The Gulf Data Center Race: Saudi Arabia, the UAE and Qatar

Gulf data center investment has become one of the most ambitious parts of the global AI build-out. Saudi Arabia, the United Arab Emirates and Qatar are pairing state capital and low-cost power with US technology partners to turn oil-era wealth into AI compute. This guide explains who is building what, why the region thinks it can compete, and the new risks after the 2026 strikes. It is part of the Data Center Economy hub; for comparison, see US data center investment and India’s data center build-out.

This article covers investment and infrastructure, not security policy. Figures are based on company announcements and reporting as of late October 2026, and Gulf project timelines in particular are prone to revision.

Why are Gulf states investing so heavily in data centers?

Gulf governments see AI compute as a way to diversify away from hydrocarbons and capture high-value digital industries. They can deploy sovereign capital quickly and offer cheap, abundant energy and large sites. Their aim is to become the third pole of AI infrastructure after the US and China.

The region’s pitch rests on three advantages. The first is capital. Sovereign funds such as Saudi Arabia’s Public Investment Fund (PIF), Abu Dhabi’s Mubadala and MGX, and the Qatar Investment Authority can fund multi-billion-dollar projects without waiting for third-party debt. The second is power: gas-rich grids, rapidly falling solar costs and state utilities able to dedicate generation to large campuses. Our guide to the power bottleneck explains why that matters when US and European grid queues stretch for years. The third is land. The UAE-US AI Campus alone covers 19.2 square kilometres, according to The National.

There is also a geopolitical logic. The US July 2025 AI Action Plan created an American AI Exports programme to sell full-stack US AI technology to allies. Gulf governments have paired these technology deals with pledges to invest in the US, and in return gain access to the advanced chips that make their data centers useful for frontier AI. The region’s main weaknesses are heat, which raises cooling costs; water scarcity; a thin local talent pool; distance from major user populations; and, as 2026 showed, physical security.

What is Stargate UAE and how big is the UAE-US AI Campus?

Stargate UAE is a 1 GW AI cluster developed by G42 with OpenAI, Oracle, Nvidia, Cisco and SoftBank. It sits inside the planned 5 GW UAE-US AI Campus in Abu Dhabi. Its first 200 MW phase was scheduled to come online in 2026, making it the first Stargate site outside the US.

G42’s data center arm, Khazna, is building the facilities. In December 2025, Mubadala’s Khaldoon Al Mubarak said the first phase would be ready in the third quarter of 2026, and G42 said the full 1 GW would take about three years. Epoch AI’s satellite-based tracker estimated that the first two buildings would reach 200 MW by around December 2026, with 1 GW by early 2028. Its imagery dates from November 2025, so later progress is not reflected. Reports have valued the first gigawatt at about $30 billion.

The project has been affected by the regional conflict. Diario Bitcoin, citing a Reuters report based on anonymous sources, said in September 2026 that the UAE was considering splitting Stargate UAE into several smaller sites across the country. Hardening measures reportedly under study include underground halls, blast-resistant structures and duplicated power and cooling. No revised master plan has been published, and the first phase was reported to be continuing. Spreading the cluster out would raise costs and could reduce training performance, because frontier model training depends on tightly interconnected chips.

Microsoft and G42

Microsoft is the UAE’s largest Western technology investor. In November 2025, it said it would spend $15.2 billion in the country from 2023 through 2029. That includes a $1.5 billion equity stake in G42, more than $4.6 billion of data center capex to the end of 2025, and more than $5.5 billion of further AI and cloud capex from 2026 to 2029. Microsoft also announced a 200 MW expansion of its UAE capacity with G42, built by Khazna and due to start coming online before the end of 2026. US export licences approved in September 2025 allow Microsoft to ship the equivalent of about 60,400 additional Nvidia A100 chips to the UAE, in the form of GB300 GPUs.

What is HUMAIN and what is Saudi Arabia building?

HUMAIN is Saudi Arabia’s national AI company, created by the Public Investment Fund in May 2025. It is building data centers, buying chips and partnering with US firms. Its CEO, Tareq Amin, has set targets of 250 MW online by the end of 2026, 1.7 GW in 2027, about 3 GW by 2030 and 6 GW by 2034.

HUMAIN’s first two facilities, in Riyadh and Dammam in the Eastern Province, were designed with 100 MW of initial capacity each and were expected to go live in Q2 2026, according to Data Center Dynamics. The company has secured 211 land plots across the kingdom for future campuses. Its partner list is long: a $5 billion AI Zone with AWS, a 500 MW data center for Elon Musk’s xAI, an AMD deal for 500 MW of compute and a 1 GW joint venture with Saudi Telecom Company. It also has Qualcomm AI100 inference racks, Groq inference chips, and $1.2 billion of financing from the National Infrastructure Fund for 250 MW. It had bought 18,000 Nvidia GB300 chips, with far larger volumes reportedly planned subject to US approval.

Funding is becoming more structured. In a September 2026 interview with Enterprise AM, Amin said HUMAIN was preparing a domestic data center fund reportedly seeking an initial $2.5 billion, alongside a GPU and IT equipment fund and a previously planned $10 billion global AI venture fund. Enterprise estimated the 2027 build at about $14.5 billion, assuming $10–12 million per MW; that estimate is the publication’s, not the company’s. Our guide to data center construction cost per MW explains where such per-MW figures come from.

DataVolt and Oxagon

The other big Saudi project is DataVolt’s AI campus at Oxagon, NEOM’s Red Sea industrial zone, where HUMAIN is a partner. Construction began in 2026 on a 360 MW first phase, of which 100 MW is moving first. Phase one is due to be operational in 2028, and the full campus targets 1.5 GW, powered by NEOM’s wind and solar assets. Smaller local players are also raising money. Axelerated Solutions is seeking about $179 million through a sharia-compliant fund to build two AI-ready facilities of 10–20 MW each.

How is Qatar positioning itself in the AI data center race?

Qatar is taking a more measured, finance-led approach. In December 2025, Qai, the AI subsidiary of the Qatar Investment Authority, formed a $20 billion joint venture with Brookfield to fund integrated AI compute facilities in Qatar and potentially abroad.

The Brookfield–Qai venture includes an integrated compute centre intended to broaden access to high-performance computing in Qatar. The partners also plan to explore co-developing AI infrastructure internationally. The venture is linked to Brookfield’s AI Infrastructure Fund, part of a wider $100 billion global AI infrastructure programme. No capacity figure has been disclosed.

On the operating side, Ooredoo’s digital infrastructure arm, Syntys, bought Q Data’s two Tier III facilities in the Qatar Free Zones in January 2026, taking its live Qatar capacity to about 26 MW. It targets more than 120 MW across the Middle East and North Africa by 2030. Ooredoo has launched Nvidia-powered sovereign AI cloud services and an Oracle sovereign cloud partnership. Microsoft operates a cloud region in Qatar and, with Meeza, plans a sovereign private cloud. Qatar’s market is far smaller than Saudi Arabia’s or the UAE’s in megawatts, but QIA’s capital gives it reach well beyond its borders.

Source: The National; Microsoft (November 2025); Enterprise AM (September 2026); Data Center Dynamics; w.media (Brookfield–Qai, December 2025; Syntys, January 2026). Targets are company plans, not operational capacity.

How do US chip export rules shape Gulf data centers?

Gulf data centers are only as useful as the chips inside them, and the most advanced Nvidia GPUs need US export licences. Washington has approved large but conditional shipments, tying chip access to security controls, reporting and reciprocal Gulf investment in the US.

In November 2025, during the Saudi crown prince’s visit to Washington, the Commerce Department authorised exports of the equivalent of up to 35,000 Nvidia Blackwell GB300 chips each to G42 and HUMAIN, up to 70,000 in total. Both companies must meet strict security and reporting requirements monitored by the Bureau of Industry and Security. G42’s deployments fall under a Regulated Technology Environment framework approved under Commerce guidelines. Separately, Microsoft’s licences cover the equivalent of more than 80,000 A100-class chips in the UAE across two approvals.

The conditions reflect US concern that advanced chips could be diverted to China, which is why licences carry anti-diversion and reporting obligations. For investors, licence approvals are the gating factor in project timing: a data center hall without GPUs earns nothing. The chip trade is covered in more detail in our guide to the data center supply chain.

How did the 2026 strikes on Gulf data centers change the investment case?

On 1 March 2026, Iranian drone strikes hit three AWS data centers, two in the UAE and one in Bahrain. It was the first confirmed military attack on a hyperscale cloud provider. AWS’s UAE region lost two of its three availability zones, disrupting banks, payments and enterprise software.

Rest of World reported that AWS advised customers to move workloads out of the affected region while it rebuilt. The damage was serious but recoverable, and multi-zone redundancy had not been designed for simultaneous physical attacks. In April 2026, Iran’s Revolutionary Guard publicly named Stargate UAE as a potential target, according to The Next Web and other outlets. Companies planning Gulf projects found that standard commercial insurance excludes acts of war, so war-risk cover has become a separate cost line.

Analysts quoted by Rest of World expect a cooling of Gulf AI investment rather than a reversal. Facilities already built will not be abandoned, but unbuilt projects face more scrutiny. Saudi Arabia’s PIF has reportedly signalled a 15% cut in capital spending as it reassesses priorities. On the other side, Gulf governments are expected to reassure US partners and hold them to their commitments, and sovereign funds such as ADIA and Mubadala are built to withstand volatility.

Will the Gulf become a global AI compute hub?

The Gulf has the capital, power and political backing to become a major AI compute exporter by the end of the decade. Whether it does depends on continued US chip approvals, delivering announced campuses on schedule, and convincing tenants that the region is physically and politically safe.

The bull case is the scale of committed partnerships. OpenAI, Oracle, Microsoft, AWS, xAI and AMD have all tied capacity plans to the region, and sovereign funds are investing in global data center platforms. Abu Dhabi’s MGX, for example, has invested alongside Western infrastructure funds in large data center platform acquisitions, as described in our article on the biggest data center M&A deals. Cheap power also matters more as AI clusters grow: the Gulf can offer gigawatt connections on timelines that are hard to match in Virginia or Frankfurt.

The bear case is that most Gulf capacity is still on paper. In late 2026, operational AI capacity was in the hundreds of megawatts, against multi-gigawatt targets. Chip export policy could tighten again, regional conflict has already damaged facilities, and demand depends heavily on a few US AI labs whose own finances are scrutinised in our AI bubble debate explainer. Heat and water constraints also push up operating costs compared with Nordic sites.

Frequently Asked Questions