Microsoft and Abu Dhabi-based G42 have announced a substantial expansion of the United Arab Emirates’ data centre footprint: a 200-megawatt increase in capacity to be delivered through G42’s Khazna Data Centers subsidiary, scheduled to begin ramping up operations before the end of 2026. This move is packaged within Microsoft’s broader Gulf strategy — an investment commitment that totals roughly
$15.2 billion for the UAE over the coming years — and is paired with U.S. licences allowing Microsoft to export advanced Nvidia AI accelerators to the country, a regulatory breakthrough framed by officials as a tightly controlled precedent in technology diplomacy.
Background
Microsoft-G42 strategic ties: a brief history
The Microsoft–G42 relationship has evolved quickly from commercial collaboration to strategic co-investment. Microsoft took a minority equity stake in G42 in 2024 and secured a board seat as part of a roughly
$1.5 billion deal; since then the partnership has broadened into a multi‑billion-dollar infrastructure and AI collaboration centered on regional cloud services, AI research, and digital transformation initiatives across the UAE. The most recent announcement formalises a capital‑intensive phase focused on hyperscale compute capacity and sovereign cloud services.
Khazna Data Centers: the local delivery vehicle
Khazna, G42’s dedicated data centre platform, has been the operational anchor for this build‑out. Khazna’s published roadmap already includes AI‑optimised facilities — including a marquee 100MW AI data centre in Ajman and further Abu Dhabi sites described as AUH4 and AUH8 — with modular designs, advanced cooling technologies and LEED Gold environmental targets explicitly stated. The new 200MW commitment from Microsoft and G42 will be integrated into Khazna’s network, accelerating timelines and capacity for both cloud and AI workloads across the UAE.
What was announced: the technical and commercial headline items
- 200 megawatts of additional data centre capacity to be built and operated through Khazna Data Centers, coming online before the end of 2026.
- Microsoft’s UAE investment envelope tied to the project rises to roughly $15.2 billion, composed of prior and planned capital and operating spend across 2023–2029.
- The U.S. Commerce Department has approved export licences allowing Microsoft to ship advanced Nvidia AI GPUs (reported to be equivalent to tens of thousands of A100-class chips, implemented via Nvidia’s newer GB300/Grace‑based hardware) to UAE facilities — a regulatory precedent emphasised by stakeholders.
- The build will emphasise AI‑optimised infrastructure: high‑density compute halls, liquid cooling and energy efficiency measures, aligned to Khazna’s ongoing projects and standards (LEED Gold, modular architectures).
These core facts form the foundation of what will be one of the largest regional data infrastructure investments in recent years, and they materially change the capacity and capability available in the Gulf for cloud providers, AI developers and sovereign workloads.
Why this matters: strategic implications for cloud, AI and sovereignty
Strengthening Azure in the Gulf
The added capacity injects direct, near‑term compute and storage resources into the UAE’s local Azure footprint, improving latency, resilience and the ability to host sensitive workloads domestically. For customers in finance, oil & gas, government and healthcare who demand
data residency and regulatory compliance, expanded sovereign-capable Azure infrastructure materially improves the value proposition of using Microsoft cloud services inside the UAE.
A test case in tech diplomacy and export controls
The U.S. approval to export advanced Nvidia accelerators — reported to enable Microsoft to move the equivalent of more than 60,000 A100 chips’ worth of compute — is not purely a commercial detail. It represents a diplomatic and national‑security judgment: the U.S. weighed risks of high‑end AI hardware leaving U.S. jurisdiction against the strategic benefits of anchoring AI development to a partner nation that pledges reciprocal investment and safeguards. The license has been described as a precedent that could shape future export decisions, balancing industrial policy, alliance politics and non‑proliferation concerns.
Local industrial strategy and talent development
Beyond raw compute, the announcement includes commitments to workforce development and local engineering capabilities. Microsoft’s planned centers, partnerships and operating projects — coupled with Khazna’s ongoing facilities and financing initiatives — create an ecosystem for AI labs, startups and applied research to access hyperscale resources without moving data offshore, an important lever for any national AI strategy.
Technical anatomy: how modern AI data centres differ (and what Khazna will likely deploy)
High‑density compute and cooling
AI infrastructure is dominated by accelerator‑heavy racks that generate concentrated heat and draw significant power. Modern AI facilities therefore emphasise:
- Liquid cooling (direct or immersion) to efficiently remove heat from GPU clusters and reduce PUE (Power Usage Effectiveness).
- Modular hall designs allowing repeatable, rapid expansion while minimising construction waste and deployment time.
- Energy management software, often AI‑driven, to dynamically schedule workloads and reduce peak grid draw.
Khazna has publicly described both AI‑optimised data halls and liquid‑cooling integration on its Ajman project, signalling that the new capacity will align with the best practices now standard for large‑scale AI compute.
Power and grid considerations
A 200MW expansion is not a modest load: data centres of this scale require secure, often dedicated power feeds, transmission upgrades and, increasingly, long‑term energy contracts. Documents from Khazna and regional partners note plans to work with local utilities to secure supply and pursue efficiency improvements; observers will watch for:
- The mix of grid vs. on‑site generation and the role of renewables.
- Arrangements to stabilise supply during peak demand periods to avoid grid stress.
Geopolitical and regulatory risk assessment
Export controls, audits and operational constraints
Export licences for high‑end accelerators tend to come with significant compliance conditions: restrictions on who may access sensitive systems, ongoing audits, personnel vetting, and limits on sharing models or datasets with prohibited parties. The Microsoft licence reported in recent coverage was described as requiring strict safeguards and monitoring, which will create a complex compliance matrix for Microsoft, G42 and any third‑party customers hosted on the new infrastructure. These obligations elevate operational cost and governance burdens.
Reputation and congressional or multilateral scrutiny
The partnership has already drawn attention from benches of lawmakers and policy analysts concerned about the potential for technology flows to end‑use parties that could indirectly benefit geopolitical competitors. Even with strict controls, the mere scale and visibility of the project makes it a target for oversight. Companies and public institutions procuring services from the expanded infrastructure will likely face increased due diligence requests from boards, regulators and insurers.
Supply dependency on GPU manufacturers
The programme’s performance will be tightly coupled to Nvidia’s supply roadmap and global chip availability. Any geopolitical frictions that affect chip exports, or supply‑chain constraints that raise prices or delivery lead times, will delay utilisation of the new capacity and raise effective cost per compute hour. The licence itself addresses export permission, but not downstream supply risks that are external to Microsoft and G42.
Commercial and market implications
For enterprise customers
- Faster cloud performance for AI workloads in the region due to reduced latency and localised GPUs.
- New cloud sovereignty choices: organisations that must keep data inside the UAE now have a larger domestic Azure/Khazna option with high‑performance AI capability.
- Higher procurement complexity: customers must weigh compliance obligations and potential access restrictions that may come with certain hardware or managed services.
For cloud competitors
AWS, Google Cloud, Oracle and regional providers will face direct pressure to match both capacity and the sovereign‑cloud guarantees advertised by Microsoft and G42. Expect competitive responses: accelerated regional builds, sovereign cloud products, and tighter regional partnerships with local operators. This can lower price and increase choice — but it also raises the infrastructure bill for all hyperscalers competing at sovereign‑grade levels.
For G42 and Khazna
The Microsoft commitment de‑risks Khazna’s expansion and improves access to capital (Khazna itself disclosed significant financing plans and facilities in recent briefings). The partnership also embeds international cloud operations expertise into Khazna’s management stack, increasing the company’s profile as a regional hyperscaler. However, deeper foreign partnerships also amplify scrutiny and necessitate stronger corporate governance and compliance programs.
Environmental and sustainability considerations
Energy intensity and decarbonisation
AI compute is power‑hungry, and large GPU farms can materially raise a country’s electricity demand profile. While Khazna projects emphasise LEED Gold goals, adiabatic cooling techniques and energy efficiency, the net effect on national emissions depends on the power mix used to supply the facilities. Real sustainability gains rely on:
- Power sourcing contracts tied to renewable generation.
- Waste‑heat recovery or cogeneration integrations where feasible.
- Transparent PUE metrics and independent verification.
Any claims of “green AI” require verifiable data on the grid mix and energy procurement strategies. At present, public statements promise efficiency and design standards, but implementation and reporting will determine real environmental impact.
Water, cooling and local resources
Large data centres can also put pressure on water resources in arid regions if evaporative or closed‑loop cooling uses significant make‑up water. Khazna’s references to adiabatic and liquid cooling aim to reduce overall energy and water use compared with legacy approaches, but regional water management and reuse strategies will matter for long‑term sustainability.
What’s next: timelines, unknowns and watchpoints
Near‑term timeline
Microsoft and G42 expect the new capacity to begin operations before the end of 2026. Khazna’s earlier projects show that modular builds and aggressive timelines are central to their approach, but actual commissioning of high‑performance GPU clusters will hinge on chip deliveries and final regulatory clearances tied to export licenses.
Key unknowns to monitor
- Exact scope of licence conditions imposed by the U.S. Commerce Department and how they will shape who can access hosted systems.
- The precise mix of GPU models and quantities that will be deployed at each Khazna facility and the delivery cadence from Nvidia.
- Power contracts and renewable procurement that will determine long‑term sustainability and operating costs.
- Third‑party customer access policies for sensitive workloads, especially where export controls or national security concerns might require segmentation or additional vetting.
Practical guidance for IT leaders and procurement teams
- Audit data‑location requirements: map current and planned workloads to evaluate whether sovereign cloud placement in the UAE materially resolves compliance or latency needs.
- Request full compliance documentation: demand written details on access controls, audit windows, and personnel vetting tied to any export‑controlled hardware.
- Validate energy and sustainability claims: require PUE targets, renewable energy purchase agreements, and water‑use metrics before signing long‑term contracts.
- Model supplier risk: include GPU supply and export‑licence contingencies in service‑level and capacity planning.
- Examine contractual exit and portability terms: ensure that data and workloads can be migrated if geopolitical or regulatory conditions materially change.
These steps mitigate the operational and governance risks introduced by the convergence of hyperscale AI capacity, cross‑border licences and sovereign requirements.
Strengths and potential pitfalls: a balanced evaluation
Notable strengths
- Rapid capacity scale‑up dramatically increases available AI compute in the Gulf and creates new local options for enterprise AI hosting.
- Regulatory breakthrough in export licences opens the door for more advanced deployments under controlled frameworks, smoothing a path for others if safeguards are proven effective.
- Integrated local operator (Khazna) with announced financing and modular project experience reduces execution risk and supports regional industrial strategy.
Potential pitfalls
- Export‑control complexity may restrict third‑party access, complicate tenancy models, and increase compliance costs.
- Supply and pricing pressure for GPUs remains a systemic risk that could hamper initial utilisation and raise compute costs.
- Reputational and oversight exposure: high visibility projects tied to geopolitics will invite scrutiny from legislators, civil society and customers concerned about governance, privacy and misuse.
Conclusion
The Microsoft–G42–Khazna announcement marks a turning point for large‑scale AI infrastructure in the Middle East: a significant infusion of compute capacity, multinational capital, and a regulatory precedent for the export of advanced AI accelerators under strict safeguards. For enterprises and governments the outcome is pragmatic: more local options and greater performance for AI workloads, combined with new layers of compliance and geopolitical risk.
Delivering the promise will require disciplined execution across procurement, energy contracts, compliance programmes and transparency reporting. Where those elements align, the UAE could become a major regional hub for sovereign AI services; where they do not, the project will be a high‑profile test of how commercial ambition, national strategy and export‑control regimes can be reconciled in the age of accelerated AI demand.
Source: Mobile World Live
Microsoft teams with G42 on UAE data centre capacity expansion