Patsnap, a Singapore-based AI platform for patent and R&D intelligence, hit unicorn status in 2021 with a valuation over $1 billion after a $300 million Series E round led by SoftBank Vision Fund 2 and Tencent. High-profile clients include NASA, Tesla, Vodafone, MIT and General Electric. The company was founded in 2007.
After the CEO of the company, Jeffrey Tiong, submitted an article to R&D World, my interest in the platform was piqued as someone who has explored patent filings for a range of stories and projects. I asked for a free trial, which they were nice to provide for one month.
After firing up Eureka Scout, a core module in Patsnap’s Eureka platform, I found a platform more flexible than others I’ve tried. The AI-driven suite for R&D and IP workflows handles company and technology scouting by tapping into millions of patents and papers. Its website said its Eureka tool has 20,000 R&D engineer customers.
A proprietary IP GPT
Patsnap uses its own proprietary LLM called PatsnapGPT, trained on billions of data points across patents, papers and technical documents. In the user interface, you can ask open-ended questions about patents. As with large-language models generally, users are rewarded for formulating clear and targeted queries. It can wander if you ask a question that is nebulous or outside of its scope.
Patsnap now claims over 3.5 billion data points, including more than 190 million patents and over 200 million non‑patent literature entries.
The tool aims to cut down on manual research time, delivering data on competitor profiles or tech trend reports. It’s part of a lineup that includes agents for idea generation, patent protection, IP filing, life sciences analysis and materials formulation. In my tests, it performed well at answering specific queries related to patent filing trends of various companies.
A flexible AI-powered patent research platform
I tested queries on major electronics companies like Samsung, Sony, and LG, all of which are prolific patent filers, and it unearthed a trove of information while also highlighting key themes. The platform pulls visuals like publication trend charts, word clouds for keyword clusters (showing terms like “electronic components,” “printed circuit boards” and “multilayer ceramics”), and matrices mapping technologies across different application domains.
The platform delivers visualizations including patent application trend charts, word clouds for keyword clustering (displaying terms like “electronic components,” “printed circuit board,” and “multilayer ceramics”) and tech application matrices that cross-reference specific inventions—such as optical elements, printed circuits and semiconductors—against targeted improvements like enhanced efficiency, higher resolution, reduced power use or stronger bonding. These matrices use bubble sizes or numbers to indicate the volume of patents or innovations in each intersection, helping users spot where tech overlaps with practical goals like preventing leaks or boosting sensitivity in devices.
Rich data visualizations and technology mapping supported
The interface shows multiple integrated modules: Eureka Scout for company and tech scouting, Generate Ideas for innovation exploration, Protect Solutions for IP protection, and specialized tools for Life Sciences and Materials analysis. Each module offers specific capabilities—for instance, the Technical Q&A feature can answer queries about companies’ most widely cited patents, while the Idea Validation tool helps assess patent similarities for new concepts.
Beyond landscape analysis, Patsnap’s modules address specific workflow needs throughout the innovation lifecycle. The Idea Validation feature exemplifies this focus, allowing researchers to assess white space or whether their concepts might infringe on existing patents. The tool can also help spot potential collaboration opportunities.
Electronics giants and Big Pharmas under the IP microscope
In testing, queries about major electronics companies generated extensive data visualizations. The patent trend analysis revealed filing patterns over time, with Samsung’s data showing peak activity in the mid-2010s before declining sharply after 2020. The platform can also analyze specific therapeutic areas, as demonstrated by searches returning detailed information about Pfizer’s multiple myeloma treatments.
What sets this apart from traditional patent search tools is the AI’s ability to identify non-obvious connections. For instance, when analyzing Samsung’s display technologies, the system surfaced unexpected cross-references to their battery management patents, for instance.
The platform’s versatility extends beyond electronics into life sciences, where patent complexity reaches another level entirely. When querying the platform about Pfizer patents, it demonstrated sophisticated handling of chemical structures and therapeutic applications.
The word cloud visualization below shows how the system identifies and weights technology relationships. Check out how ‘Printed circuit board’ emerges as a central node, with related technologies like “Camera module,” “Multilayer ceramic” and ‘Coil component’ forming distinct but interconnected clusters.
The word cloud visualization above highlights the platform’s ability to identify key technology clusters, while the Tech Application Matrix below provides a cross-functional view of how different technologies apply across various industrial domains. The bubble sizes indicate patent volume at each intersection. See how optical elements dominate efficiency improvements (75 patents), while printed circuit manufacturing shows broader distribution across multiple application areas. This matrix essentially maps where innovation is happening and where white space opportunities might exist.
For researchers and journalists tracking innovation trends, the platform offers practical applications for competitor benchmarking and technology landscape analysis. You can also ask it seemingly off-the-wall questions like “How to reduce motorcycle noise?” and it will do a deep literature search, all while showing its thinking traces. The quality and depth of results can vary depending on the complexity of queries and the specific data being searched. In some cases, you might still want a fallback for deeper queries or a more targeted approach based on SQL for instance. But the barrier to entry for such databases is greater.
(Disclosure: Patsnap’s enterprise pricing is not publicly disclosed, and specific pricing for individual researchers or journalists was not provided during my trial.)
