all 670 classes by classified-patent count. A single patent can carry several CPC codes, so a patent may be counted in more than one class; the figures measure activity per technology area, not a partition of unique patents. Read the full site methodology →
Data provenance and standards
Every figure above is computed deterministically from a single ingested snapshot of the public USPTO PatentsView release, never estimated, scraped from secondary summaries, or adjusted by hand. Our pipeline disambiguates assignees, normalizes corporate naming variants, aggregates grants by year and classification, and stores the reconciled results so that any number is reproducible and auditable against the original government records. Where a measurement carries unavoidable caveats — examination lag in recent grant years, the overlapping nature of classification codes, imperfect entity reconciliation across subsidiaries — we surface those limitations explicitly rather than presenting a tidier picture than the evidence supports. This commitment to transparency, verifiable provenance, and honest uncertainty is what separates rigorous analysis from decorative statistics, and it governs every study we publish.
Primary data sources: USPTO PatentsView — CPC classifications; Cooperative Patent Classification. Reviewed by PlainPatent Editorial · 2026-06-02.