Servers in a data centre used by Wikimedia projects. Image by Helpameout, licensed under CC BY-SA 3.0.
By turning data, computing and knowledge into instruments of decision-making, artificial intelligence has become a question of international political economy. Its value appears when prediction, classification and automation reduce coordination costs across different sectors. These gains require a material base. Servers, chips and energy define the technical reach of AI. Reliable data, specialists and rules for linguistic adaptation define its social usefulness.
The economic contest around AI goes beyond the question of which countries will “use” the technology. The decisive issue is who controls the inputs that make its use possible. Companies that concentrate models, cloud services and distribution channels can capture a rising share of the value created by automation. States that control semiconductors and standards can turn that capacity into industrial and diplomatic advantage. The result is ambivalent: the same technology that opens development opportunities can deepen dependence among countries, firms and workers.
Summary
- AI can raise productivity and improve public policy. To do so, it needs digital infrastructure, reliable data, energy, skills and institutions able to assess risks.
- Global inequality in AI comes less from access to a single tool than from the concentration of compute, capital, dominant languages and regulatory capacity.
- In labour markets, AI tends to automate tasks and reorganize occupations. Its effects depend on training, social protection and business adaptation.
- Competition among the United States, China and the European Union shows that AI is industrial policy, supply-chain control and a struggle over technical standards.
- International governance combines universal organizations, economic forums and regional initiatives. The main gap lies in financing, participation and implementation.
- For developing countries, the central agenda is to turn access to AI into local productive capacity, rather than merely consuming foreign platforms.
Why AI Is An International Economic Issue
Artificial intelligence affects the global economy because it changes how information becomes decision-making. In companies, AI models help forecast demand and organize inventories. In governments, they can support risk triage, public-service translation and document analysis. In production systems, they connect machines and logistics chains. The economic effect comes from a common mechanism: reducing the costs of search, prediction and coordination.
This capacity matters to countries because productivity is never only technical. An economy that incorporates AI widely can produce more with the same resources, raise service quality and create new sectors. Adoption depends on accumulated conditions. Connectivity and stable electricity support everyday use. Firms, universities and digital government turn that use into learning. Without that base, AI remains a superficial layer: users access imported tools, while the country does not control technology, data or industrial learning.
The distinction between use and capacity is central. A ministry can purchase a foreign system to classify documents. That can improve routine work without creating a national AI base. Capacity emerges when local technicians understand the system, adapt data and test for bias. Public procurement completes the circuit by creating demand for infrastructure and skills. Without that local layer, modernization remains apparent and the dependent position in the digital chain persists.
The Material Base Of AI
The public image of artificial intelligence often focuses on models and interfaces. Its material base has three layers: data centres, communication networks and semiconductor chains. Generative AI systems require training on large volumes of data and continuous operation on high-capacity servers. The user sees a textual answer. Behind it lies a physical chain of energy, cooling and cloud infrastructure.
This infrastructure changes the international distribution of power. Countries that host cloud companies and chipmakers concentrate instruments of economic influence. They can restrict access to components, define security standards and shape how other countries use digital services. Countries with abundant renewable energy and good connectivity, in turn, can compete for data centres. In other words, the ability to host computing at scale becomes part of economic competition.
There are environmental and geopolitical costs. Data centres consume electricity and require cooling. Chip production depends on specialized equipment, ultrapure water and concentrated industrial chains. Critical minerals enter the electrical and digital base that sustains this infrastructure. When demand for AI grows quickly, these chains become objects of industrial policy and export controls. In this sense, AI is not an immaterial technology: it increases the diplomatic importance of energy, semiconductors and digital infrastructure.
Productivity And Development
For developing countries, the promise of AI is real when it connects to concrete problems. In agriculture, models can improve climate forecasting and water management. In health, they can support triage and image reading. In public services, they can organize documents and identify fraud. In industry, they can bring maintenance, quality control and production planning closer together.
These applications work well only when there are local data, institutions able to interpret them and people responsible for decisions. A model trained on another country’s data can fail in diagnosis, language or administrative priority. A tool applied to a public service can reproduce inequality if previous data records exclusion, informality or discrimination. The practical rule is simple: AI improves public policy when it increases state capacity and worsens inequality when it automates poor procedures.
This distinction explains the importance of digital public infrastructure. Digital identity, payments and interoperable registries give the state an organized basis for using AI under public control. The Global Digital Compact adopted by the UN in 2024 treats digital public goods, digital public infrastructure and open models as instruments of inclusion. The goal is to reduce exclusive dependence on proprietary platforms. When governments can organize data and services safely, AI can support development instead of only extracting value from users.
Work, Skills And Inequality
AI’s impact on work does not appear as a uniform replacement of people by machines. The change tends to affect tasks. Some administrative and service activities can be automated. In other occupations, tools accelerate search, writing and diagnosis. The outcome depends on how each society organizes work, social negotiation and training.
In 2024, the IMF estimated that about 40% of global employment is exposed to AI, with higher exposure in advanced economies. The ILO found that generative AI is more likely to augment than to destroy entire jobs, although it can change intensity and autonomy at work. These diagnoses point to the same mechanism: AI redistributes advantages between workers who can use it and workers trapped in replaceable or poorly paid tasks.
This process can deepen inequality within countries. Workers with education and connectivity can gain productivity. Workers without training can lose entry-level tasks or face algorithmic surveillance. Women may be more affected in administrative sectors where they are overrepresented. Younger workers tend to adapt faster, while older workers face higher retraining costs. Public policy needs to treat AI as a labour transition. The innovation agenda works only when it incorporates that social dimension.
Corporate Concentration And Platform Power
The economics of AI favours concentration as advanced models require scale. Training competitive systems demands capital, specialized teams and access to cloud-based chips. Few companies can combine those elements. As a result, many businesses use AI through APIs and platforms supplied by large foreign companies. Entry becomes easier for final users. Dependence on central providers increases.
This dependence appears at three levels. Economically, part of the value generated by local firms pays for licences and proprietary models. Informationally, data and usage patterns can remain under the control of external platforms. In regulatory terms, public authorities without sufficient technical access struggle to audit bias and security. When the AI layer becomes essential infrastructure, control over digital platforms becomes both market power and political power.
Intellectual property and data reinforce the problem. Companies want to protect models and training sets. Societies need transparency, contestation and protection against discrimination. The tension is real: full openness can expose security and privacy. Full closure blocks public audit. The economic governance of AI must calibrate these interests so that innovation does not become a permanent argument against accountability.
The Geopolitics Of Chips, Data And Cloud
Competition between the United States and China shows how AI has become a matter of economic security. The United States concentrates leading firms in chips, software, cloud services and frontier models. Its influence increases through alliances at decisive points of the semiconductor chain. China invests in technological autonomy, national platforms and data centres. Between these poles, the European Union tries to use its market to define obligations on transparency, risk and fundamental rights.
Export controls on advanced semiconductors reveal the logic of this contest. High-capacity AI models depend on specialized chips and manufacturing equipment concentrated in a few countries. Limiting access to these inputs can slow competitors and protect military advantages. In response, affected countries seek domestic substitution and partnerships. The AI supply chain becomes an instrument of foreign policy.
Data and cloud services complete the dispute. States want to protect sensitive data and attract global digital services. Companies want to operate at transnational scale. Middle-income countries try to avoid dependence on a single technological sphere. This dilemma is concrete: when a government hosts public services in foreign clouds or buys closed models without audit capacity, its digital choices start to depend on other countries’ rules and corporate decisions.
Global Economic Governance
International governance of AI has advanced in layers. UNESCO adopted a global recommendation on the ethics of AI in 2021, grounded in human rights, human oversight and inclusion. The OECD and the Global Partnership on AI contributed principles for trustworthy AI. The G7 developed the Hiroshima Process for advanced models. In 2024, the UN adopted resolutions on safe AI and international capacity-building, while also including the issue in the Global Digital Compact.
The Global Digital Compact is especially relevant to the political economy of AI because it links technical safety to development. It calls for representation of developing countries, data cooperation and interoperable standards. The agenda also includes accessible computing, open models and voluntary financing to reduce AI divides. The political shift lies in asking who will have the capacity to participate in the benefits and the rules.
Other forums enter through specific angles. The G20 discusses innovation and digital infrastructure. The WTO observes how AI affects trade in services, intellectual property and technical barriers. Development banks can finance connectivity, energy and digital government. These forums still do not form a single regime. There are too many principles, too little financing and a large distance between countries that design rules and countries that must implement them.
Brazil And The Global South
Brazil illustrates an intermediate position. The country does not control the global frontier of large models. Even so, it has a large market, universities, digital public-sector experience and a relatively clean electricity matrix. That base is completed by capabilities in agriculture, health and specific industrial sectors. The Brazilian Artificial Intelligence Plan 2024-2028 sought to connect AI to economic and social development. Its axes include public services, business innovation, infrastructure, training and Portuguese-language models. The ambition is right: without language, data and domestic capacity, the country consumes foreign systems without fully shaping them.
Brazil’s challenge shows the difficulty facing middle-income countries. Strategy must move from documents into budgets, public procurement and technical training. Research centres, data protection and independent evaluation give continuity to that effort. Policy must also avoid two extremes: blocking innovation through regulatory fear or accepting technological dependence as inevitable. An AI policy for development must combine rapid adoption with local learning.
For the Global South, the common agenda involves finance, South-South cooperation and digital public infrastructure. Local language models and access to computing complete this base. Small countries, island states and landlocked economies face even higher costs because they depend on connectivity and scale they may not possess. Inclusion, in this context, is not merely a seat at a conference. It is the effective capacity to test, adapt, contest and produce technology.
Limits
AI will not solve development problems on its own. It can improve an agricultural policy without replacing credit and infrastructure. It can support education without replacing teachers and connectivity. It can reduce administrative fraud without replacing institutional reform. When AI is treated as a shortcut, it tends to hide distributive conflicts that still exist.
There is a risk of dependence through efficiency. A foreign tool may be cheap and work well in the short term. Over time, the organization becomes dependent on its formats, updates and prices. Replacement becomes costly, data become locked in and internal capacity atrophies. This is a classic mechanism of technological dependence, now applied to cloud services, models and platforms.
Finally, governance faces a permanent political tension. States want cooperation to avoid harms and compete for industrial advantage. Companies defend responsible standards while seeking scale and profit. Developing countries demand inclusion, although they do not always have institutions able to implement complex rules. The likely result is fragmented governance, with national laws, technical standards and private contracts coexisting unequally.
Conclusion
Artificial intelligence reorganizes the global economy by changing the relationship between knowledge, infrastructure and power. Its benefits reach productivity, health, education, public administration and scientific research. Those benefits depend on material and institutional conditions that are unequally distributed. The economic future of AI will be decided less by the existence of the technology than by the distribution of compute, data, energy, talent, regulation and finance.
For diplomacy, the central issue is to prevent AI from creating a new divide between countries that control models and countries that merely provide data, markets and energy. That requires public capacity, competition rules and digital infrastructure. Scientific cooperation and financing complete responsible adoption. AI can expand development when it is treated as political and economic infrastructure, more than as a neutral product.
The decisive question is not whether artificial intelligence will be used in the global economy. It is already being used. The question is under which rules, with which inputs, for whose benefit and with what capacity for contestation. To govern AI is to contest how future productivity will be distributed among states, firms and societies.