AI Strains or AI Gains: Powering the Gulf’s Digital Futur...

Saudi Arabia and the UAE want to be more than consumers of AI, they intend to host and export it. That ambition is now colliding with a hard constraint that sits outside most national AI strategies; electricity. Projects now under development mean that AI infrastructure will no longer be a marginal addition to electricity demand. It is instead a new class of energy-intensive industry, with direct implications for generation adequacy, transmission deliverability and summer peak reliability.

Globally, policymakers are also revising their assumptions. The IEA projects that data center electricity consumption could more than double by 2030, reaching roughly 900–1,000TWh under central scenarios, with AI a key driver. What distinguishes the Gulf is not the direction of travel, but the decision to concentrate a meaningful share of that growth inside some of the world’s most heat-stressed power systems. That choice creates strategic opportunity, but it also exposes the power sector to a hard systems test.

NATIONAL GRIDS WILL DETERMINE THE OUTCOME

The electricity consequences of AI in the Gulf will be resolved primarily within national power systems, not through regional balancing mechanisms. Saudi Arabia and the UAE are where the largest AI campuses, hyperscale cloud regions and sovereign AI programs are being developed. These are the systems with the deepest balance sheets and the strongest political incentives to guarantee supply domestically, particularly for infrastructure tied to national security, industrial policy and digital sovereignty.

It is no surprise, therefore, that these are the states already investing the most in strengthening grid stability, even before the arrival of new compute. In its 9M 2025 earnings release, Saudi Electricity Company (SEC) reported capital expenditure of SAR 74bn ($9.7bn), and around 60% of this is going on grid enhancements (MEES, 31 October 2025).

What AI changes is not only the volume of electricity consumed over a year, but the character of the load. AI campuses are concentrated, high-availability facilities, often clustered in specific locations, which can stress transmission corridors, substations and redundancy margins.

For system planners, the binding constraint is often deliverability rather than installed capacity. Transmission lead times, substation reinforcement, right-of-way approvals and N-1 redundancy requirements for critical loads all become decisive variables. These constraints rarely feature in headline AI investment announcements, but they shape how quickly and at what cost infrastructure can be absorbed.

SAUDI ARABIA’S AI BUILDOUT IS MOVING INTO POWER-SECTOR TERRITORY

Humain’s launch and mandate have been reinforced through major partnerships with AMD, Nvidia, AWS and others, and by national planning documents that link compute buildout to domestic industrial strategy and Vision 2030 goals. Nvidia says it would supply “hundreds of thousands” of its Blackwell-class AI chips over time, with an initial tranche of around 18,000 units, and AMD has announced a $10bn partnership with Humain focused on building AI infrastructure with targeted capacity of up to 500MW.

Humain’s first data centers in Riyadh and Dammam are expected to go live in early 2026, each initially supporting around 100MW of load. By 2034 it targets 6GW.

The most explicit grid signal came in December 2025, Saudi Telecom Company (STC) announced that it will form a joint venture with Humain to advance data center development, targeting infrastructure capable of supporting up to 1GW of load, with an initial phase of up to 250MW. At that scale, AI demand becomes a structural feature of an electricity system rather than a niche load.

UAE AND QATAR FACE SIMILAR, THOUGH NOT IDENTICAL, PRESSURES

The UAE is pursuing a parallel strategy, anchored in sovereign AI development and hyperscale cloud infrastructure. In 2024 and 2025, Microsoft and Abu Dhabi-based G42 announced plans to expand data center capacity through Khazna Data Centers, including projects that could together add on the order of 200MW of additional capacity, with operations expected to begin by end-2026.

While the UAE benefits from a more diversified supply mix, including large-scale solar and nuclear baseload, clusters of AI-driven demand still raise questions around network reinforcement, redundancy standards and the value of flexible load. In contrast to this, Qatar’s power system remains gas-dominated and it has little in the way of plans to diversify this. AI could therefore become a material addition to the national emissions profile at a time when scrutiny of the carbon footprint of digital services is intensifying.

FROM STRAIN TO ASSET, THE POLICY CHOICES THAT MATTER

AI does not have to become a reliability or emissions problem. The pivot point is whether Gulf regulators and utilities treat data centres as an inflexible industrial entitlement or as a class of load that must participate in system optimization. Three design choices will determine the outcome.

First, flexibility needs to be a connection condition, not a voluntary add-on. Not every AI workload is time-critical. Training and batch processing can be shifted if operators face the right commercial incentives. Connection agreements can embed demand-response obligations for the flexible share of load, supported by time-varying tariffs and enforceable curtailment protocols during periods of system stress.

Second, compute should be sited where it strengthens the grid, not where it strains it. If AI campuses cluster in already congested urban nodes, they trigger expensive and time-consuming network upgrades. Proactive siting can align new demand with grid headroom, planned transmission corridors and low-carbon supply, reducing both reinforcement costs and reliability risk.

Third, credibility depends on transparent adequacy planning. Saudi Arabia’s AI narrative now carries an implicit promise, reliable power at scale. SEC’s capex trajectory signals that the utility is moving aggressively. The credibility test is whether AI demand is explicitly integrated into public adequacy planning, including network reinforcement and redundancy pathways, rather than assumed to be absorbed by general capacity growth.

WHAT TO WATCH IN 2026

If AI is becoming a new class of power demand in the Gulf, 2026 is the year it will start to appear clearly in operational milestones and grid investment priorities.

Key signals include commissioning timelines for Humain’s first Riyadh and Dammam facilities and what their initial ~100MW scale implies for local network upgrades. Another is how quickly the stc–Humain joint venture converts its ~250MW initial target into contracted projects and grid connection requests, and whether “up to 1GW” evolves into a firm delivery schedule.

Equally important will be whether tariff structures and connection codes evolve to require demand flexibility from large digital users, rather than assuming flat-load service, and whether Saudi adequacy discussions begin to explicitly reference AI load clusters alongside SEC’s ongoing capex ramp.