A person using artificial intelligence today encounters something remarkably simple. They open an application, type a question and receive an answer. They might ask for a document to be prepared, an image to be created or a difficult subject to be explained. Within moments, the computer produces something that might previously have required considerable time and effort.

What the user rarely sees is the enormous industrial system working behind that apparently effortless exchange.

Somewhere, powerful processors are performing calculations inside data centres that consume electricity on the scale of substantial industrial facilities. Those processors were designed by specialist companies, manufactured through some of the most sophisticated production processes ever developed and assembled into computing systems costing billions of dollars.

The artificial intelligence model itself may have required months of research, extensive collections of data and large teams of engineers before it became available through an ordinary screen.

The contrast is striking. One of the most accessible technologies of the modern age depends upon infrastructure that relatively few organisations can afford to develop independently.

This brings us to an important question in the history of industrial revolutions.

Who controls the machinery of a new industrial age?

In the previous issue, we examined how artificial intelligence is changing human work, particularly in occupations involving language, analysis and professional judgement. We saw that the technology can improve productivity, although its consequences for employment will depend upon how businesses and workers adapt.

But the question of who benefits from greater productivity is not settled by the technology alone.

The companies developing AI models, the businesses operating computing infrastructure and the people building services with those models occupy different positions in the emerging economy. Some provide the machinery, others supply the intelligence, while many more are beginning to use it in their ordinary work.

Their interests overlap, but they are not identical.

Understanding those relationships is essential to understanding the Fourth Industrial Revolution, because the history of industrial progress has always been shaped as much by access to productive resources as by the inventions themselves.

Every Industrial Revolution Created New Centres of Economic Power:

When steam engines began transforming industry in the eighteenth century, owning machinery became an increasingly important source of economic advantage.

A manufacturer who could invest in mechanised equipment could produce goods at a scale that was difficult for individual artisans to match. Factories brought machinery, workers and capital together, changing how production was organised and who could compete effectively.

The steam engine did not make every factory owner wealthy, nor did mechanisation eliminate small producers overnight. Nevertheless, the ability to finance machinery and organise production became increasingly important in determining which businesses could expand.

The Second Industrial Revolution carried that process further.

Steelmaking, petroleum refining, electrical infrastructure and mass manufacturing demanded substantial investment. Building a steelworks or an oil refinery was beyond the financial capacity of most individuals, while establishing a large automobile factory required machinery, premises, materials, workers and access to reliable transportation.

Industries became increasingly associated with large enterprises capable of coordinating enormous productive systems.

Yet the economic importance of these technologies extended beyond the companies that owned the factories.

A business did not need to construct a power station to benefit from electricity. Once electrical networks developed, thousands of smaller enterprises could purchase power and use it to operate machinery, illuminate premises or provide new services.

This distinction between owning infrastructure and gaining access to its productive capacity became an important feature of industrial development.

The Third Industrial Revolution introduced another variation.

Computers and software made it possible to develop products that could be distributed without manufacturing a new physical copy for every customer. A small team could create a computer program and make it available to users far beyond the company’s immediate geographical market.

The internet expanded those possibilities considerably.

A small business could establish a website, reach international customers and use services supplied by companies whose infrastructure it did not own.

At the same time, the digital economy produced its own concentrations of power.

Search engines, social networks, online marketplaces and cloud computing providers became important intermediaries in how information and commercial activity moved across the internet.

Their advantages came partly from technology, but also from scale, investment and the relationships between users.

A service that attracts large numbers of participants can become more useful to others precisely because those participants are already there. Businesses operating such networks may therefore enjoy advantages that are difficult for newcomers to reproduce.

Artificial intelligence combines features of these earlier revolutions.

Like the great industrial enterprises of the nineteenth century, the most advanced AI developers depend upon expensive machinery and substantial capital. Yet, like the software companies of the digital age, they can distribute the resulting capabilities to enormous numbers of users through networks.

This creates a new economic arrangement in which sophisticated technology can become widely available even though the resources required to develop it remain concentrated.

That arrangement offers both opportunities and potential difficulties.

It allows small businesses to use capabilities they could never afford to build independently, but it can also make them dependent upon companies controlling the underlying infrastructure.

The Machinery Behind Artificial Intelligence:

The physical foundations of modern AI are more complicated than the applications through which most people encounter it.

At the centre are specialised computer processors capable of performing enormous numbers of calculations. Training a large AI model involves repeatedly processing data and adjusting the model’s internal parameters, a task that requires considerable computing power.

Companies such as Nvidia design processors widely used for this work. But designing an advanced processor and manufacturing it are different activities.

The production of leading chips depends heavily upon highly specialised semiconductor manufacturers, particularly Taiwan Semiconductor Manufacturing Company, better known as TSMC.

Manufacturing these chips requires facilities equipped with extraordinarily precise machinery, as well as considerable technical expertise and investment. Building a competitive production operation is therefore not something that a new company can accomplish simply by hiring talented programmers.

The concentration is considerable.

Stanford University’s 2026 AI Index reported that TSMC manufactured almost all leading AI chips, illustrating how much of the industry’s advanced hardware supply depended upon one company.

This concentration does not mean TSMC owns the AI systems that use those processors. A manufacturer, a chip designer and an AI developer can be separate businesses with different customers and commercial interests.

It does, however, show why the supply of advanced computing hardware has become strategically important.

Once manufactured, the processors must be installed in data centres equipped to operate them.

These facilities require electricity, cooling systems, networking equipment, maintenance and physical security. Their operators must also manage the substantial costs of purchasing equipment that can become outdated as semiconductor technology advances.

Some of the world’s largest technology companies already possess extensive cloud computing infrastructure, developed long before the recent expansion of generative AI.

Amazon, Microsoft and Google operate services through which businesses can rent computing capacity rather than establish their own facilities.

This model has been enormously important to the digital economy. A company developing an application can rent servers, storage and other resources according to its needs, avoiding some of the cost and complexity of owning the equipment.

Artificial intelligence is extending that arrangement, although the largest AI projects can place exceptional demands upon computing infrastructure.

Electricity has become an especially important consideration.

The International Energy Agency’s Key Questions on Energy and AI, published in April 2026, estimated that data centres worldwide consumed approximately 485 terawatt-hours of electricity in 2025. Its central projection suggests that demand could rise to about 950 terawatt-hours by 2030.

Those figures include conventional data centre activities as well as AI, and the projection remains sensitive to investment, efficiency improvements and the capacity of electricity networks.

Nevertheless, they reveal how closely the expansion of artificial intelligence is connected to the physical economy.

A data centre cannot operate merely because its software is sophisticated. It must have reliable electricity, suitable equipment and the infrastructure required to support that equipment.

In this respect, the Fourth Industrial Revolution has not escaped the industrial foundations established by the earlier revolutions.

It continues to depend upon manufacturing, energy and transportation, even while appearing to operate entirely through digital services.

The considerable investment required to provide these resources helps explain why relatively few organisations can compete at the largest scale.

Yet the companies that control the infrastructure do not necessarily determine every application that will be developed upon it.

That distinction is becoming increasingly important.

Large Companies Hold Advantages but Competition Remains:

The enormous cost of AI infrastructure has encouraged substantial investment and close relationships among technology companies.

Businesses developing advanced models frequently require access to computing systems supplied by other firms. Cloud providers, in turn, have an interest in attracting model developers whose services can increase demand for their infrastructure.

These relationships can benefit both parties.

An AI developer gains access to expensive computing resources without independently constructing every facility it requires. A cloud provider secures customers for its services and may gain access to products that complement its existing business.

But the arrangements can also influence competition.

In January 2025, the United States Federal Trade Commission published a report examining major partnerships involving Microsoft and OpenAI, Amazon and Anthropic, and Google and Anthropic.

The report considered how investment agreements, access to computing capacity and commercial commitments might affect competition. Among the concerns identified were the possibility of higher costs for developers seeking to change providers and difficulties for smaller competitors attempting to obtain essential resources.

These were potential competition concerns, rather than findings that every partnership was unlawful.

The distinction matters because cooperation between large companies is not inherently harmful. Developing expensive infrastructure often requires investment, and partnerships can bring new technologies to market more quickly.

The question is whether such arrangements leave sufficient room for competitors, customers and new businesses to make meaningful choices.

The evidence suggests that competition remains active, particularly in the development of AI models.

In July 2026, the Organisation for Economic Co-operation and Development published Artificial Intelligence Markets: Recent Developments and Competition Issues.

Its assessment found a mixed picture.

Certain parts of the industry, especially advanced hardware and cloud infrastructure, remain highly concentrated and difficult for new entrants to challenge. Yet the market for AI models has been more competitive, with new developers achieving impressive results and prices falling substantially.

This is an important qualification to the idea that a few technology companies have already secured permanent control of artificial intelligence.

Developing a capable model remains challenging, but competition has not been confined to the largest American technology companies.

Chinese developers and other research organisations have demonstrated that alternative approaches can produce highly capable systems. Models made available under open licences have also expanded the choices available to developers and businesses.

The term open requires some care.

A model may be available for download and modification without every detail of its original training process being disclosed. Different licences may also place different conditions upon commercial use.

Nevertheless, accessible models can reduce the dependence of businesses upon a single commercial provider.

A company developing specialised software may choose to obtain AI services through a cloud provider, adapt an available model or operate a smaller model on its own equipment.

Each arrangement involves different costs and technical requirements.

There is also a difference between paying less for a unit of AI processing and spending less overall. As businesses find more uses for AI, they may make substantially more requests, increasing their total expenditure even while individual operations become cheaper.

The future market is therefore unlikely to be determined by a single factor.

Access to hardware matters, but so do model quality, reliability, cost, technical expertise and the ability to build products that solve practical problems.

The companies controlling infrastructure possess important advantages. Whether those advantages become lasting economic dominance will depend partly upon competition and the alternatives available to customers.

Economic Value Extends Beyond Those Who Own the Models:

The discussion of AI ownership often concentrates on the companies developing the largest models. They attract considerable public attention because their products are highly visible and their investments can be extraordinary.

But economic value does not arise only where a technology is invented.

Consider electricity.

Power companies generate and distribute electrical energy, yet much of the economic activity made possible by electricity occurs in businesses that simply purchase and use it.

A bakery benefits from electric ovens without owning a power station. A manufacturer operates machinery using electricity supplied through a network, while a hospital depends upon reliable power to provide services.

Artificial intelligence may develop a comparable economic relationship.

A company supplying accounting software can incorporate an existing AI model into its product. It may then use that technology to help customers organise records, explain financial information or prepare documents.

The accounting software company does not need to own the original AI model, but it may still create considerable value by understanding accounting, designing a useful application and serving customers effectively.

A small agricultural technology company could use available AI tools to make technical information more accessible to farmers. A legal technology business might develop software that helps practitioners organise documents and conduct research.

The underlying models could be supplied by larger technology companies, while the smaller businesses contribute specialised knowledge and practical applications.

This distinction is especially important because many commercial problems require more than powerful general-purpose technology.

A system designed for a particular industry must understand the information, working conditions and expectations of its users. It may need access to specialised databases, established procedures or locally relevant terminology.

The quality of the application can therefore depend as much on its design and the knowledge of its developers as on the capabilities of the underlying model.

Data introduces another dimension.

AI models learn from information, and businesses with access to large quantities of useful data may enjoy important advantages. But the value of data depends upon its quality, relevance and the lawful conditions under which it can be collected and used.

A hospital’s records, for example, may contain information useful for developing better systems, but that does not mean the information can be freely transferred or exploited without regard to patient privacy and applicable law.

Similarly, a business may possess years of customer information without automatically gaining the expertise necessary to turn it into a useful AI application.

Owning data, controlling access to it and possessing the right to use it for a particular purpose are different matters.

These distinctions complicate the economics of artificial intelligence, but they also explain why the emerging industry offers opportunities beyond its largest participants.

A business may succeed because it understands customers better, maintains reliable information or develops a service that works effectively under local conditions.

There are reasons to believe these opportunities will matter.

The OECD’s 2026 research on competition in the age of AI identified potential advantages for smaller firms using generative AI, alongside continuing benefits for companies with stronger existing resources.

Neither outcome is guaranteed.

Small businesses may become more competitive through affordable technology, but they may also face rising dependence upon providers whose pricing or access conditions they cannot influence.

Their commercial position will depend partly upon the availability of alternatives and their capacity to adapt.

That makes access to AI an economic issue in its own right.

The important question is not simply which organisation possesses the most advanced model. It is whether businesses and individuals can obtain useful technology on terms that allow them to build, compete and retain a reasonable share of the value they create.

Africa Can Participate Without Owning Every Part of the Industry:

For countries such as Uganda, the economics of artificial intelligence presents a different challenge from that facing the largest technology companies.

Constructing advanced semiconductor factories or operating the enormous computing facilities used to train leading AI models would require levels of investment and technical capacity beyond the immediate reach of most developing economies.

It would make little economic sense to suggest that every country must establish such facilities before it can benefit from artificial intelligence.

The World Bank’s World Development Report 2026, published in August, makes a related argument.

Developing economies can benefit from AI by adopting existing tools, adapting them to local circumstances and gradually building the technical capacity needed to develop more advanced systems where appropriate.

The greatest immediate opportunities may therefore lie not in reproducing the infrastructure of the world’s largest AI companies, but in making available technology useful to local businesses, public institutions and communities.

That requires more than purchasing software.

A system developed primarily around information from wealthy economies may not respond adequately to questions involving local languages, agricultural conditions, public services or cultural circumstances.

The usefulness of AI depends partly upon whether it can work with information relevant to the people it is intended to serve.

Uganda offers an example of what local adaptation can look like.

In October 2025, Sunbird AI introduced Sunflower, a language model developed with a focus on Ugandan languages. In September 2026, the organisation announced Sunflower v2, expanding its suite of speech and language models to cover 67 African languages.

The project builds upon existing AI technology while contributing locally relevant data, language resources and further training.

According to Sunbird AI’s September 2026 release, the newer systems include models designed for translation and speech recognition, alongside smaller models intended for less powerful devices.

The organisation has published performance evaluations, although its reported results should be distinguished from independent verification across every language and practical setting.

The significance of the initiative is not that Uganda has suddenly acquired computing infrastructure comparable to that of the world’s largest technology companies.

It is that useful innovation can occur through adapting existing models to solve problems that larger developers may not have prioritised.

That principle extends beyond language.

An agricultural advisory service could combine an existing AI model with reliable information about local crops, weather conditions and farming practices. A public-service application could make government information easier to understand in widely spoken local languages.

Educational tools could be designed to accommodate schools with limited connectivity or students who access information primarily through mobile phones.

Such applications would still require careful testing, accurate information and suitable professional oversight. AI cannot independently remedy shortages of teachers, health workers or agricultural specialists, and poor infrastructure can undermine even well-designed software.

Nevertheless, these are plausible areas in which locally relevant technology could improve existing services.

Uganda’s wider digital transformation efforts provide part of the necessary foundation. In June 2026, the National Information Technology Authority–Uganda launched a five-year strategic plan aimed at expanding access to electronic government services, strengthening infrastructure and improving the delivery of public services through technology.

Those objectives concern digital development more broadly, but reliable connectivity, accessible public information and institutional capacity are also important conditions for effective AI adoption.

The country’s challenge will be to move beyond treating artificial intelligence as a collection of imported applications and develop the expertise needed to adapt technology to its own economic circumstances.

That includes training software developers, supporting relevant research, improving digital infrastructure and creating opportunities for businesses capable of building specialised products.

It also involves making sensible choices about investment.

Not every worthwhile AI application requires expensive computing hardware. Some services can operate through rented cloud infrastructure, while smaller models may be suitable for particular tasks or devices.

For developing economies with competing priorities, the ability to select technologies appropriate to local needs may be more important than attempting to match the spending of wealthy technology companies.

There is also a broader question of economic participation.

If African businesses increasingly use AI systems developed elsewhere, a portion of the value generated may flow to the companies supplying those systems. That is not necessarily undesirable; international trade has always allowed businesses to benefit from technologies they do not manufacture themselves.

The concern arises where excessive dependence limits choice, makes services unaffordable or prevents local enterprises from developing their own competitive capabilities.

Building skills, maintaining access to alternative technologies and supporting local innovation can reduce some of those risks.

The future of AI in Africa will therefore depend not only upon the technology available to the continent, but upon the businesses and institutions capable of applying it productively.

This brings us back to the larger history of the industrial revolutions.

The first transformed production through machinery and factories. The second expanded industrial scale through electricity and mass manufacturing. The third created an economy increasingly dependent upon software, information and global networks.

The fourth draws upon all of those foundations while introducing a technology capable of assisting with intellectual work.

Like earlier revolutions, it creates valuable new capabilities while changing the economic importance of particular resources.

Advanced processors, computing infrastructure, data and specialised expertise have become important sources of commercial advantage. The organisations possessing them may occupy influential positions in the developing AI economy.

But history also shows that the value of an industrial technology is not confined to those who manufacture its machinery.

Electricity transformed businesses that never owned power stations. Computers created opportunities for companies that did not manufacture semiconductors, while the internet allowed enterprises of every size to reach customers far beyond their immediate markets.

Artificial intelligence may follow a similar course.

The companies developing its most advanced infrastructure will remain important, but much of its eventual economic contribution may come from the businesses, institutions and individuals who discover how to use it effectively.

The question is whether those opportunities will remain sufficiently accessible for new participants to enter the market and build useful services of their own.

That outcome is not predetermined.

It will depend upon competition, investment, public policy, technical expertise and the choices made by organisations adopting the technology.

The Fourth Industrial Revolution may be distinguished by machines capable of producing increasingly sophisticated intellectual work, but the economic history surrounding those machines remains familiar.

Ownership matters because it influences who can invest, compete and determine the conditions under which technology is supplied.

Access matters because it determines who can put that technology to work.

And the greatest measure of the revolution may ultimately be found not in the companies that possess its most powerful machines, but in how widely their capabilities improve the lives and work of the people who use them.

Next in The Trend Report:

The Geography of Artificial Intelligence Is Reshaping Global Competition

Computing Infrastructure and Technological Capability Are Becoming Matters of National Strategy