August 25, 2026: The global artificial intelligence economy is expanding at extraordinary speed, with annual revenue now exceeding $175 billion and generative AI alone generating around $110 billion in sales over the past year.

The global AI market is valued at more than $610 billion, while private and corporate investment has reached $252.3 billion, driven largely by enormous spending from major technology companies.

As AI economy expands, however, countries are facing a new strategic question: how much control can they retain when businesses, governments and citizens increasingly depend on AI services operated by companies based in other countries?

The issue is moving beyond technology. It is becoming a question of economic security, data protection and national sovereignty.

The Rising Cost of AI Dependence

Artificial intelligence systems process information through tokens, small units of text or data used by AI models to generate responses and make predictions.

AI providers generally charge customers according to token consumption. While current prices may appear relatively low, the sheer volume of future usage could transform AI spending into a significant economic cost.

Goldman Sachs estimates that global token consumption could reach 120 quadrillion tokens every month by 2030.

A German research institute has estimated that Germany could eventually spend about €157 billion annually on token-based AI services under current adoption assumptions. That would equal around 3.51% of the country’s GDP.

The figure is an estimate rather than a fixed prediction. But it illustrates how AI could create a new form of recurring international expenditure.

For countries without strong domestic AI infrastructure, increasing reliance on foreign platforms could result in significant foreign-exchange outflows.

A New Digital Trade Deficit

Researchers have begun describing this emerging phenomenon as a potential “digital trade deficit”.

Traditional economic statistics do not yet fully capture the movement of money associated with AI services. The underlying dynamic, however, is straightforward.

When businesses and government institutions use foreign AI platforms, money flows toward the companies that own the models, cloud infrastructure and computing resources.

American companies currently account for more than 80% of the global enterprise AI market, according to estimates cited in the analysis.

Cloud infrastructure is also highly concentrated. In Canada, around 85% of cloud resources are controlled by Amazon, Microsoft and Alphabet.

Such concentration means many countries could end up effectively renting their AI infrastructure instead of developing and controlling it themselves.

Data Is the Second Challenge

Cost is only one part of the problem.

Data represents another major source of dependence.

Every interaction with an AI system can involve processing prompts, documents and contextual information. For companies, that information could include customer records, proprietary documents and internal business processes.

Government agencies could potentially process administrative records, archives and institutional knowledge through foreign AI platforms.

This creates an unusual relationship. Users pay companies for access to AI systems while simultaneously providing information that may hold significant economic or strategic value.

The concern extends beyond data security.

If a country’s language, culture and institutional knowledge are increasingly processed through foreign AI systems, the way that information is represented could be influenced by the models, training methods and policies of overseas providers.

Some researchers describe this as “epistemic dependence” — a situation in which a country’s knowledge and institutional information become increasingly mediated through external technology.

Geopolitics Can Change AI Access

The third major issue is access.

Developments involving Anthropic in June demonstrated how geopolitical decisions can potentially affect access to advanced AI systems.

The company announced restrictions following a US government export-control order affecting access to certain models. Some restrictions were later eased for more than 100 US institutions.

The episode highlighted a broader vulnerability.

AI services supplied by American companies remain subject to US regulatory and export-control frameworks. Organisations that build essential operations around those services could therefore face risks if geopolitical conditions change.

A company may have access to an AI model today but could potentially lose that access tomorrow because of regulatory decisions made in another country.

Developing Countries Face a Different Risk

The impact of AI dependence is not identical everywhere.

For countries across the Global South, including much of Asia, Africa and Latin America, affordability and computing infrastructure are major concerns.

Europe faces greater questions around technological autonomy and data governance.

For US allies in Asia, dependence on foreign AI systems is increasingly connected to strategic and defence considerations.

The common issue is control.

Countries that rely heavily on external providers may have limited influence over pricing, availability, data policies and future technological development.

Venice Film Festival Adds Gaza Documentary NAZA to Golden Lion Competition

Three Routes Toward AI Resilience

Complete technological independence may not be realistic for every country. But governments and businesses have several options to reduce excessive dependence.

Open-source and open-weight AI models can allow organisations to run systems on their own infrastructure. This can reduce data-residency concerns and provide greater protection against external access restrictions.

Sovereign AI infrastructure offers another option. Countries can invest in domestic computing capacity, locally trained models and national knowledge systems.

Regional cooperation could provide a third path.

Countries could share computing infrastructure, develop multilingual datasets and explore local-currency payment mechanisms for AI services.

Each approach has challenges.

Building domestic AI infrastructure requires substantial investment. Regional projects also require political cooperation, technical standards and long-term funding.

AI Is Becoming Infrastructure

The strategic question is therefore not whether countries can eliminate all foreign AI dependence.

Instead, governments must determine where dependence is acceptable and where technological resilience is essential.

As artificial intelligence moves from being a specialised software tool to becoming part of economic and public infrastructure, control over computing power, AI models and data will become increasingly important.

The token used by an AI model may be tiny.

The economic and sovereignty questions behind that token are becoming much larger.