Innovation rarely happens in isolation. Patents frequently list multiple assigneesassigneeThe entity (corporation, university, government, or individual) that owns the rights to a patent. or inventors from different organizations, revealing a rich tapestry of collaborative relationships. Network analysis uncovers the hidden structure of these connections -- which organizations co-patent together, which inventors form long-term collaborative teams, and how these networks have shaped the direction of technological progress.
Organizational Co-Patenting
When two or more organizations appear as assignees on the same patent, it signals a substantive collaborative relationship -- joint research ventures, licensing agreements, or cross-firm R&D partnerships. The network below maps these co-patenting relationships among organizations with significant collaboration ties.
Co-Patenting Network (All Organizations)
Co-patenting network among organizations with significant collaboration ties. Node size = total patents; edge width = shared patents. Hover over nodes for details; drag to reposition.
The rise of co-patenting reflects both the growing complexity of innovation and strategic inter-firm collaboration in technology development.
The network exhibits a clear core-periphery structure. A dense core of highly connected major patent holders sits at the center, while specialized firms and research institutions occupy the periphery with fewer but often strategically important connections. This structure reflects the hierarchical nature of industrial R&D, where large firms anchor collaborative ecosystems.
Inventor Co-Invention
Beyond organizational partnerships, individual inventors also form collaborative networks. When the same inventors repeatedly appear together on patents, it reveals stable research teams that persist across projects and years. These co-invention ties represent some of the most productive relationships in the innovation ecosystem.
Co-Invention Network (All Inventors)
Co-invention network among inventors with significant collaboration ties. Edges = shared patents. Node size = total patents. Hover over nodes for details; drag to reposition.
The increasing connectivity of the co-invention network means knowledge can diffuse faster, but may also create path dependencies in innovation direction.
Comparing the two networks reveals a fundamental asymmetry: organizational co-patenting networks are relatively dense and interconnected (many firms co-patent with many others), while inventor networks are sparser and more clustered (inventors tend to work within small, stable teams). This suggests that cross-firm collaboration is driven more by institutional partnerships than by individual inventor relationships.
The US-China Decoupling
International co-invention rates — the share of US patents with inventors from multiple countries — reveal the evolving geography of collaborative innovation. US-China co-invention grew steadily from near zero in the 1990s to a peak around 2017–2018, coinciding with trade tensions, entity list restrictions, and tightening export controls. The pattern is sharpest in semiconductors and AI.
US Co-Invention Rates by Partner Country
Share of US patents co-invented with each partner country, 1976–2025. A co-invented patent has at least one inventor in the US and at least one in the partner country.
US-China co-invention grew rapidly after China's WTO entry in 2001 but flattened around 2018, coinciding with trade tensions and technology restrictions. US-India collaboration has emerged as a growing alternative.
US-China Co-Invention by Technology Area
Annual count of US patents co-invented with Chinese inventors, by CPC section.
The US-China collaboration pattern varies dramatically by technology. Electricity (H) and Physics (G) — which include semiconductors, AI, and telecommunications — show the most pronounced flattening, consistent with targeted technology restrictions in sensitive areas.
Talent Flows Between Companies
When inventors file patents at different organizations over their careers, they create talent flowstalent flowThe movement of inventors between organizations, tracked by consecutive patent filings with different assignees. Net talent flow reveals which companies are gaining vs. losing inventive talent. that transfer knowledge between companies. The Sankey diagramSankey diagramA flow diagram where the width of links is proportional to the quantity of flow. Used here to visualize inventor talent flows between companies. below maps these inventor movements between major patent filers.
Inventor Talent Flows
Movement of inventors between top patent-filing organizations, based on consecutive patents with different assignees (gap ≤ 5 years). Blue = net talent importer; red = net exporter.
Large technology companies tend to be net talent importers, drawing inventors from smaller firms and universities. The bidirectional nature of many flows suggests active talent cycling within industry clusters.
Competitive Proximity Map
How similar are companies' patent portfolios? By computing cosine similaritycosine similarityA measure of similarity between two vectors based on the angle between them. Values range from 0 (completely different) to 1 (identical). Used to compare patent portfolio compositions between companies. between CPC subclass distributions and projecting with UMAPUMAPUniform Manifold Approximation and Projection — a dimensionality reduction technique that preserves both local and global structure, used to visualize high-dimensional data in 2D., we can visualize the competitive landscape of innovation.
Patent Portfolio Proximity (UMAP)
Each dot represents a company. Proximity reflects similarity in CPC subclass distributions. Color indicates industry cluster.
Companies cluster by industry, but the boundaries are increasingly blurred. Technology conglomerates sit at the intersection of multiple clusters, reflecting diversified portfolio strategies.
Innovation Strategy Profiles
Each company pursues a distinct innovation strategy that can be characterized across multiple dimensions. The radar chartradar chartA multi-dimensional chart that plots values along multiple axes radiating from a center point. Useful for comparing strategy profiles across several dimensions simultaneously. below compares strategy profiles across 8 dimensions for the most prolific patent filers.
Innovation Strategy Radar
Eight-dimensional strategy profile comparing selected companies. All dimensions normalized to 0–100 scale across the top 30 assignees.
Companies show distinctive strategy fingerprints. Some emphasize breadth and collaboration (diversified conglomerates), while others optimize for depth and defensiveness (focused technology leaders).
Corporate Innovation Speed
How quickly do different companies move from application to patent grantgrant lagThe time (usually in days or years) between a patent application filing date and the date the patent is granted. Also called patent pendency.? Innovation speed varies dramatically across firms, reflecting both technology complexity and patent prosecution strategy.
Median Grant Lag by Company Over Time
Median days from application filing to patent grant for the top patent filers, by year.
Grant lag patterns reflect both the technology mix of a company's portfolio and its patent prosecution efficiency. Companies filing primarily in software and electronics face longer pendency, while mechanical and design patents move faster.
Having explored how inventors and organizations collaborate, the next chapter traces the knowledge flows that these connections enable. Patent citations -- the formal references linking new inventions to prior art -- reveal the deeper structure of how knowledge accumulates, diffuses, and shapes the direction of technological progress.
Co-patenting identifies patents with 2+ distinct organizational assignees. Co-invention identifies inventors who share multiple patents. Edge weights represent the number of shared patents. Only connections above significance thresholds are shown to reduce visual clutter. Talent flows track inventor movements between assignees based on consecutive patent filings with gap ≤ 5 years. Portfolio overlap uses cosine similarity of CPC subclass distributions projected to 2D via UMAP. Strategy profiles normalize 8 innovation dimensions to a 0–100 scale across the top 30 assignees.