The global artificial intelligence (AI) economy is generating more than $175 billion in revenue annually, with the generative AI sector alone recording around $110 billion in sales over the past 12 months. The global AI market is currently valued at more than $610 billion, while private and corporate investment in the sector soared to $252.3 billion, driven primarily by massive capital spending by tech giants.
The AI sector is likely to expand rapidly in the coming years, with Statista projecting the global AI market to reach $1.42 trillion by 2032, while the UN Trade and Development (UNCTAD) estimates that it could surge to $4.8 trillion by 2033.
As the AI economy rapidly develops, the more relevant unit is increasingly the token — the tiny fragment of text or data that AI models process to generate answers, predictions and decisions. As governments, businesses and individuals increasingly rely on AI services, countries are beginning to face a basic question: what does it mean when a nation’s access to AI depends on companies whose pricing, data policies, and availability are ultimately governed by another country’s regulatory environment?
Read: Alibaba launches Wan3.0 AI video model after $10 billion share sale
AI dependence has at least three dimensions: cost, data, and access.
Large language models process information in tokens, and users pay according to the number of tokens consumed. Mainstream AI service providers currently charge roughly $2.50 per million tokens, but the scale of future demand could radically change token expenditure into a significant component of national economies.
According to Goldman Sachs estimates, global token consumption could reach 120 quadrillion tokens a month by 2030. Under current adoption assumptions, Germany’s annual expenditure on token-based AI services could reach about €157 billion, equivalent to 3.51% of its GDP, per a study conducted by a German research institute.
That figure is an extrapolation rather than a forecast written in stone. Yet it shows the potential scale of what could become an “intelligence-as-a-service” economy. For countries without domestic AI infrastructure, rising consumption could mean a recurring outflow of foreign exchange.
This emerging pattern has been described by some researchers as a potential “digital trade deficit”. Traditional balance-of-payments statistics do not yet fully capture it, but the fundamental dynamic is simple: as domestic users consume foreign AI services, revenue flows towards the firms that own the models and infrastructure.
According to estimates, American firms control over 80% of the global enterprise AI market. In Canada, about 85% of cloud resources are controlled by three US companies: Amazon, Microsoft and Alphabet. Similar monopolisation exists in many other markets too. The result: countries can effectively end up leasing their AI stack rather than owning it.
The second issue is data. Every interaction with an AI model can involve prompts, documents and contextual information being processed by the service provider. For businesses, this could include customer information, proprietary documents and internal processes. For governments, it could involve archives, administrative records and other institutional knowledge. This creates paradoxical dependence: users pay for access while at the same time generating information that can contribute value to the systems they use.
At the commercial level, this issue came to the fore in 2026 when startup DigitalApplied used Anthropic’s Claude API to develop a feature allowing coding sessions to be exported as shareable HTML pages. Shortly before its launch, Anthropic introduced an equivalent feature into its own product, effectively absorbing the startup’s differentiation.
Read More: As AI agents act in unexpected ways, is ‘rogue AI’ really here?
The concern is broader at the national level. If a country’s linguistic, cultural, and institutional information is processed through foreign AI systems, decisions about how that information is represented and reproduced are influenced by the providers’ models, training choices and policies. Some analysts describe this as a form of “epistemic dependence” — a situation in which a country’s knowledge, language, and institutional information are increasingly mediated through foreign systems.
The third dimension is access. The June 2026 episode involving Anthropic showed how geopolitical and regulatory decisions can affect the availability of frontier AI. On June 12, Anthropic announced that, in compliance with a US government export-control order, it would suspend access to its Fable 5 and Mythos 5 models for foreign nationals, including those in the United States and some of its own employees. Some of the curbs were subsequently lifted on June 27 for more than 100 American institutions.
This points to an alarming structural vulnerability: AI services supplied by American companies remain subject to US export-control authority. That creates a supply-chain risk similar to other forms of cross-border technological dependence for organisations that have incorporated such systems into essential operations.
The exposure varies across regions. The exposure varies across regions. For the Global South, including countries in Asia, Africa and Latin America, the problem is primarily affordability and infrastructure. For Europe, it is autonomy and data governance. For US allies in Asia, it increasingly has a strategic and defence dimension.
That does not mean dependence is inevitable. There could be three options: open-source models, sovereign AI infrastructure, and regional cooperation.
Open and open-weight models can allow countries and companies to deploy AI on their own infrastructure, reducing data-residency concerns and exposure to external access restrictions. Some Chinese-developed models have also emerged at substantially lower token costs, while the performance gap between US and Chinese frontier models has narrowed substantially.
Sovereign AI infrastructure offers another route through domestic computing capacity, locally trained models and nationally curated knowledge systems. Regional cooperation could complement these efforts through shared computing facilities, multilingual datasets, and local-currency settlement mechanisms. None comes without costs. Building domestic AI capacity requires massive investment, while regional cooperation requires political coordination and technical compatibility.
Also Read: Chinese teenagers flock to tech frontline to touch the future amid AI study-tour boom
The strategic choice, therefore, is not between complete independence and total dependence. It is about deciding where dependence is acceptable and where resilience is essential.
As AI becomes infrastructure rather than simply software, the countries that control compute, models and data will possess an increasingly important source of economic power.
The token may be tiny. The sovereignty question behind it is not.
from Latest Technology News, Tech News Pakistan | The Express Tribune https://ift.tt/yZ1Szdb
via IFTTT
0 Comments