Global AI Analysis

Semiconductors

Semiconductor Supply Chain by Subcategory & Country

The global semiconductor supply chain reveals stark geographic concentration across all four resource types. Design resources command the largest share at $648B, led by US-headquartered fabless firms and IP licensing giants. South Korea and Taiwan anchor Process capabilities through Samsung and TSMC's combined foundry scale, while Japan's Materials strength forms an indispensable but often underappreciated layer of the stack.

$648B
Design is the largest supply chain segment — dominated by US fabless firms and IP licensing. Click any box to drill down into sub-segments.

Loading interactive visualization...

Source: CSET · SIA · McKinsey Global Semiconductor Report

Country Rankings in Global Semiconductor Supply Over Years

Between 2019 and 2024, rankings within the semiconductor supply chain have remained broadly stable at the top, masking important shifts in mid-tier positions. China has moved up steadily despite export restrictions, reflecting sustained domestic capacity investment. The Netherlands holds an outsized position — disproportionate to its economy size — through ASML's monopoly on EUV lithography tooling, making it a pivotal single point in the global supply network.

Netherlands
Small economy, singular leverage: ASML's EUV monopoly makes the Netherlands an irreplaceable chokepoint in advanced chip manufacturing.

Loading interactive visualization...

Source: CSET · SIA · Rank by share of total supply chain value · 2019 → 2022 → 2024

Global Semiconductor Market Breakdown by Country and Resource Type

The US dominates Design resources by a wide margin, reflecting the concentration of leading fabless chip companies (Nvidia, Qualcomm, AMD) and EDA software vendors that underpin global chip development. Taiwan and South Korea lead in Process through world-class foundry infrastructure, while Japan's Materials and Tool strengths complete a triangulated dependency that no single nation can replicate alone. This interdependence is simultaneously a stabilizing force and a geopolitical vulnerability.

US $373B
US leads Design — the highest-value segment — by a margin no other country approaches. Reflects deep concentration of fabless IP and EDA tooling.

Loading interactive visualization...

Source: CSET · SIA · Sorted by total market size

AI Innovation

Global AI Innovation & Workforce by Country

The US and China present two contrasting profiles in AI capacity. The US dominates deployed workforce with 170,061 AI workers versus China's 5,342, reflecting deep integration of AI talent into industry. China counters with sheer publication and patent volume — 22,665 papers and 48,660 patents — suggesting a strategic emphasis on output and IP claims over workforce deployment. These profiles reveal fundamentally different approaches: US depth versus Chinese breadth.

31×
The US has 31 times more deployed AI workers than China — the largest capability gap between the two leading AI powers.

Loading interactive visualization...

Source: PARAT · OECD AI Policy Observatory · Area proportional to capacity across papers, patents, and workers

Global AI Patents By Sub-Domain

Computer Vision dominates the AI patent landscape with 27,304 filings — more than double any other sub-domain — reflecting its commercial maturity in surveillance, autonomous vehicles, and consumer devices. Control systems (11,662) and Analytics (10,733) follow, signaling that applied industrial AI is drawing serious IP investment alongside consumer applications. Emerging domains like NLP and AI Safety remain comparatively thin, suggesting the patent race has yet to fully reach generative and safety-critical AI.

27,304
Computer Vision patents — more than double the #2 domain. Reflects commercial maturity in surveillance, autonomous vehicles, and consumer hardware.

Loading interactive visualization...

Source: PARAT · USPTO · WIPO · Patent volume across AI technology sub-domains

Research Capacity & Collaboration

AI Research Capacity: Publications vs. Citations

Research output and citation impact are highly unequal: a small cluster of elite institutions — primarily US and Chinese universities and research labs — account for a disproportionate share of high-citation work. The top 20 organizations sit in an exponential outlier cluster well above the trend line, meaning their research is cited far more than their publication count alone would predict. Volume and impact diverge sharply; the citation leaders define the field's direction.

Top 20
Organizations account for a disproportionate share of total citations despite representing less than 1% of all institutions tracked.

Loading interactive visualization...

Source: PARAT · Semantic Scholar · Top 20 organizations highlighted by total citation impact

Global Research Collaboration Network

The US-China research collaboration link is the single thickest arc in the global network, representing 162,646 co-authored articles — far larger than any other bilateral pair. This paradox of deep scientific interdependence amid rising geopolitical tension is the defining dynamic of the field. Decoupling AI research would require severing one of the world's most productive scientific partnerships, with cascading effects on citation networks, talent exchange, and benchmark development.

162,646
US-China co-authored AI papers — the world's most productive bilateral research relationship, and a structural barrier to decoupling.

Loading interactive visualization...

Source: PARAT · Semantic Scholar · Arc thickness proportional to co-authorship volume

Global Cross-Border AI Research Collaboration

Cross-border AI research collaboration grew substantially across all tracked fields from 2015 to 2023. Computer Vision and core Artificial Intelligence saw the sharpest absolute increases, driven by shared benchmark datasets and open-source frameworks that lower collaboration barriers. Emerging fields like AI Safety and Large Language Models show accelerating growth from a small base — suggesting the collaborative frontier is shifting toward the most consequential and contested areas of the field.

All 8 fields
Show consistent year-over-year growth — collaboration has expanded through policy tensions, not despite them.

Loading interactive visualization...

Source: PARAT · Semantic Scholar · Cross-border co-authorship volume by research field · 2015–2023

AI Research Collaboration by Field — US & China

Despite escalating geopolitical friction, US-China AI research collaboration remained robust through most of the 2015-2023 period before showing signs of deceleration at the tail end. The collaboration is deepest in Computer Vision and core AI, where shared conference culture (NeurIPS, CVPR) and open-source norms keep researchers connected. Any structural decoupling would need to overcome deeply entrenched researcher-level networks that have produced some of the field's most-cited and foundational work.

2018–2020
Collaboration persisted through the trade war years, peaking around 2021 before policy tightening began to show measurable effects.

Loading interactive visualization...

Source: PARAT · Semantic Scholar · Collaboration trends across 8 AI research fields · 2015–2023

Partner Diversity — Global AI Research Hubs

The upper-right quadrant — high partner diversity and high collaboration volume — identifies the world's true AI research hubs. The US occupies this position most definitively, collaborating with the widest range of countries at the greatest aggregate volume. Countries in the middle cluster show that raw collaboration volume often concentrates with just a few major partners, limiting the diversity of ideas and talent networks. Hub status confers compounding advantages in attracting talent and setting research agendas.

US leads both
Highest collaboration volume and widest partner diversity simultaneously — a dual dominance that reinforces US centrality in global AI research.

Loading interactive visualization...

Source: PARAT · Semantic Scholar · Top 15 countries highlighted by unique collaboration partners

Policy & Investment

US AI Policy Stance Over Time (2019–2025)

US AI policy has shifted decisively toward restriction since 2022, with Restrictive measures consistently outnumbering Enabling ones from 2023 onward. The 2024 peak of 65 Restrictive policies reflects a wave of export controls on advanced chips (H100, A100), investment screening via CFIUS expansions, and computing access restrictions targeting Chinese entities. Balanced and Neutral policies remain present but are now secondary to a national security frame that has fundamentally reoriented US AI governance.

2024: 65 vs. 38
Restrictive policies now outpace Enabling ones by nearly 2:1 — a structural reversal from the 2019-2021 period that favored AI enablement.

Loading interactive visualization...

Source: OECD AI Policy Observatory · CSET · Policy count by stance category · 2019–2025

Global AI Patent Landscape — Sector Categories

AI patent activity clusters in industrial and consumer applications, with Healthcare, Mobility, and Finance representing the dominant commercial sectors globally. Asia-Pacific leads by volume across most sectors, driven by Chinese strategic filing at scale, while US and European patents tend to concentrate in higher-value, more specialized categories. This divergence reflects fundamentally different IP strategies: volume-based portfolio building in Asia versus targeted protection of core innovations in the West.

Asia-Pacific
Leads patent volume in most sectors. Western filings are fewer but concentrate in higher-value niches — revealing two distinct IP strategy philosophies.

Loading interactive visualization...

Source: PARAT · WIPO · USPTO · Click any slice to drill down by region, country, and sector

Global AI Investment Heatmap — Top 15 Countries & Fields

US AI investment is concentrated overwhelmingly in Mobile/Internet services at $168B, reflecting Big Tech's platform-scale bets on AI-native products. Healthcare and Data & Analytics represent the next priority clusters, both areas where AI offers clear productivity leverage. China's profile is more evenly distributed at lower absolute values, while European nations show targeted investment in industrial automation and specialized verticals — revealing a fundamental difference between platform-building and sector-specific application strategies.

US $168B
In Mobile/Internet alone — more than the total AI investment of most other nations combined. Reflects Big Tech's outsized role in US AI commercialization.

Loading interactive visualization...

Source: CSET · PitchBook · AI investment across countries and technology categories

Publications

Global AI Publications Racing Bar (2015–2023)

China's rise in AI publication volume is the defining story of the 2015-2023 period. China surpassed the US around 2017 and reached 103,672 publications by 2023 — more than double the US total of 50,359. India's emergence as a significant third contributor (43,816) is the decade's secondary story, while European nations have grown at slower but steady rates. The raw volume race, however, obscures quality differences better captured in citation analysis — China's publication lead does not directly translate to equivalent research influence.

103,672
China's 2023 AI publication count — 2× the US total and growing faster year-over-year. India at 43,816 is rapidly closing the gap with the US.

Loading interactive visualization...

Source: PARAT · Semantic Scholar · Top 15 countries by annual AI publication count · 2015–2023