The Middle East data center market is predicted to be $19.89 billion by 2030, growing at a 14.92% CAGR, according to research firm Arizton. IDC puts AI spending across the Middle East, Türkiye and Africa at $4.5 billion in 2024, rising to $14.6 billion by 2028. And in the final quarter of 2025, IDC reported that Saudi Arabia and the UAE recorded the strongest AI infrastructure growth anywhere in the world, driven by sovereign AI programs.

Those announcements are now turning into buildings. HUMAIN’s first sites were scheduled to come online in the second quarter of 2026, CNBC reported, and Stargate UAE's opening 200MW is targeted for the third quarter of 2026, according to The National. The region's AI infrastructure story is moving from press release to commissioning schedule, and that changes the questions worth asking.

The gap between announced and usable capacity

Computer Weekly reported in April 2026 that the region’s operational third-party data center capacity stands at roughly 500MW, forecast to triple to around 1.5GW by 2030.

Both numbers are accurate. They describe different things. The Abu Dhabi campus hosting Stargate UAE is planned at 5GW, which OpenAI describes as the largest AI campus announced outside the United States. HUMAIN, Saudi Arabia’s Public Investment Fund-backed AI company, targets around 6GW by 2034, per CNBC. Usable capacity is what has been built, energized, cooled, certified and connected, and it is still counted in hundreds of megawatts.

Part of the difference is category. The 1.5GW forecast counts third-party colocation supply, while the headline campuses are largely sovereign self-builds whose capacity may never reach the open market. A CIO cannot lease a slice of a national AI program. The capacity available to enterprises is the smaller number, which is exactly why it is worth understanding what limits it.

For a CIO in Riyadh, Dubai or Doha, the strategic question is not whether the Gulf’s build-out is real. It is how much of it you can contract, in which jurisdiction, and in which quarter. Four constraints decide the answer.

CONSTRAINT 01
Power decides the time

An AI training rack draws more than 100kW against 5-10kW for a traditional enterprise rack, an order-of-magnitude jump that explains why, as Computer Weekly put it in 2026, "AI doesn't simply sit on top of digital infrastructure, it requires a redesign of the entire datacentre stack."

The redesign starts at the substation, and it is the reason even flagship projects sequence delivery in disciplined phases. Stargate UAE's first 200MW tranche is targeted for the third quarter of 2026, with structural work advanced and the first critical mechanical systems already on site, according to G42's construction updates. HUMAIN's first Riyadh and Dammam facilities are due online in the second quarter of 2026 at roughly 100MW each, CNBC reported in August 2025.

Read those schedules the way an operator does. New capacity arrives in 100-200MW increments, gated by grid interconnection and generation rather than by land or capital. The practical consequence: contract against energization dates, not campus masterplans, and treat the delivery quarter as the only number that matters.

CONSTRAINT 02
Chips set the compute economics

Power builds the shell. Accelerators make it AI capacity. Stargate UAE will run on NVIDIA's GB300-class systems, according to the project's partners, and HUMAIN's partnership with NVIDIA covers AI factories of up to 500MW, announced in May 2025. Those allocations, not floor space, are the scarce unit in Gulf compute economics.

Allocation is also policy-shaped. Advanced accelerators reach the region under export approvals set in Washington, which makes chip supply a contractual risk to underwrite rather than a line item to assume. The questions now asked in colocation and sovereign cloud negotiations: who holds the accelerator allocation, and what substitution rights exist if a shipment slips?

The economics improve once facilities run. AI applied to data center operations can cut operating costs by up to 30%, chiefly through energy conservation and predictive maintenance, according to equipment manufacturer Maysteel. Density inflates the capital bill; intelligent operations claw part of it back.

CONSTRAINT 03
Data residency decides where capacity counts

A gigawatt in the wrong jurisdiction is zero usable gigawatts for a regulated workload. Saudi Arabia's Personal Data Protection Law has been fully enforceable since its grace period ended in September 2024, and its Article 29 treats cross-border remote access as a data transfer in its own right, according to law firm DLA Piper's data protection guidance. The UAE's Federal Decree-Law 45 of 2021 sets the federal data protection baseline. Microsoft's Azure region for Saudi Arabia, under construction and expected in 2026, shows how seriously hyperscalers now take in-country processing.

The shift runs deeper than compliance paperwork. "Digital sovereignty is ultimately about operational continuity," Nischal Kapoor, chief revenue officer at e& enterprise, told Computer Weekly in April 2026. "If critical functions like security, identity and incident response are dependent on external jurisdictions, resilience is compromised." Sovereignty, as the same Computer Weekly report frames it, now spans four layers: data, infrastructure, operations and AI.

The cost calculation has flipped with it. For infrastructure strategy, that means classifying workloads by where they may legally and safely run, then buying capacity against that map rather than against price per kilowatt alone.

CONSTRAINT 04
Cooling in a water-stressed region

The Gulf is building some of the world's most thermally demanding compute in one of its driest places. The wider region contains 16 of the world's 20 most water-stressed countries, Computer Weekly noted in March 2026, and one forecast cited in the same analysis puts GCC data center water demand at 426 billion liters a year by 2030.

Engineering answers exist. Direct liquid cooling cuts a facility's direct water use by 90-98% compared with conventional evaporative systems, per the same Computer Weekly analysis, and closed-loop designs recycle the same water across years of operation. The subtler point from that reporting: roughly 60% of a data center's water footprint sits in electricity generation rather than in the facility itself, so cooling choices and power sourcing are one decision, not two.

Operators are turning the constraint into differentiation. Khazna Data Centers' DXB8 facility in Dubai became the first data center globally to be certified zero waste by SCS Global Services, diverting 99.55% of operational waste, Computer Weekly reported in April 2026; the company plans more than 1GW of hyperscale capacity by 2030. Expect water usage effectiveness and diversion rates to appear in procurement scorecards alongside power usage effectiveness (PUE), because ESG reporting teams now read colocation contracts too.

Five questions for CIOs to ask before signing

Sources

• Arizton, Middle East Data Center Market: Industry Outlook & Growth Forecast · arizton.com
• Computer Weekly: capacity feature (Apr 2026) · sovereign-first (21 Apr 2026) · geopolitics (18 May 2026) · zero waste (28 Apr 2026) · water-smart (25 Mar 2026) · AI redesign (2026) · computerweekly.com
• The National, Stargate UAE first phase (5 Dec 2025) · thenationalnews.com
• G42 construction updates (2025) · g42.ai | OpenAI, Introducing Stargate UAE (May 2025) · openai.com
• CNBC, HUMAIN (27 Aug 2025) · cnbc.com | NVIDIA × HUMAIN partnership (May 2025) · nvidianews.nvidia.com
• IDC, AI spending in META (2025) · idc.com | DLA Piper, Saudi Arabia data protection (accessed Jun 2026) · dlapiperdataprotection.com
• Microsoft, Azure Saudi Arabia region (expected 2026) · microsoft.com | Maysteel, AI and data center operating costs · maysteel.com