The Digital Oasis: Data Centers and the Gulf’s AI Ambitions

The Gulf’s next economic transformation is taking shape in data centers. Saudi Arabia and the United Arab Emirates are seeking to turn capital, energy resources, and international partnerships into a stronger position in the global digital economy. But building computing capacity is only part of the challenge. The harder question is whether that capacity can become economically productive, environmentally sustainable, and strategically resilient.

From hydrocarbon revenues to digital infrastructure

For Gulf economies, data centers are more than buildings filled with servers. They are infrastructure for a broader diversification strategy: cloud services, financial technology, public-sector digitalization, and artificial intelligence all depend on reliable computing capacity.

The political economy is familiar, even if the technology is new. Governments can use sovereign capital and long-term planning to coordinate investments that private firms might struggle to finance independently. Power supply, telecommunications, land, and specialist infrastructure must develop together. The potential reward is an economy that captures more value from knowledge and digital services.

Yet investment expenditure is not the same as economic transformation. The lasting gains will depend on whether local firms use this infrastructure, whether specialist skills develop around it, and whether the resulting services are competitive beyond the domestic market. A large facility is an input into diversification, not proof that diversification has succeeded.

Reading the capacity announcements carefully

Headline capacity figures can obscure important distinctions. A national target, a company’s development pipeline, and an operating facility describe different things. They may also overlap. Adding them together can therefore exaggerate the scale of infrastructure actually available.

OpenAI’s May 2025 announcement described Stargate UAE as a planned 1 GW cluster, with an initial 200 MW expected in 2026. Those figures describe the announced project and its expected first phase; they do not independently establish what is operating today. [1] Similarly, the Saudi Press Agency described Hexagon’s 480 MW as the facility’s future total capacity when reporting the laying of its foundation stone. It should not be treated as 480 MW already online at that announcement. [2]

The advantage—and limits—of energy and geography

The original attraction is clear. The region combines substantial investment resources with energy expertise and a location between major markets in Asia, Europe, and Africa. These advantages support the case for developing a larger digital infrastructure industry.

But cheap generation alone does not guarantee cheap computing. A data center needs dependable power delivered around the clock, alongside grid connections, backup systems, cooling equipment, and network access. A low solar-generation price cannot be compared directly with an electricity tariff that includes the cost of delivering reliable power to a customer.

Geography also needs to translate into functioning connections. For applications sensitive to latency, the relevant question is the quality and resilience of the network linking users to the facility. Location creates an opportunity; infrastructure and service reliability determine how much of that opportunity can be realized.

The engineering challenge of desert computing

My paper identifies cooling as a central constraint on the region’s ambitions. High-density computing produces heat, and hot ambient conditions make removing that heat more demanding. The relevant economic cost extends beyond the electricity used by processors to the energy and equipment needed to keep them operating reliably.

Power Usage Effectiveness, or PUE, compares total facility energy with the energy used by IT equipment. A PUE of 1.8 means that every unit of IT energy requires 0.8 additional units for facility overhead. That overhead represents about 44% of total facility energy, and includes more than cooling alone.

This distinction matters when evaluating efficiency claims. A lower PUE indicates less facility overhead, but it does not by itself reveal carbon emissions or water consumption. Cooling designs must be assessed as a complete system, including how heat is ultimately rejected and how much water the system requires.

The sustainability paradox

The region’s energy advantage creates a tension. Electricity can be commercially attractive while remaining carbon intensive. The IEA’s 2025 regional assessment reports that oil and natural gas account for more than 90% of electricity generation across the Middle East and North Africa. This regional average should not be mistaken for the exact electricity mix of every country or facility. [3]

Hassan, ElZeftawy, and Mahmoud examine this issue through an empirical study of language-model inference across the UAE, Iceland, Germany, and the United States. Their experiment uses a specific model and code-generation task, providing evidence about geographical trade-offs rather than a universal emissions estimate for every AI workload. [4]

The implication for my argument is that efficiency and decarbonization must be evaluated separately. An efficient facility can still draw carbon-intensive electricity. Conversely, cleaner electricity does not remove the need to manage equipment efficiency, cooling demands, and resource use.

Building a credible path to cleaner compute

Solar and nuclear energy feature prominently in the paper’s discussion of potential solutions. The broader challenge is to connect low-carbon supply with the continuous demand of data centers. Solar output can help meet daytime demand, but round-the-clock operations still require a reliable combination of generation, storage, and grid support.

A credible strategy therefore needs more than a renewable-energy headline. It needs to explain when clean electricity is available, how the facility handles supply gaps, and how emissions are measured. The development of digital infrastructure and electricity infrastructure must proceed together.

Digital sovereignty still depends on partnerships

Computing capacity is also a strategic asset. Governments seeking greater control over data and AI services must secure access to chips, software, networks, and specialist expertise. Physical ownership of a data center does not remove dependence on these international supply chains.

The Stargate UAE announcement illustrates this interdependence. Its named partners include G42, Oracle, NVIDIA, Cisco, and SoftBank, and the agreement was developed in coordination with the US government. [1] The pursuit of domestic AI capability is therefore closely tied to cross-border commercial and political relationships.

My interpretation is that Gulf states are seeking greater strategic agency within global technology networks. Their ability to finance infrastructure gives them bargaining power, but the durability of that position will depend on trusted partnerships, domestic capabilities, and the resilience of their supply chains.

What will determine success?

The Middle East’s data-center expansion brings together industrial policy, energy economics, and geopolitical strategy. Its significance lies in the possibility of converting existing financial and energy strengths into new productive capabilities.

The strongest measure of progress will not be the largest announced capacity figure. It will be infrastructure that is commissioned, used productively, supported by reliable and increasingly clean power, and connected to a growing base of local skills and services. That is the test of whether the Gulf’s digital oasis can become a durable foundation for economic diversification.