The State of AI Competition in Advanced Economies

October 06, 2025

The State of AI Competition in Advanced Economies

Alex Haag

Global competition in artificial intelligence (AI) has intensified in recent years. Some assessments emphasize US exceptionalism, while others argue that China is eroding US dominance. By contrast, the progress of other advanced foreign economies (AFEs) receives far less attention. Existing cross-country comparisons rely largely on composite indices that, while useful as benchmarks, are subject to weighting and aggregation biases that may obscure important dimensions of AI capacity. A clearer understanding of cross-country AI capabilities can help better contextualize global AI competition.

This note brings together international comparisons on key metrics to assess countries' relative preparedness and performance in AI. The analysis shows that the United States retains important advantages in infrastructure (with the exception of electricity-related infrastructure), compute capacity, and investment conditions. AFEs face greater challenges in scaling compute resources and attracting investment. China has achieved rapid gains in research output and adoption, though its position in enabling infrastructure is less clear.¹ Taken together, these comparisons suggest a global landscape in which, despite increased competition in research and applied domains, US strengths in core AI enablers remain durable.

1. Cross-Country Comparisons Using AI Indices

We start with several widely cited indices of AI capacity and preparedness to provide an initial view of the global landscape (Table 1).² The United States consistently ranks at or near the top, with the United Kingdom close behind and Italy lowest among G7 members. China's ranking varies from just below the United States on some indices (e.g., GII, Stanford) to significantly lower than most AFEs on others (e.g., IMF, Oxford).

These indices are a useful starting point but have limitations. Methodological approaches, such as indicator selection, weighting, and aggregation, can potentially obscure core metrics of AI performance and penalize diverse national AI strategies (Greco et al., 2018). Still, these indices highlight key areas for comparison: physical and digital infrastructure, investment and firm dynamics, and adoption across the economy. The rest of this note delves deeper into these dimensions, focusing on key metrics to provide a clearer picture of global AI capabilities.

Table 1: Cross-Country Comparison in Selected Technology and Artificial Intelligence Indices

2. Physical and Digital Infrastructure

Physical and digital infrastructure form the foundation for AI adoption. Physical infrastructure includes data centers and high-performance computing hardware, supported by reliable energy, telecommunications, and manufacturing capabilities. Digital infrastructure adds cloud computing, data access, and other tools for widespread model deployment.

As shown in Figure 1, the United States made early, outsized investments in computing, software, and databases, with annual real investment in these areas growing over tenfold from 1995 to 2021, far outpacing AFE peers, whose growth was two- to fourfold. These early investments provided the computing power, networks, and hardware that positioned the United States to lead early in AI-related innovation and diffusion.

Compute, the processing power and network resources used for AI training and inference, is the clearest measure of a country's ability to develop and deploy AI. Data centers, processors (CPUs and GPUs), and broader computing systems support the growing demands of modern AI workloads.

Data center construction in the United States surged over the past decade.³ The United States hosted an estimated 4,049 data centers as of 2024, far more than the EU (~2,250), the UK (484), and China (379).⁴ In 2024 alone, the United States added 5.8 gigawatts (GW) of data center capacity, compared with 1.6 GW in the EU and 0.2 GW in the United Kingdom.⁵ On a per capita basis, the US server base stands at 99.9 per 1,000 people, far surpassing other advanced foreign economies and China (Figure 2A).

High-end computing, particularly AI supercomputers, is particularly critical for large-scale model training and more recently inference-based tasks. The United States dominates cumulative AI supercomputer capacity, controlling an estimated 74 percent of global high-end AI compute, while China holds 14 percent and the EU 4.8 percent (Figure 2B).

3. Energy Infrastructure Risks

The rapid growth of AI workloads is driving a substantial increase in data center power demand, highlighting the importance of electricity generation and transmission capacity in supporting AI infrastructure. RAND estimates that AI could require 117 GW globally by 2028, while the IEA projects that global electricity consumption from data centers will more than double by 2030, with the United States and China accounting for 80 percent of the growth.⁶

China outstripped the United States in electricity generation capacity more than a decade ago (Figure 3), with roughly 3,200 GW of installed capacity compared with 1,293 GW in the United States and 1,125 GW in the EU. China added 429 GW of net electric generation capacity in 2024 alone, over 15 times the net capacity added by the United States (Chan et al., 2025).

It remains unclear whether US power generation and transmission capacity will keep pace with rapidly growing data center demand. Installed data centers currently account for about 8.9 percent of average US energy use, compared with 4.8 percent in the EU and 2.3 percent in China, highlighting the scale of additional supply that may be required in the United States to support continued expansion of AI infrastructure.⁷

4. Private Investment and R&D

US private investment in AI far outpaces that of other advanced economies. Figure 4 shows that, from 2013 to 2024, cumulative private AI investment in the United States exceeded $470 billion, compared with roughly $50 billion across EU countries, $28 billion in the United Kingdom, $15 billion in Canada, and $6 billion in Japan.⁸ The gaps between the United States and other advanced economies in private investment are most pronounced in AI infrastructure, research, and data management and processing.⁹ The United States also accounts for the majority of global venture capital investment in AI-related data startups and over 75 percent of reported funding in generative AI ventures.¹⁰

Firm-level research and development in AI-adjacent sectors further underscores US AI leadership. R&D intensity, measured as R&D spending relative to sales, is highest among US firms in both ICT hardware and software sectors (Figure 5). The EU trails slightly in hardware but falls far behind in software R&D, despite strong growth over the past decade. Chinese firms scaled hardware R&D compared to a decade prior but reduced software intensity, both of which remain far behind the United States. These patterns align well with recent OECD findings showing that a larger share of US firms conduct AI-specific R&D for commercial use than their AFE peers (OECD 2025).¹¹

5. Domestic Adoption of AI

Comparing AI adoption across countries faces several challenges. Survey-based measures from private institutes, companies, and universities often report much higher adoption than national statistics, largely because large firms in AI-intensive sectors are overrepresented (larger firms in general have higher adoption rates across countries and surveys). It is therefore unsurprising that the United States ranks highest in surveys such as Stanford's AI Index. National statistics aim for representative samples, which include more small and medium-sized enterprises that are less likely to use AI for various reasons (see Appendix C).

Differences in definitions further complicate cross-country comparisons. For example, the US Census Bureau's biweekly BTO survey measures consistent AI use within the past two weeks, while Eurostat reports annual adoption with less specificity on intensity. As highlighted by Crane, Green, and Soto 2025, accounting for employment weighting in these surveys substantially raises estimated US adoption, suggesting that firm-level responses alone may understate the diffusion of AI.¹²

Table 2: Selected Survey Results on AI Adoption

Alternative diffusion indicators, such as those focused on AI-related skills in the workplace, show the United States largely outpacing AFE peers. For example, only France surpasses the United States in fostering occupation transitions from non-AI to AI employment (a proxy for upskilling and reskilling of the workforce toward AI, Figure 6a). The United States far outpaces AFE counterparts in the incorporation of AI into production processes in manufacturing (Figure 6b), as well as in education, financial, and technology service sectors (Table A3).¹³ Further development and standardization of AI adoption metrics by firms and individuals will continue to clarify where advanced economies stand with regard to economy-wide AI diffusion.

Conclusion

This note documents how the United States largely outperforms AFEs across key areas of AI capacity, while China remains remotely competitive in some areas, with its largest advantage in energy infrastructure. The combination of robust physical and computing infrastructure, sophisticated investment climate, and emphasis on diffusing AI use across the economy will likely allow the United States to capitalize on future benefits from advances in AI, while deficiencies in one or more of these areas may limit the gains from AI in AFEs over the near term.

Appendix A: Additional Data

I. Compute
A common metric widely used to quantify compute capacity is Top500's supercomputer list.¹⁴ There are, however, multiple issues that should raise caution in interpreting Top500 results. The data represents a rotating list of the most powerful commercial supercomputers updated every six months. As such, figures across time may not reflect the same dataset of supercomputers. In addition, data suffers from self-reporting issues, particularly in the context of China, which stopped reporting its supercomputers in 2023, leading to potential overestimation for other countries on this list.

Table A1: Additional Compute Indicators

II. Investment
The United States has accounted for most global venture capital (VC) investment in AI and data startups and more than 75 percent of reported private VC investment in generative AI startups (Table A2). US VC investments to date were concentrated in IT, media/marketing, healthcare, and financial firms, where it maintains an outsized lead compared to both China and AFE counterparts. China maintains a slight advantage over the United States in private VC funding toward education and training startups. China also invests heavily in AI compute startups, the only country remotely close to the US level of VC investment in the sector. European and Japanese counterparts, in contrast, lag behind the United States and China across all categories by a considerable margin.

Table A2: Cumulative Venture Capital Investment in AI (Millions USD; 2012-2024)

III. Labor Markets

Table A3: Relative AI Skills Penetration in Advanced Economies (Global Average = 1)

Appendix B: Contextualizing Competition with China

This appendix highlights important considerations when comparing AI developments between the United States and China. Comparisons between the United States and China are complicated by several factors, including a lack of transparency into Chinese data associated with AI but also a markedly different approach to AI investment and adoption across the domestic economy. We highlight only a few of these differences here.

The rise of DeepSeek, which performs at or near US model capabilities across several benchmarks, demonstrates that in terms of model development, China maintains a competitive position relative to the United States. Experts emphasize, however, that these model capabilities are largely isolated to China's model training capabilities.¹⁵ This model training tells little of the broader AI ecosystem, composed of compute infrastructure, investment, and adoption across the economy that facilitate widespread AI diffusion.

Data availability on Chinese compute capacity is opaque, partly resulting from ongoing geopolitical tensions and export restraints on high-end computing materials. In response to these restraints, China has focused on a combination of circumvention and building domestic capacity, both of which it keeps largely veiled. As such, key indicators of China's position relative to the United States in AI, such as high-end compute capacity, are unreliable.¹⁶ Widespread smuggling of high-end chips and other circumvention of US export controls may also result in underestimating China's overall compute capacity.¹⁷

Some metrics, such as private investment, do not adequately capture the nature of AI investment in the Chinese economy, which focuses on a state-driven approach with local governments as diffusing agents. For example, Stanford researchers recently reported that over the past decade, China plowed $912 billion of investment into several key sectors through vehicles known as local government venture capital funds, with roughly $184 billion going to nearly 10,000 AI-related firms across China (Beraja et al., 2024; Omaar 2024). These local government VC funds tend to invest in a broader geographical profile across China compared to private VCs (which largely cluster in coastal regions), which may result in several benefits including more widespread dispersion of AI across the entire Chinese economy. In March 2025, China announced an additional venture capital guidance fund dedicating ~$138 billion over 20 years to AI and quantum technology.¹⁸ These funds, however, could likewise generate poor innovation returns on investment if funds are poorly allocated across firms due to, among other things, local government corruption.

Figures on China's AI infrastructure buildout may not fully capture inefficiencies that currently limit China's ability to close to the gap with the United States. For example, local outlets as recently as March 2025 reported that up to 80 percent of China's newly built computing resources were idle, as many companies running the data centers struggled to stay afloat.¹⁹ A misallocation of resources by both local officials and corporations seeking investment alternatives during the real estate slump reportedly led to the construction of data centers that fell short of industry standards and suitability requirements for ongoing AI market developments (e.g., the shift from pre-training workloads to inference-based tasks). This shift to inference and real-time reasoning, particularly in the advent of DeepSeek, reinforced the need for lower latency, which disadvantages Chinese data centers recently built in rural areas that benefit from cheap land and energy prices but face latency issues. In response to this data center glut, the Chinese government is currently conducting an assessment to regulate and optimize existing resources and sell off unused compute power (Reuters 2025). It remains unclear at present whether China can successfully repurpose these facilities for productive means, AI or otherwise.

Appendix C: Reasons for Lagging Adoption among Other AFEs

Drawing from a range of country-specific reports, this appendix reviews the main barriers that help explain why AFEs lag the United States in AI adoption. Some factors repeat earlier themes, such as limited computing power, while others relate to costs, skills, and finance.

• Lack of Computing Power Infrastructure: As discussed in Section 2, many AFEs face a shortage of computing capacity (Dobbs and Hirsch-Allen 2024; Schnabel 2024; Li 2024). The Euro Area missed out on earlier buildouts of IT and data infrastructure, while Canada reports gaps in AI-specific infrastructure. The United States has more than four times the compute performance capacity of its closest G7 peer, though this lead narrows when adjusting for GDP or population. In Japan and the United Kingdom, infrastructure exists but is not broadly accessible, especially for small and medium sized enterprises who struggle to compete for compute access with larger firms.
• Skills and Training Gaps: Most AFEs report shortages of workers with AI expertise (Canadian Chamber of Commerce 2024; Oliveira-Cunha et al., 2024; Hoffman and Nurski 2021). Firms also struggle to retrain existing employees for AI tasks, indicating a general mismatch in skills. As previously shown in Figure 6, for some sectors like manufacturing, AI skill penetration in several G7 economies is at or below the global average.
• Access to Equity and External Investment: In the United Kingdom and Euro Area, many firms cite limited access to equity investment as a barrier (European Commission 2025). The issue is particularly acute for SMEs in Europe, which cite high upfront costs to AI implementation and more binding credit constraints than larger firms (Hoffman and Nurski 2021). In the European case, one avenue to alleviate these constraints could be the deepening of the internal market and further progress on the savings and investment union (Airaudo et al. 2025).
• Energy costs: High electricity prices, particularly in Europe, might constrain adoption. As shown in Figure C1, country level data from the past two years show a slight negative correlation between energy costs and AI use among firms, with the effect significantly stronger for large enterprises. This trend is complicated by the fact that increased AI infrastructure buildout is causing a surge in electricity demand, putting upward pressure on prices.

Table Appendix D: Brief Literature Review of AI's Effects on Productivity and Employment

References

• Airaudo, Florencia, François de Soyres, Alexandre Gaillard and Ana Maria Santacreu. 2025. « Recent Evolutions in the Global Trade System: From Integration to Strategic Realignment". ECB Forum on Central Banking, Sintra, July 2025.
• Beraja, Martin, Wenwei Peng, David Y. Yang, and Noam Yuchtman. 2024. "Government as Venture Capitalists in AI." NBER Working Paper 32701. July.
• Chan, Kyle, Gregory Smith, Jimmy Goodrich, Gerard DiPippo, and Konstantin F.Pilz. 2025. "Full Stack: China's Evolving Industrial Policy for AI." RAND Corporation. June 26.
• Chen, Caiwei. 2025. "China built hundreds of AI data centers to catch the AI boom. Now many stand unused." MIT Technology Review. March 26. https://www.technologyreview.com/2025/03/26/1113802/china-ai-data-centers-unused/
• Crane, Leland, Michael Green, and Paul Soto. 2025. "Measuring AI Uptake in the Workplace," FEDS Notes. Washington: Board of Governors of the Federal Reserve System, February 05, 2025.
• Dobbs, Graham and Jake Hirsch-Allen. 2024. "Canada's plans to bridge the AI compute gap and how it can make industry policy inclusive and sustainable." AI Wonk Blog, OECD. April 16.
• Dutta, Soumitra, Bruno Lanvin, Lorena Rivera Leon, and Sacha Wunsch-Vincent.(2025). "Global Innovation Index 2024: Unlocking the Promise of Social Entrepreneurship (PDF)." World Intellectual Property Organization.
• European Commission. 2025. "The EU invests in artificial intelligence only 4% of what the U.S. spends on it." January 11.
• Gil, Yolanda and Raymond Perrault. 2025. "Artificial Intelligence Index Report 2025." Stanford University.
• Gill, Patrick and Andrew DiCapua. 2024. "Prompting Productivity: Generative AI Adoption by Canadian Businesses." Canadian Chamber of Commerce.
• Greco, Salvatore, Alessio Ishizaka, Menelaos Tasiou, and Gianpiero Torrisi. 2023. "On the Methodological Framework of Composite Indices: A Review of the Issues of Weighting, Aggregation, and Robustness." Social Indicators Research Vol 141 (2023): 61-94.
• Heim, Lennart. 2025. "China's AI Models Are Closing the Gap - but America's Real Advantage Lies Elsewhere." RAND Corporation. May 2.
• Hoffman, Mia and Laura Nurski. 2021. "What is Holding Back Artificial Intelligence Adoption in Europe?" Policy Contribution Issue no. 24/21, Bruegel. November.
• International Energy Agency. 2025. "Energy and AI (PDF)." World Energy Outlook Special Report. April.
• Li, Guajiao. 2024. "Behind the Technology Screen: Japan's Silent Progress in AI." Acuity Knowledge Partners.
• OECD/BCG/INSEAD. 2025. The Adoption of Artificial Intelligence in Firms: New Evidence for Policymaking, OECD Publishing, Paris.
• Oliveira-Cunha, Juliana, Bruno Serra-Lorenzo, and Anna Valero. 2024. "What an LSE-CBI survey found about AI adoption in UK firms." LSE Business Review Blog.
• Omaar, Hodan. 2024. "How Innovative is China in AI?" Information Technology & Innovation Foundation. August.
• Pilz, Konstantin, Yusuf Mahmood, and Lennart Heim. 2025. "AI's Power Requirements Under Exponential Growth." RAND Corporation. January 28.
• Pilz, Konstantin, Robi Rahman, James Sanders, Lennart Heim, 'Data on GPU Clusters'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/data/gpu-clusters' [online resource]. Accessed 25 Aug 2025.
• Reuters. 2025. "China plans network to sell surplus computing power in crackdown on data centre glut." July 24. https://www.reuters.com/technology/china-plans-network-sell-surplus-computing-power-crackdown-data-centre-glut-2025-07-24/.
• Schnabel, Isabel. 2024. "From Laggard to Leader? Closing the Euro Area's Technology Gap." Inaugural Lecture of the EMU Lab at the European University Institute. February 16.
• Westgarth, Tom, Melanie Garson, Kayla Crowley-Carbery, Alex Otway, Joseph Bradley, and Jakob Mokander. 2024. "State of Compute Access 2024: How to Navigate the New Power Paradox." Tony Blair Institute for Global Change. November 18.
• Wu, Zijing and Eleanor Olcott. 2025. "Nvidia AI Chips Worth $1bn Smuggled to China after Trump Export Controls." Financial Times. July 25. https://www.ft.com/content/6f806f6e-61c1-4b8d-9694-90d7328a7b54.

Data

• EpochAI GPU Cluster Dataset.
• EUKLEMS & INTANPROD - Release 2025.
• Eurostat.
• International Monetary Fund AI Preparedness Index.
• OECD AI.
• Oxford Insights. Government AI Readiness Index 2024. https://oxfordinsights.com/ai-readiness/ai-readiness-index/
• Stanford 2025 AI Index Report Data. https://drive.google.com/drive/folders/1AxxxL9-AsaeMdDKtTNHCR1KqEJTsHCod
• Tony Blair Institute for Global Change.
• Top500.
• United States Census Business Trends and Outlook Survey.
• World Intellectual Property Organization. Global Innovation Index.