As the world's leading powers invest hundreds of billions of dollars in artificial intelligence infrastructure, middle-income economies face a different challenge: how can they turn technology largely developed elsewhere into a source of national economic capability? The World Bank's 2026 World Development Report offers a framework for addressing this dilemma. For Morocco, which is pursuing its own industrial and technological ambitions, sovereignty may depend less on the sophistication of the models it possesses than on the strength of the institutions capable of using, adapting and controlling them.

A technological revolution that does not begin in the same place for everyone

In Rabat, in September 2026, the presentation of progress under the Morocco AI 2030 roadmap gave concrete expression to the kingdom's digital ambitions. The announced targets are substantial: generating MAD 100 billion in additional GDP value by 2030, creating 50,000 direct and indirect jobs, and training or certifying 200,000 people. Behind these objectives lies an emerging architecture of computing infrastructure, data centres, research institutes, training programmes and applications intended for both public institutions and private enterprises.

The strategy is taking shape as artificial intelligence becomes one of the principal arenas of industrial competition between the United States, China and a limited number of other technological powers. Investment is increasingly concentrated in semiconductors, data centres, electricity networks and foundation models. Training and operating the most advanced systems now requires a combination of capital, energy, researchers and infrastructure that relatively few countries can mobilise.

For a middle-income economy, the temptation is to reproduce this architecture on a smaller scale: establish domestic data centres, develop national models, attract technology manufacturers and gradually reduce dependence on foreign suppliers.

But this approach raises a fundamental economic question. Does owning technological infrastructure guarantee the ability to extract economic value from it?

Published on August 4, 2026, the World Bank's World Development Report 2026: The Promise of Artificial Intelligence offers a more nuanced answer. Developing countries do not necessarily need to build the world's most powerful models to benefit from artificial intelligence. They need the infrastructure, skills, data and institutions required to integrate these technologies into productive activities.

The report proposes a progression through three stages: adopting existing tools, adapting them to local needs and progressively advancing towards more sophisticated technological capabilities.

This distinction shifts the centre of gravity of digital sovereignty. It suggests that the defining divide between economies will not simply separate those that own AI models from those that import them. It will increasingly separate countries capable of converting artificial intelligence into productivity gains from those that remain primarily consumers of technological services.

The paradoxical advantage of middle-income economies

Artificial intelligence is spreading considerably faster than the major industrial technologies that preceded it.

According to the World Bank, the steam engine took approximately 80 years to reach low-income countries, electricity around 40 years and the internet roughly two decades. Six months after ChatGPT was launched, middle-income economies already accounted for half of its global traffic.

This acceleration changes the possibilities for technological catch-up. A company in Morocco, Vietnam or Indonesia can now access certain information-processing capabilities that, only a few years ago, would have required specialised teams and considerable investment.

Yet the diffusion of technology does not mean that its economic benefits are distributed automatically.

World Bank estimates reveal an important contrast. In low- and middle-income countries, 4.5% of jobs are classified as exposed to the risk of automation by generative AI, compared with 14.2% in high-income economies. Meanwhile, 16.2% of jobs in developing economies could experience significant productivity gains, relatively close to the 18.7% estimated for wealthy countries.

These figures describe exposure and potential, not observed job losses or productivity improvements. Nor are they specific forecasts for Morocco.

Nevertheless, they illuminate a structural characteristic of middle-income economies: a substantial share of employment remains concentrated in physical activities, local services, agriculture, construction and manufacturing. Such occupations are generally less directly susceptible to automation by generative models than certain administrative, analytical or document-intensive functions.

Artificial intelligence may therefore play a greater role in complementing human labour than in immediately replacing it.

An industrial technician equipped with diagnostic software, a physician supported by analytical tools, a farmer receiving agricultural forecasts adapted to local conditions, or a small business automating part of its commercial management can improve productivity without eliminating the underlying occupation.

The opportunity lies in this augmentation of human capabilities.

But realising it requires tools that are accessible, reliable and adapted to actual production conditions. An AI system capable of generating sophisticated analyses in English but unable to interpret local administrative documents, agricultural practices or national languages may have limited economic value for large segments of the population.

Technological development consequently becomes a question of social and institutional organisation as much as computing power.

Morocco's infrastructure decision

Morocco is not approaching this transformation without a strategy. Its Morocco AI 2030 roadmap, connected to the broader Digital Morocco 2030 programme, includes the development of sovereign digital infrastructure, a national network of specialised institutes and the integration of artificial intelligence into public administration and industry.

The planned Dakhla data centre illustrates the industrial dimension of these ambitions. Its announced target capacity could reach approximately 500 megawatts, with electricity expected to come from renewable energy sources. Another project in Rabat concerns a 50-megawatt facility intended, among other purposes, to support sovereign cloud services and the national research and innovation ecosystem.

These figures refer to planned capacities, not infrastructure that has already been fully constructed and brought into operation.

If implemented, such projects could strengthen Morocco's technological attractiveness, support hosting and computing activities, and create new opportunities for businesses and public institutions.

They also introduce a series of trade-offs.

A data centre represents an investment in property, electricity infrastructure, computing equipment and financing. Its economic viability depends on utilisation rates, the services it provides, energy costs, international connectivity and actual demand from customers.

It is also essential to distinguish activities frequently grouped under the same label. Conventional hosting, cloud computing, high-performance computing and the training of artificial intelligence models do not require identical equipment or expertise.

Substantial electrical capacity does not, by itself, guarantee the emergence of a domestic AI industry.

Sovereignty over data storage does not necessarily translate into sovereignty over the entire technological value chain. A data centre located on national territory may still depend on foreign processors, proprietary software, international cloud providers and overseas maintenance contracts.

The relevant question is not whether these dependencies can be eliminated entirely. For an economy of Morocco's size, complete technological autonomy would be extremely expensive and probably unrealistic in the short term.

The more useful question is which dependencies must be controlled, diversified or made reversible.

Legal control over sensitive data, the ability to switch providers, system interoperability, cybersecurity and the availability of domestic expertise may, in certain applications, provide more meaningful guarantees of sovereignty than the nationality of an AI model's developer.

This is precisely where institutions become decisive.

Public administration is the first infrastructure of artificial intelligence

A public administration does not become more efficient simply because it acquires a conversational assistant. It becomes more efficient when it can transform its procedures, organise its information and reduce the time required to deliver services.

The distinction is fundamental.

In many countries, public systems still rely on fragmented databases, inconsistent administrative procedures and applications that communicate poorly with one another. Adding an artificial intelligence layer to these structures may improve individual tasks without resolving their underlying dysfunctions.

A model can analyse a file rapidly, but it cannot guarantee the reliability of information that was never properly recorded. It can recommend a decision, but it cannot replace the rules determining who has the authority to make, review and challenge that decision.

Institutional modernisation therefore becomes a prerequisite for technological modernisation.

In Morocco, plans for a national Data Factory and the sharing of software components across public institutions partly address this problem. Their purpose is to make administrative data more usable, improve its quality and facilitate the reuse of digital tools across government bodies.

Their potential importance extends well beyond artificial intelligence.

Consistent data repositories, common application programming interfaces and regulated information-sharing procedures can reduce administrative costs, accelerate service delivery and limit duplicated investment.

But their effectiveness will depend on implementation.

Public institutions need personnel capable of defining requirements, supervising contractors, evaluating results and detecting errors. They must also protect personal information and preserve citizens' rights.

An AI system used to guide a patient, detect tax fraud or analyse an administrative application does not require the same degree of oversight in every case. The necessary safeguards depend on the potential consequences of errors.

Governance therefore cannot be reduced to declarations of ethical principles. It requires identifiable responsibilities, audit mechanisms, opportunities for appeal and the technical capacity to verify how systems perform.

An administration that purchases an AI tool without knowing how to assess its performance may become dependent on its supplier even when the underlying data remains hosted domestically.

Institutional sovereignty consists precisely in preserving the capacity to evaluate, control and ultimately replace such systems.

Language: an underestimated economic frontier

Morocco has a distinctive characteristic that makes local adaptation particularly important: its linguistic environment.

Modern Standard Arabic, Moroccan Arabic or Darija, Amazigh and French coexist across administrative, economic, educational and everyday settings, although their respective roles vary.

This diversity is a cultural asset, but it also creates technical challenges for systems developed primarily using international datasets with uneven linguistic coverage.

A model that performs well in a dominant language may produce considerably less reliable results when interpreting dialect expressions, local administrative documents or situations involving several languages simultaneously.

In commercial settings, these limitations can reduce a tool's usefulness. In healthcare, law or public administration, they can have more serious consequences.

Developing high-quality linguistic datasets, evaluating models in national languages and building specialised applications are therefore strategic investments.

They do not systematically require training a large general-purpose model from scratch.

Adapting existing models, using document-retrieval systems, developing specialised models or combining several technologies may address specific needs with more limited resources.

The JAZARI institute network, whose development forms part of Morocco's AI strategy, could help structure these efforts by bringing universities, public institutions and businesses closer together.

But the value of such a network should not be measured solely by the number of laboratories established or models announced.

It will depend on its ability to produce reusable resources, reliable datasets, evaluation methods and applications that are actually deployed.

The linguistic challenge reveals a deeper dimension of sovereignty: the ability of a society to use advanced technologies without systematically reshaping its own practices around the cultural and linguistic assumptions of their developers.

Productivity cannot be achieved by decree

One of the risks inherent in national AI strategies is the confusion between technological investment and economic transformation.

Announcing a data centre, training programme or industrial partnership represents a step forward. It does not yet demonstrate an improvement in productivity.

For a company to benefit from artificial intelligence, several conditions must be satisfied.

It must identify tasks that can be improved, possess usable data, integrate tools into its organisation and train its employees. It must also be able to finance the transition and measure the results.

These conditions may appear straightforward. Yet they represent substantial obstacles for small and medium-sized enterprises, which account for an essential part of the productive structure of middle-income economies.

In Morocco, the opportunities are particularly significant in automotive manufacturing, aerospace, agribusiness, logistics, outsourced services and certain agricultural activities.

In manufacturing, artificial intelligence can improve predictive maintenance, quality control, supply-chain planning and energy efficiency.

In logistics, it can support the optimisation of flows and demand forecasting.

In agriculture, forecasting and decision-support tools may contribute to more efficient water use, disease detection and harvest planning, provided reliable local data is available.

In services, document automation and specialised assistants can accelerate certain operations while exposing repetitive functions to more profound restructuring.

But these gains are neither automatic nor evenly distributed.

A large exporting company with structured data and technical expertise will generally be able to integrate these technologies more rapidly than a small business whose management processes remain largely informal.

Without appropriate support policies, artificial intelligence could therefore widen productivity gaps within the Moroccan economy itself.

The challenge is not merely to train highly specialised experts. It is also to spread practical capabilities among technicians, managers, teachers, civil servants and SME executives.

Training 200,000 people may be a useful policy target. But introductory awareness, certification, advanced technical education and the ability to deploy applications in professional environments are very different achievements.

The decisive measure will be less the number of certificates issued than the transformation of productive processes.

Technological dependence does not disappear with a national model

The World Bank report highlights the exceptional concentration of the artificial intelligence value chain.

Advanced processors, computing infrastructure, foundation models and distribution platforms are dominated by a limited number of companies and countries.

This concentration gives suppliers influence extending beyond ordinary commercial relationships.

Prices, access conditions, privacy policies, technical restrictions and export controls can all change the capabilities available to foreign users.

For a country seeking to develop its own applications, the risk is that parts of its digital economy become dependent on infrastructure whose rules may change without its participation.

But responding to this dependence by systematically developing national alternatives carries another danger: dispersing public resources across expensive projects that may quickly become obsolete or remain insufficiently used.

Sovereignty cannot be equated with technological autarky.

A more selective policy would identify genuinely strategic functions: sensitive administrative data, certain critical infrastructures, language technologies, cybersecurity systems and the tools necessary to maintain essential public services.

In these areas, developing domestic expertise, controlled hosting capabilities and alternative solutions may be justified even when their immediate cost exceeds that of a foreign commercial service.

For other applications, adopting international technologies may be economically preferable, provided contracts preserve data portability, interoperability and the possibility of changing suppliers.

This approach also requires distinguishing open software from complete autonomy. A model whose parameters are publicly available may still depend on foreign hardware, specialised software libraries and computing infrastructure that is difficult to replace.

Absolute independence is rarely attainable. Reducing critical vulnerabilities, however, is an operational objective.

An industrial policy that begins with demand

Morocco possesses a potential advantage in this transformation: several sectors of its economy are already integrated into international value chains.

This integration makes it possible to identify specific industrial needs and directly observe the productivity, quality and traceability requirements imposed by major customers.

It could provide the foundation for an artificial intelligence policy driven by demand rather than technological supply alone.

Instead of multiplying programmes without clearly identified applications, the state could concentrate part of its resources on measurable economic problems: reducing administrative delays, improving energy productivity, optimising logistics, strengthening water management or facilitating SME access to specialised tools.

Such an approach does not exclude fundamental research. On the contrary, it can provide that research with industrial applications.

Universities and specialised institutes could develop solutions based on needs identified by companies and public institutions, while public procurement could help domestic suppliers emerge.

But this mechanism requires rigorous evaluation.

Projects should be compared with existing methods, subjected to performance measurements and discontinued when they fail to produce the expected results.

Artificial intelligence must not become an end in itself.

In some cases, a better database, simpler regulation or conventional software application may create more value than a complex generative model.

Establishing this hierarchy of needs is probably one of the most important decisions facing a middle-income economy.

It allows the most expensive investments to be reserved for areas where they respond to demonstrable demand or a clearly established strategic necessity.

Institutions can also become a source of dependence

The framework proposed by the World Bank nevertheless deserves critical examination.

The report recommends, among other measures, drawing initially on certain voluntary private standards to establish AI governance while progressively strengthening public regulatory capabilities.

This approach offers a practical advantage: countries with limited administrative resources can use existing frameworks rather than immediately developing entirely new systems.

But it also raises a difficulty.

When technical standards, evaluation methods and compliance criteria are largely defined by dominant companies, their adoption can reinforce those same companies' influence over emerging markets.

A standard may facilitate interoperability and security. It may also impose compliance costs or technical choices that favour established suppliers.

Building national governance therefore requires more than transposing international frameworks.

It requires the capacity to determine which rules respond to domestic needs, which obligations should apply to suppliers and which rights must be guaranteed to users.

For Morocco, these questions concern personal data protection, digital public procurement, cybersecurity and accountability for automated systems.

Institutional quality cannot be measured by the volume of regulation produced. It depends on the ability to enforce coherent rules, resolve disputes and oversee technologies whose complexity is evolving rapidly.

A country may possess an ambitious legal framework while remaining unable to audit the systems used by its own public institutions.

Conversely, institutions equipped with strong technical expertise can exercise meaningful oversight without multiplying formal obligations.

The challenge is to build public supervisory capacity proportionate to both technological risks and available resources.

The risk of a new divergence between economies

The promise of artificial intelligence rests on a favourable hypothesis: the rapid diffusion of advanced cognitive capabilities could help developing countries overcome certain skills shortages and accelerate economic growth.

But the opposite scenario remains possible.

Wealthier economies generally possess better connectivity, stronger education systems, more digitally mature companies and institutions better equipped to integrate new technologies.

They could therefore capture a disproportionate share of productivity gains even as AI tools become accessible elsewhere.

This would produce a new form of divergence: not simply unequal access to technology, but unequal capacity to exploit it.

Countries that already possess complementary infrastructure would advance more rapidly, while those simultaneously needing to improve electricity networks, public administration, skills and information systems would progress more slowly.

The World Bank identifies precisely these complementarities as an essential condition for economic transformation.

For Morocco, this observation places artificial intelligence within a broader development strategy.

Investment in education, energy, digital networks, research and administrative quality is not peripheral to AI policy. It determines a substantial part of its economic return.

A national strategy may therefore fail despite significant technological investment if companies lack the necessary expertise or institutions remain unable to transform their working methods.

Conversely, relatively unspectacular institutional improvements can generate considerable benefits when they enable the effective diffusion of technologies that are already available.

Competition will not concern computing power alone. It will also concern the speed at which societies can reorganise their activities around new technological possibilities.

Morocco's real strategic trade-off

Morocco's strategy therefore faces a choice that extends beyond the distinction between national models and imported solutions.

The country must simultaneously strengthen its digital infrastructure, develop skills, modernise public administration, support domestic innovation and preserve its negotiating capacity with major international technology providers.

These objectives are complementary, but they do not involve the same costs or investment horizons.

Large infrastructure projects require substantial capital and long-term commitments. Training programmes produce their effects gradually. Administrative and industrial applications can sometimes generate faster returns, but only when organisations are prepared to integrate them.

Public policy must therefore establish a credible sequence.

Initially, adopting proven tools can help identify the most productive applications. Adapting those tools to national languages, datasets and procedures can then support the emergence of specialised domestic companies. Finally, the accumulation of expertise and demand may justify the development of more advanced technological capabilities in selected areas.

This progression should not be interpreted as a prohibition on investing immediately in research or strategic infrastructure. It means that such investments must be connected to identifiable needs and domestic operational capabilities.

Morocco can reasonably seek to become a regional hub for digital services, applied research and industrial artificial intelligence applications.

But the credibility of that ambition will depend on verifiable results: companies that have genuinely improved productivity, public institutions that have reduced processing times, linguistic tools that are actually used, infrastructure operating under economically sustainable conditions, and expertise capable of moving between research and industry.

The MAD 100 billion in additional value added targeted for 2030 is a public policy objective, not an achieved result. Assessing progress will require distinguishing between investment expenditure, economic activity directly created and productivity gains attributable to AI adoption.

This measurement discipline is essential if the strategy is to be evaluated through more than announcements, partnerships and installed equipment.

Sovereignty as the capacity to choose

Artificial intelligence introduces a contradiction into development policy.

On one hand, it makes advanced technological capabilities accessible to economies that were previously excluded from them. On the other, it concentrates a growing share of global technological power among a handful of companies possessing the capital, data and infrastructure required.

For middle-income economies, the response cannot be either passive dependence on these suppliers or systematic imitation of their investment strategies.

The World Bank's 2026 report invites policymakers to treat adaptive capacity as a central driver of development. Morocco's experience will provide a test of that proposition in a country seeking simultaneously to modernise its productive economy, strengthen its infrastructure and build domestic technological expertise.

The decisive question is ultimately not whether Morocco can possess an artificial intelligence model bearing its own name. It may develop, adapt or host several models, with widely differing levels of sophistication and autonomy.

The real question is whether it will possess institutions capable of determining when those models are useful, measuring what they contribute, controlling their risks and replacing them when better solutions become available.

Technological sovereignty is not simply about owning machines or software. It is about retaining control over the economic, administrative and political choices that determine how those technologies are used.

A country can own its AI models without controlling its digital transformation. It can also use technologies developed elsewhere while building genuine autonomy, provided it retains control over its data, skills, rules and decisions.

That distinction may ultimately determine an important part of Morocco's position in the global artificial intelligence economy.

Main Sources

  • World Bank, World Development Report 2026: The Promise of Artificial Intelligence, August 4, 2026. Report, key messages and chapter summaries. https://www.worldbank.org/en/publication/wdr2026
  • World Bank, AI Offers Lifeline to Developing Economies in an Era of Weak Growth, press release, August 4, 2026. https://www.worldbank.org/en/news/press-release/2026/08/04/ai-offers-lifeline-to-developing-economies-in-an-era-of-weak-growth
  • Morocco's Ministry of Digital Transition and Administrative Reform, presentation of progress under Morocco AI 2030, September 8, 2026. https://www.mmsp.gov.ma
  • Official Portal of the Kingdom of Morocco, presentation of the principal developments under Morocco AI 2030, September 2026. https://www.maroc.ma
  • Reuters, Morocco targets $10 billion AI contribution to GDP by 2030, January 12, 2026. https://www.reuters.com/sustainability/climate-energy/morocco-targets-10-billion-ai-contribution-gdp-by-2030-2026-01-12/
  • Médias24, Maroc IA 2030 : JAZARI, data centers, IA souveraine… où en est concrètement la stratégie marocaine ?, September 9, 2026. https://medias24.com/2026/09/09/maroc-ia-2030-jazari-data-centers-ia-souveraine-ou-en-est-concretement-la-strategie-marocaine-1752025/

Methodological note: International estimates of occupational exposure are drawn from the World Bank's published findings and should not be interpreted as forecasts specific to Morocco's labour market. Morocco's 2030 targets and announced data-centre capacities are distinguished from completed investments and operational infrastructure. The analysis draws on publicly available documents and chapter summaries rather than claiming an exhaustive review of the full report.