The AI revolution may be unfolding at breakneck speed, but not everyone is riding the same wave. In its newly published Economic Index, frontier AI lab Anthropic — creator of the Claude family of large language models — provides a data-driven look at the global and enterprise-level uptake of AI, revealing wide disparities across countries, industries, and socioeconomic lines.
The report is more than a vanity metric on usage. It sheds crucial light on where, how, and why certain regions and companies are embracing AI rapidly — while others risk being left behind.
Measuring AI Impact: Anthropic’s Methodology
Anthropic constructed the Economic Index using two key pillars:
- Consumer Interaction Metrics: Measured via per capita use of Claude.ai across countries and U.S. states, normalized by working-age population.
- Enterprise API Usage: Captures task volumes and complexity through Claude’s API integrations, indicating which business functions are being delegated to AI.
This dual approach allows Anthropic to distinguish between mass-market consumer use and strategic enterprise adoption, offering rare insight into AI’s real-world diffusion.
High-Income Countries Lead the AI Pack
The most immediate and striking conclusion is that AI usage correlates strongly with GDP per capita. Wealthier nations are not only early adopters — they’re accelerating faster.
- Singapore emerged as the global AI powerhouse, clocking nearly 4.6 times the expected usage given its population.
- Canada followed closely with 2.9× expected usage, alongside Germany, Australia, and the U.K. showing robust adoption.
- Meanwhile, major developing economies like India (0.27×), Indonesia (0.36×), and Nigeria (0.2×) remain well below expected usage thresholds, despite large populations and growing tech sectors.
Anthropic attributes these divides to a mix of factors — including access to high-speed internet, affordability of compute resources, language localization, and digital literacy.
U.S. Breakdown: Utah and D.C. Dominate AI Usage
Within the United States, the disparities are equally pronounced:
- Washington, D.C. and Utah top the leaderboard with 3.82× and 3.78× expected per capita usage, respectively. Utah’s rise is especially notable, reflecting a strong startup ecosystem and tech-forward education system (Axios report).
- California (home to Silicon Valley) and New York maintain high engagement but are notably behind Utah in relative usage.
- States in the Southeast and Midwest tend to under-index, suggesting economic, educational, or infrastructure gaps.
This usage data doesn’t just reflect access — it signals how embedded AI already is in the professional and personal workflows of these regions.
AI at Work: Enterprise API Use Centers on Automation
The enterprise API adoption metrics tell a slightly different story — one of increasing trust in AI to not just assist, but to act:
- A striking 77% of API usage involves full automation of tasks (as opposed to co-piloting or augmentation).
- Enterprise usage skews heavily toward coding, admin work, and technical writing, with far less uptake in fields like education, legal reasoning, or creativity — areas more common on the consumer-facing Claude.ai.
- API users also pay more per task, suggesting that firms are investing in high-value, mission-critical AI deployments.
Yet, there’s a catch: the most sophisticated enterprise use cases often stall due to lack of integration infrastructure. According to Anthropic, the main constraint is organizational readiness — firms that can’t provide structured context or real-time feedback see lower returns on AI.
Usage Patterns Reveal Automation vs Augmentation Divide
One of the most intriguing findings in the report is a behavioral pattern shift among AI users. Over the past year, both individuals and businesses have moved from using Claude as a collaborator to treating it as a full agent:
- Between January and September 2025, the share of “directive” (i.e., full-delegation) conversations rose from 27% to 39%.
- Lower-AI-Index countries tend to rely more on automating coding tasks, while higher-usage countries spread AI across education, business planning, and scientific exploration.
- The richer the context a user or firm can give Claude, the broader and more efficient the AI’s performance becomes.
This shift points to a growing cultural and technical confidence in AI’s autonomy.
Global Risks: Will AI Widen Economic Inequality?
The data paints a worrying picture for those concerned with global equity and technological inclusion. Just as previous industrial revolutions concentrated early gains among the few, AI appears to be reinforcing — not reversing — global and intra-national inequalities.
- Regions lacking infrastructure, funding, or skilled labor fall behind not just in adoption, but in economic competitiveness.
- High-AI-usage countries and companies will likely see faster productivity growth, attracting more capital and talent.
- Without intervention, this could trigger a feedback loop where the AI-rich get richer — and the AI-poor stagnate.
What Needs to Change: Policy, Access, and Literacy
Anthropic’s findings raise pressing policy questions:
- Can emerging economies catch up?
Public and private investment in broadband, compute, and AI literacy programs is urgently needed. - Will enterprises become AI monopolists?
If only the largest firms can afford deep integrations, they’ll outpace smaller rivals — exacerbating economic concentration. - How can AI usage be democratized?
Incentives for open-source alternatives, subsidized access, and public cloud infrastructure could help level the field.
Importantly, Anthropic has pledged to triple its international workforce in the next year to support broader global access (Reuters report). Whether that includes regional language models, local servers, or education programs remains to be seen.
Final Thought: The Next Phase of AI is Distribution
The Claude-powered AI boom is no longer just about what AI can do. It’s increasingly about who gets to use it, where, and how effectively. Anthropic’s Economic Index provides one of the clearest signals yet that the next frontiers of AI progress won’t be solely technological — they’ll be geographic, social, and political.
Without action, the divide between AI haves and have-nots could harden into an economic fault line.
