New usage data from Anthropic and OpenAI indicates that AI adoption is accelerating, but unevenly. Affluent nations and tech-concentrated U.S. states are seeing much higher usage, while many poorer countries and less connected regions lag significantly. For investors, this isn’t just a socioeconomic story—it points to where demand, infrastructure, and regulatory tailwinds are likely to emerge.
Key Findings
From the recent reports:
- Task Usage Differences: Claude (Anthropic) users are heavily weighted toward “computer & mathematical tasks”—around 36% of usage—whereas OpenAI’s ChatGPT is more used for information lookup (~18%) as well as writing and teaching assistance. This suggests specialization in use cases depending on user base.
- Geographic Concentration:
- Wealthier countries (Israel, Singapore, Canada, U.S.) show higher per-capita usage relative to their working-age populations.
- Within the U.S., states like California, D.C., Virginia are leaders; southern and plains states are lagging.
- Implication of Inequality: The data reinforces concerns that AI might widen economic inequality: those with better infrastructure, higher educational attainment, digital literacy, and access to fast internet are already pulling ahead.
Broader Context & What It Signals
To understand why this matters, here are important contextual factors:
- Early Stage Diffusion: AI is still in early phases of diffusion. New technologies historically start concentrated in educated, urban, rich regions before spreading out. Infrastructure (internet, devices), regulatory clarity, education, and access drive that lag. The reports confirm that pattern.
- Task Bias & Commercial Use-Cases: Regions or users that lean toward more complex, technical tasks (coding, mathematical work) may extract more economic value, higher productivity gains, than those using AI for simpler tasks (information lookup). So differences in usage reflect not just usage volume but potential value per user.
- Skill & Education Gaps: Regions with higher tech industry presence and more skilled workforce are likely not only to adopt AI faster but to benefit more—through augmentation of work, automation of repetitive tasks, and scaling of high value tasks.
- Regulatory & Policy Differences: Countries with clearer regulation, data privacy laws, investment incentives for AI infrastructure, or governmental support tend to enable faster adoption. Conversely, regulatory ambiguity or connectivity constraints hamper adoption.
- Potential Market Size vs Adoption Barriers: Though under-penetrated, many lower-income geographies represent latent demand. Once broadband, device access, cost, language models localized, etc., improve, these markets could see strong growth.
Investment Plays to Consider
As an investment manager, here’s where I see potential strategic plays, and how to position:
| Play Area | Why It Looks Attractive | How to Execute / Companies to Watch |
|---|---|---|
| Infrastructure & Connectivity | Under-served geographies with low AI usage often suffer from poor connectivity, high device costs, or lack of data center infrastructure. Improving those is a prerequisite for adoption. | Invest in companies building out broadband, edge computing, compact data centers, and local cloud/AI hosting. Also hardware/device suppliers for low-cost smartphones or low-power compute. Partnerships with telcos, governments. |
| Localized AI Services & Models | Global models trained for major markets may not serve language, cultural, regulatory, or task needs in smaller or developing economies. Local specialization could unlock higher per-user value. | Firms that build or fine-tune AI models per region, in local languages. Also potentially companies offering cost-efficient inference/prediction stacks optimized for low resource settings. |
| Enterprise / API Adoption in High Growth Regions | Even in lower use places, enterprise demand may pick up faster once infrastructure is in place, because business productivity gains are compelling. Cloud providers, SaaS vendors embedding AI offer leverage. | Watch cloud providers expanding regionally (data centers), enterprise AI vendors (customer service, internal tools, automation), startups in emerging markets adopting AI to scale. |
| Education / Reskilling Providers | Skill gaps (digital literacy, AI tool fluency) are going to be a major constraint. Firms that help train or provide tools for adopting AI will be in demand. | EdTech platforms, bootcamps, online skill courses, AI tool providers with educational verticals. Governments may subsidize; opportunity for public-private partnership. |
| AI-Efficiency & Edge / Low Resource Models | Models that require less compute, bandwidth, or expensive GPUs will be essential for adoption in poorer or more remote regions. Those that reduce inference cost per user may unlock new addressable markets. | Invest in startups or incumbents optimizing model sizes (pruning, quantization), hardware accelerators, or edge inference. Companies like those building compact LLMs or optimizing inference pipelines. |
| Regulation & Policy Tech / AI Governance | As adoption increases, so will regulatory scrutiny. Companies that help other firms comply (data privacy, safety, fairness) may have tailwinds. Also, firms pushing for standards / transparency may benefit reputationally. | Legal-tech firms, compliance SaaS, model auditing tools. Stakeholders engaging in policymaking; firms with strong governance practices may see premium valuations. |
Risks & Counterweights
A nuanced view requires recognizing what might go wrong, or limit upside:
- Infrastructure Bottlenecks: Even if device cost and software models become cheaper, without reliable internet, electricity, data centers, many regions remain limited. Investment takes time.
- Regulation & Localization Hurdles: Some governments may restrict AI use, either for political or security reasons. Localization of language/data, content moderation, censorship constraints, etc., can slow growth or lead to fragmented markets.
- Cost vs Monetization Mismatch: High usage doesn’t always equate to high revenue. If users in lagging regions are lower income, monetization per user may be much lower. Companies need business models tailored to lower willingness/ability to pay.
- Competition & Copycats: Big players with deep resources (OpenAI, Anthropic, Google) may move aggressively to capture new markets, lowering entry barriers and squeezing smaller, localized players.
- Education & Talent Constraints: Deploying AI well requires trained users, engineers, prompt-engineers, local support. Without sufficient education or talent pipeline, usage may be superficial and lower productivity gains.
- Uneven Productivity Gains: Even within high-usage regions, gains may be concentrated in certain sectors (tech, education, professional writing, programming) and leave others behind, potentially worsening inequality rather than distributing gains evenly.
What To Watch, Key Metrics
To track whether this divide is narrowing or widening—and to signal which investment plays gain momentum—monitor:
- Changes over time in AI usage per capita in emerging markets / lower-income countries.
- Increases in infrastructure investment: broadband access, mobile internet quality, local data centers, edge compute deployment.
- Growth in AI model localization: multilingual models, regionally focused agents, apps in local languages.
- Enterprise AI adoption in these regions: number of API calls, enterprise customers outside major tech hubs.
- Policy actions: subsidies for AI infrastructure, digital literacy programs, regulation that enables or restricts AI deployment.
- Monetization metrics: average revenue per user (ARPU) in different geographies, willingness to pay, freemium vs paid adoption.
Example Hypothetical Scenarios
- Base Case: Infrastructure improvements, moderate expansion of usage into middle-income regions. Growth in enterprise AI adoption. Marginal clearing of digital divide. Key beneficiaries: cloud providers, enterprise SaaS, efficiency startup tools.
- Upside Case: Rapid improvements in connectivity, AI tools optimized for low-resource usage, strong policy support (education, subsidies, regulatory clarity). Local AI ecosystems emerge (startups, localized content, models). Significant ARPU lift from new markets.
- Downside Case: Regulatory backlash (data laws, export controls), infrastructure lag, models that remain resource-intensive and expensive to deploy. Big incumbents dominate, and smaller/new entrants or local players cannot compete. Divide persists or even worsens.
A Potential Investors Take
As someone managing capital with an eye toward AI’s long wave, this story is a key red flag and opportunity:
- The data reinforces that AI is not a uniform tailwind; it’s a driven opportunity, where geography, infrastructure, policy, and skills matter as much as the technology.
- Portfolios should include exposure to both core infrastructure (cloud, data centers, efficient inference hardware) and enablement plays (education, localization, AI-optimization) — especially in under-served regions.
- Valuation assumptions for AI product companies should explicitly model differential adoption rates across geographies. Companies that claim universal addressable markets without regard for infrastructure or regulatory constraints may have overly optimistic projections.
- Investors should monitor policy risk: subsidies, regulation, export controls, internet governance. Favor companies with adaptive strategies and diversified geography.
- Lastly, as with any tech adoption curve, expect “laggards” to offer opportunities — companies or regions currently under-penetrated may have high upside if the right enabling conditions materialize.
