When Innovation Arrives Before Capacity

Science, technology and innovation are often treated as engines of development, but tools do not create development simply by arriving. A cloud platform becomes transformative only when the community receiving it has the practical infrastructure to absorb it. In the cloud era, development depends less on whether technology reaches a place than on whether local institutions can turn rented compute into usable capability.

Infrastructure-as-a-service was built on a powerful economic idea. Institutions should not need to own the full technical backbone of the internet before they can use it. Computing power could be rented as demand grew. Software could run through networked systems instead of remaining trapped inside local hardware. Digital production could begin before a local institution had the capital to build its own computing estate.

That promise remains real. Cloud computing can narrow the distance between a local problem and a sophisticated technical response. It can give smaller institutions access to systems once reserved for much larger organizations. The same model also exposes a harder divide. The old digital divide asked who could connect to the internet. The new divide asks who can compute with it.

Connectivity remains deeply unequal. In 2024, 2.6 billion people were still offline. Internet use reached 93 percent in high-income countries and only 27 percent in low-income countries. The first divide has not disappeared. The cloud economy has placed a second divide above it, where connection does not automatically become economic agency.

The problem appears most clearly when cloud deployment succeeds technically but fails institutionally. A local government may move benefit administration into a cloud environment and deliver a faster public interface almost immediately. Residents see cleaner forms and quicker transactions. Behind that improvement, the agency may still lack the internal skill to evaluate system security or control the movement of public data. It may also discover that leaving the system is harder than joining it. The service has modernized, but the institution has not gained the authority to govern the infrastructure beneath it.

The compute divide begins there: a gap between access to cloud systems and the practical power to make computation useful.

The Cloud as a Capability Test

Infrastructure-as-a-service changes the economics of adoption because it separates use from ownership. Institutions can enter advanced digital systems without owning the full stack beneath them. In theory, latecomers gain a faster path into the internet economy. In practice, the most important layer of the system may remain outside local control.

Cloud infrastructure services reached $119.1 billion in the fourth quarter of 2025 and $419 billion for the full year. Amazon, Microsoft and Google accounted for roughly two-thirds of enterprise cloud infrastructure spending in that quarter. Compute has moved from back-office technology into the operating layer of the internet economy.

Artificial intelligence is intensifying that shift. Worldwide AI spending is forecast to reach $2.59 trillion in 2026. Infrastructure is becoming the largest segment of that market because AI systems require specialized IaaS capacity and the hardware fabric behind it. Cloud capacity is now a condition for participation in the next production system rather than a narrow IT choice.

For institutions with technical depth, cloud infrastructure becomes leverage. It allows faster experimentation and more sophisticated service delivery without full ownership of the physical estate. For institutions without that depth, the same cloud can become dependency by contract. The interface becomes modern while authority over the system remains remote.

Advanced computing is now part of the market architecture of the internet rather than a narrow technical layer. Cloud computing therefore becomes more than an industry category. It becomes a test of institutional readiness. A community that can govern cloud systems gains room to build. A community that cannot govern them may still modernize, but on terms written elsewhere.

Technology Transfer After the Cloud

Once cloud infrastructure becomes the operating layer of development, technology transfer can no longer mean the arrival of a tool. A platform may be deployed and a receiving institution may be counted as modernized. The surface of adoption still says little about whether the institution understands the system it now depends on.

Real technology transfer leaves behind operational judgment. It gives institutions the ability to keep services running, protect public interest and change direction when a system no longer fits local needs. In a cloud environment, transfer must include authority over infrastructure that is used through outside providers rather than owned outright.

A cloud migration that leaves a public agency unable to change providers is not full modernization. An AI deployment that cannot be audited locally is not full capability. A data system that collects local information but returns little local knowledge has not closed the divide. It may have made the divide harder to see.

Inclusive AI sits inside a wider development problem. Economies need preparation before new systems can generate broad productivity gains. The same logic applies to cloud computing. Rented compute becomes developmental only when the receiving institution has enough practical control to use it with judgment rather than dependence.

Cloud systems make advanced capability easier to reach, but absorption still has to be built inside the institutions that use them. Dependency grows when a service works only because an outside provider retains the knowledge behind it.

Sovereignty and the New Geography of Dependence

The governance problem becomes sharper once cloud systems carry public services and economic activity. Cloud does not float above geography. It operates through legal jurisdictions and commercial contracts. Data sovereignty has become central because information is increasingly governed by the terms of the systems that store and process it.

A nation can import cloud capacity and still lose authority over its digital future if it lacks bargaining power over where data resides and how systems can be examined. A public agency can digitize services while binding itself to infrastructure it cannot inspect or replace. A local firm can gain reach while becoming dependent on standards it had no role in shaping.

Compute power has become geopolitical for the same reason. The ability to run large-scale models and process strategic data is unevenly distributed. The issue is no longer only where data is stored. It is where computation happens, whose rules govern it and whether the user has credible alternatives when access or pricing changes.

Cloud dependence is now a market-structure problem. The UK cloud services market was estimated at £9 billion, with competition not working as well as it could. Weak cloud competition can raise costs, reduce choice and limit innovation for the organizations that increasingly depend on cloud infrastructure.

European data policy is moving in the same direction. The EU Data Act began applying in September 2025 and gives cloud users stronger rights to switch providers or use several providers in parallel. Cloud sovereignty is therefore no longer only a question of where data sits. It is also a question of whether institutions can move and bargain around the infrastructure they use.

Falling behind in the compute economy no longer means being outside the internet. It can mean being inside the internet economy on weaker terms.

Data Centers Bring the Cloud Back to Earth

The sovereignty question leads directly to the physical one. If cloud capacity determines who can govern digital systems, data centers determine where that capacity can exist. The language of the cloud suggests abstraction, but infrastructure-as-a-service becomes real through the power systems and local approvals that make computation possible.

Global data-center electricity consumption is projected to more than double to around 945 terawatt-hours by 2030. From 2024 to 2030, data-center electricity use grows by around 15 percent per year. That is more than four times faster than electricity demand from the rest of the economy. In the United States, data centers are expected to account for nearly half of national electricity demand growth by the end of the decade.

Those pressures change the meaning of local development. A region cannot become compute-rich through cloud adoption alone. It needs a public bargain over who benefits when data centers arrive and who absorbs the cost of hosting them. The same cloud region that strengthens national compute capacity can place heavy demands on the local systems beneath it.

The timing gap makes the problem more acute. A data center can become operational in a few years. The energy system needed to support it often takes longer to plan and finance. Cloud expansion therefore becomes a governance problem for the communities and utilities that must carry the load, not only for the technology companies that build the demand.

Data-center expansion now sits inside national policy debates over electricity, jurisdiction and public value. The compute divide is therefore not only about who can afford cloud services. It is also about who can host the infrastructure behind them without letting the higher-value gains travel elsewhere.

Data centers can bring investment, but investment alone is not development. The cloud becomes developmental only when physical hosting connects to institutional strength.

Leaping Forward Without Becoming Dependent

The compute divide is not a case against cloud computing. It is a case for treating cloud access as the beginning of development rather than the end of it.

A public agency that rents infrastructure for faster service delivery can also build internal competence over procurement and system oversight. The same cloud environment that makes modernization easier can become a training ground for local judgment if the institution treats deployment as a capacity-building process. The point is not to own every layer of infrastructure. It is to avoid becoming dependent on systems that local institutions cannot understand or govern.

AI diffusion shows both the opportunity and the risk. Artificial intelligence reached 1.2 billion users in only three years. Nearly 70 percent of those users are in developing countries. In some high-income economies, two in three people already use AI tools. In many low-income countries, usage remains close to 5 percent. Adoption is spreading quickly, but the benefits are positioned to concentrate where connectivity, compute and regulation are already strong.

Those numbers show the danger of mistaking diffusion for development. A tool can spread quickly while capability remains scarce. Cloud and AI can compress the path to advanced services, but they cannot replace the institutional work required to govern those services.

Leapfrogging does not mean skipping capacity. It means building capacity faster because shared infrastructure lowers the cost of entry. The risk is a false leap, where a community appears modern because it has adopted the newest platform while the knowledge remains elsewhere. The interface is local. The power is remote. The result is dependence made efficient.

Regulation becomes decisive here. Poorly designed rules can raise costs without building capability. Stronger governance can turn cloud adoption into a foundation for trust and local bargaining power. The internet is now economic infrastructure whose design conditions shape productivity and risk. Cloud development has to be judged through that same frame.

Science, technology and innovation will not serve development simply because they move faster. They will serve development when communities can absorb them and build with them. The old digital divide taught that access alone does not create equality. The cloud era adds a sharper lesson: technology transfer without capacity can reproduce dependency at higher speed.

In the internet economy, the difference between falling behind and leaping forward is increasingly the ability to compute.

TL;DR Summary

• Cloud access does not automatically create digital capability.
• The old digital divide focused on connection, while the new divide focuses on computation.
• Infrastructure-as-a-service lowers entry costs but can create dependency by contract.
• Rented compute becomes developmental only when institutions can govern it.
• Cloud concentration makes bargaining power part of digital development.
• Technology transfer must leave behind operational judgment, not only tools.
• Data sovereignty now depends on portability and credible exit.
• Compute power has become a geopolitical and regulatory asset.
• Data centers make the cloud a local power and planning issue.
• AI adoption can spread faster than institutional readiness.
• Leapfrogging requires capacity building, not only faster technology deployment.
• The ability to compute is becoming a core condition of economic agency.

Sources

• ITU; Facts and Figures 2024; – Link

• Synergy Research Group; GenAI Helps Drive Quarterly Cloud Revenues to 119 Billion; – Link

• Gartner; Gartner Forecasts Worldwide AI Spending to Grow 47 Percent in 2026; – Link

• UNCTAD; Technology and Innovation Report 2025; – Link

• Cohen and Levinthal; Absorptive Capacity A New Perspective on Learning and Innovation; – Link

• UK Competition and Markets Authority; Cloud Services Market Investigation; – Link

• European Commission; Data Act; – Link

• International Energy Agency; Energy Demand from AI; – Link

• UNDP; The Next Great Divergence; – Link

• Epoch AI; Computing Power and the Governance of Artificial Intelligence; – Link

• Institute of Internet Economics; Regulation and the Economic Restructuring of the Internet; – Link

• Institute of Internet Economics; The Political Economy of AI Data Centers and Regional Digital Capacity; – Link