When Gulf officials say that data is the new oil, they usually mean it as a metaphor for economic diversification. There is, however, a more literal reading. Artificial intelligence runs on physical infrastructure – chips, concrete, transformers, cooling towers – and that infrastructure consumes the two resources the Arabian Peninsula manages most anxiously: electricity and water. Over the past eighteen months, the United Arab Emirates and Saudi Arabia have announced AI data-center programs of a scale previously seen only in the United States and China. The deals have been covered exhaustively as geopolitics and as finance. What has received far less attention is the resource ledger: how many gigawatts and how many billion liters the announced ambitions actually imply, and whether they are compatible with the very net-zero and water-security strategies the same governments have championed. This article attempts that accounting.
The Build-Out: Stargate on the Sand
The centerpiece of the Emirati program is Stargate UAE, a 1-gigawatt AI computing cluster being built by Khazna Data Centers – a G42 company – together with OpenAI, Oracle, Nvidia, Cisco and SoftBank, announced in May 2025 as the anchor of a planned 5 GW UAE–US AI Campus in Abu Dhabi. The scale is difficult to overstate: the campus site spans 19.2 square kilometers – roughly ten square miles – and is billed as the largest AI infrastructure complex outside the United States, with OpenAI noting it could serve users within a 2,000-mile radius covering nearly half the world’s population. Construction is moving at Gulf speed. By late 2025, more than 5,000 workers had poured over 100,000 cubic meters of concrete, with steelworks weighing one and a half times the Eiffel Tower; Mubadala chief executive Khaldoon Al Mubarak confirmed the first 200-megawatt phase – equipped with roughly 100,000 of Nvidia’s Grace Blackwell GB300 systems – for delivery in the third quarter of 2026. The project was unlocked by a Washington policy shift: in late 2025 the US Commerce Department authorized the export of the equivalent of up to 35,000 GB300 chips to G42, ending years of restrictions tied to concerns over the company’s former China links.
Saudi Arabia’s answer is Humain, an AI company created under the Public Investment Fund in May 2025 – launched, pointedly, one day before President Trump’s visit to Riyadh – with a mandate spanning data centers, cloud platforms and Arabic-language model. Its infrastructure roadmap is even larger than Abu Dhabi’s: first facilities in Riyadh and Dammam of up to 100 MW each launching in 2026, a target of 1.9 GW of capacity by 2030 and 6.6 GW by 2034, in a program analysts value at some $77 billion. The company has secured 211 land plots with access to 14 GW of power, plans to deploy up to 600,000 Nvidia GPUs over three years, has signed a $5 billion “AI Zone” partnership with AWS and a $10 billion joint venture with AMD and Cisco, and in 2026 tendered infrastructure works for a single 6 GW campus across 24 square kilometers of eastern Riyadh – a data-center district the size of a small city. Microsoft, Google, Oracle and Alibaba are meanwhile expanding cloud regions across the Gulf from Qatar to Kuwait.
The Power Bill
Now for the ledger. Data centers are unlike almost any other industrial load: they run at high utilization around the clock, and AI-optimised racks consumemany times the power of conventional server hardware. Globally, the International Energy Agencyestimates that data centers consumed about 415 TWh of electricity in 2024 and projects this to more than double to roughly 945 TWh by 2030 – slightly more than Japan’s entire current consumption – with AI the dominant driver and consumption from AI-focused facilities tripling over the period.
Applied to the Gulf, the arithmetic becomes striking. A simple illustrative calculation: a 5 GW campus running continuously at full load would draw around 44 TWh per year. The UAE’s entire power generation in 2024 – a record year – was 164 TWh, of which the four-reactor Barakah nuclear plant, the largest single decarbonization asset in the Arab world, contributed 40 TWh. In other words, the Abu Dhabi AI campus at full announced build-out would eventually demand more electricity than the country’s entire nuclear program produces, and over a quarter of current national generation – in a system where demand was already growing around 3% a year before AI entered the picture. Even the 1 GW Stargate cluster alone implies close to 9 TWh annually, comparable to adding a new emirate to the grid. These are upper-bound figures – real facilities ramp gradually and rarely run at nameplate – but they explain why Wood Mackenzie projects UAE data-center power demand to more than double by 2030, and why the IEA has suggested Gulf states may need to double their generation capacity by 2030 to support AI infrastructure and the desalination needed to cool it.
The governments are not blind to this. The UAE has formed a National Team for Reviewing the Impact of Data Centers on the Energy Sector, tasked with assessing grid impact and revising regulation, while the updated Energy Strategy 2050 mobilizes up to AED 200 billion by 2030 to triple renewable output. Yet friction points are real: analysts note that current UAE regulations prevent data-center operators from signing direct corporate power-purchase agreements for renewables, complicating the clean-power procurement on which hyperscalers’ own net-zero pledges depend. Saudi Arabia’s pitch, meanwhile, leans openly on subsidized electricity as a competitive advantage – economically rational for attracting compute, but in tension with the Kingdom’s parallel goal of freeing hydrocarbons for export rather than burning them for domestic megaprojects.
The Water Bill
Electricity is only half the ledger. Cooling server halls in a region where summer temperatures routinely exceed 45°C demands water – and the Gulf has almost none of its own. Industry estimates suggest a single megawatt of water-cooled data-center capacity can consume about 25 million liters per year. FTI Consulting calculates that data centers in Saudi Arabia alone consumed some 15 billion liters in 2024, with preliminary research suggesting the figure could reach roughly 87 billion liters – about 4% of the Kingdom’s current water consumption – as the build-out proceeds. In the Gulf, marginal water is desalinated water: Saudi Arabia operates two of the world’s largest desalination plants, and desalination is itself energy-hungry, creating a circular problem in which AI requires water, water requires energy, and energy generation strains the very climate commitments both states have made.
The response is emerging but uneven. Khazna is piloting treated sewage effluent for cooling at some facilities – a genuinely important innovation if scaled – while regional researchers advocate mandatory efficiency standards, non-potable cooling water and summer-neutral load design before inefficient facilities lock in decades of demand. The creation of a GCC water security task force in March 2025 signals that governments now frame water, like energy, as a shared strategic issue. Dubai’s plan to host the world’s most energy-efficient, fully solar-powered desalination plant points to the preferred escape route: coupling new water supply to clean power rather than expanding business-as-usual desalination.
Why Do It at All? The Strategic Logic – and the Risks
Given the resource costs, why is the Gulf racing to host compute? There are three reasons. First, comparative advantage: capital, cheap energy, land and speed of execution are precisely the inputs AI infrastructure needs, and the Gulf has them in greater abundance than almost anywhere else. Second, geopolitics: hosting American-supplied compute embeds the UAE and Saudi Arabia in the US technology bloc at a moment of intensifying US–China rivalry – the chip exportauthorizations of 2025 were as much diplomatic instruments as commercial licenses, creating what one analysis calls mutual dependencies extending beyond the commercial realm. Third, the diversification imperative: analysts place total announced AI data-center investment across the GCC at more than $30 billion through 2030, with projections that regional data-center markets will roughly triple in value – a genuine new non-oil sector, not a vanity project.
The risks deserve equal billing.
Demand risk: the world may not need as much Gulf-hosted inference as the announcements assume, and capacity targets – as the IEA dryly notes of the industry globally – face bottlenecks in chips, transformers and grid connections that make aggressive timelines unlikely.
Dependence risk: the entire stack, from GB300 chips to cloud software, is American, meaning Washington retains a kill-switch over the Gulf’s flagship diversification bet.
And resource risk: ESG-focused investors are reportedly already demanding detailed water-usage disclosures before committing to Gulf tech projects – an early sign that the power-and-water ledger will eventually be priced by markets even if governments defer it.
Conclusion: The Grid Is the Strategy
The Gulf’s AI build-out is real, funded and – unusually for megaprojects in the region – ahead of schedule: concrete is being poured in Abu Dhabi, and the first Saudi campuses come online in 2026. But the decisive constraint on the Gulf’s AI ambitions will not be chips, capital or talent; it will be the grid and the water system behind it. The same governments that built Barakah and the world’s largest solar parks now face a second, harder infrastructure race: expanding clean generation and reinventing cooling fast enough that the AI economy reinforces, rather than cannibalizes, their energy transitions. If they succeed, the Gulf will have done something remarkable – turned hydrocarbon wealth into computational power via a bridge of nuclear reactors and solar panels. If they fail, the region’s data centers will simply be burning gas to answer the world’s chatbot queries, and the “new oil” will look uncomfortably like the old one. Either way, the megawatts, not the models, will tell the story.
