Microsoft G42 200 MW UAE Data Center Expansion via Khazna

Microsoft and Abu Dhabi’s G42 have committed to a 200-megawatt expansion of data‑center capacity in the United Arab Emirates — to be built and operated through G42’s Khazna Data Centers subsidiary — a move the partners say will come online in phases beginning before the end of 2026 and that sits inside Microsoft’s broader, $15.2 billion UAE investment program through 2029.

Background​

Over the past two years the UAE has pursued an explicit strategy to become a regional and global hub for artificial intelligence and cloud services. Major public- and private-sector investments have focused on building local compute capacity, attracting hyperscale cloud partners, and creating governance and skills programs designed to enable sovereign hosting for regulated industries.
Microsoft’s announcement is the latest, and most concrete, expression of that strategy: the company has publicly framed a multi‑year program of capital and operating spending in the UAE totaling $15.2 billion across 2023–2029, with roughly $7.3 billion already invested by the end of 2025 and $7.9 billion earmarked for 2026–2029. Those numbers were confirmed directly by Microsoft’s corporate statement and replicated by independent reporting. At the same time, U.S. export licensing actions have enabled Microsoft to stage very large quantities of high‑end Nvidia accelerator capacity inside UAE facilities — an essential material underpinning for an “AI‑first” data‑center program. Public reporting and Microsoft’s account indicate prior authorizations equated to the compute capacity of roughly 21,500 A100‑class GPUs, and more recent approvals cover additional GB/Blackwell‑class systems that Microsoft says amount to the equivalent of tens of thousands more A100s. These GPU licensing details underpin how the planned 200 MW of power will translate into usable AI compute.

---

What the announcement actually says​

The headline facts​

• Capacity: 200 megawatts (MW) of additional IT‑load capacity in the UAE, to be delivered through Khazna Data Centers, a subsidiary of G42.
• Timeline: Initial capacity is expected to begin coming online before the end of 2026.
• Investment context: The capacity expansion is part of Microsoft’s wider $15.2 billion commitment to the UAE through 2029, with $7.3 billion spent by year‑end 2025 and $7.9 billion planned for 2026–2029.
• Hardware/export controls: U.S. Commerce Department export licenses have allowed Microsoft to deploy advanced Nvidia accelerator families (GB/Blackwell and earlier A100/H100 equivalents) in the UAE; Microsoft publicly reported deploying the equivalent of ~21,500 A100‑class units so far and cited further authorizations that expand that footprint. Independent outlets report approvals covering shipments that Microsoft and analysts characterize as equivalent to tens of thousands more A100‑class GPUs.

What was not disclosed (and what remains open)​

• Exact site locations for the new capacity were not revealed.
• Capital expenditure (capex) breakdown for the 200 MW build — total project spend, capex per MW, and financing structure — were not provided.
• Detailed energy sourcing and firming contracts (e.g., firmed renewables, PPAs, onsite generation) were not published alongside the announcement.

Those omissions matter: 200 MW of IT load is a major grid commitment that requires substation work, multi‑year power purchase agreements, and precise site engineering to realize operationally.

---

Technical anatomy: what 200 MW means in practice​

Translating megawatts to usable AI compute​

A data‑center announcement quoting 200 MW generally refers to IT load — the electrical capacity available to run servers, accelerators, storage and networking inside the data halls. This is different from a site’s total electrical footprint (which also includes facility losses, cooling, and redundancy). Practically speaking:

• Modern AI‑dense racks often draw 10 kW–40 kW+ per rack, depending on whether they house multi‑GPU nodes, liquid cooling, or specialized systems.
• At a conservative 10 kW per rack assumption, 200 MW of IT load could theoretically support ~20,000 racks; at higher per‑rack densities used for Blackwell/GB‑class accelerators, the number of racks is lower but each rack delivers far greater inferencing and training capacity.

This math is order‑of‑magnitude and sensitive to architectural choices: rack density, cooling approach (air vs. direct liquid vs. immersion), redundancy (N+1 vs. 2N), and the specific GPU family used all change the final compute available.

The GPU layer: Blackwell / GB‑class vs A100 / H100​

The recent export approvals and Microsoft’s own disclosures indicate deployment of newer GB/Blackwell family systems in addition to earlier A100/H100‑class gear. Key differences:

• Power & density: GB‑class racks (Blackwell Ultra/GB300 variants) are typically much more power‑dense than A100 generations. That increases the facility’s per‑rack cooling and distribution demands but yields greater per‑rack performance.
• Performance implications: One GB300 rack can substitute for many A100 nodes in training throughput or inference throughput depending on topology; therefore, GPU counts presented as “A100 equivalents” should be treated as a performance proxy rather than a literal inventory. This distinction matters for cost, energy, and thermal engineering.

Cooling, PUE and energy implications​

High‑density AI racks drive facilities toward liquid cooling (direct-to-chip or immersion) to achieve competitive Power Usage Effectiveness (PUE) and to keep operational costs reasonable. A 200 MW IT allocation implies:

• significant substation upgrades and redundancy feeds,
• likely adoption of liquid cooling at scale,
• multi‑year energy contracts (PPAs, renewables plus firming or storage) to meet reliability and sustainability commitments.

---

Khazna, G42 and Microsoft: roles and relationships​

Khazna Data Centers is G42’s hyperscale wholesale data‑center arm and will be the delivery vehicle for this expansion. The announcement formalizes a deeper operating relationship between Microsoft and G42 that builds on Microsoft’s prior $1.5 billion equity investment in G42 and accompanying collaboration frameworks. That local partnership model serves multiple purposes:

• Provides a domestic operator and contracting counterparty for local governments and regulated customers.
• Enables Microsoft to deliver Azure‑grade services while meeting sovereign hosting and compliance requirements.
• Helps reconcile hyperscaler platform control with national concerns about data residency and governance.

---

Strategic implications​

For Microsoft​

• Market positioning: The expansion strengthens Azure’s proposition in the Gulf for customers who require local residency, audited governance, and low latency for AI services like Microsoft 365 Copilot. Microsoft has explicitly tied part of the UAE commitment to product‑level residency assurances that enable qualified customers to keep Copilot processing in‑country.
• Supply chain and policy win: Securing U.S. export licenses for frontier accelerators is politically and operationally significant. It allows Microsoft to deliver frontier AI compute outside the U.S. while framing the move within a controlled, assurance‑based model.

For the UAE​

• AI sovereignty and ecosystem: The new capacity materially increases in‑country compute headroom, which helps attract research labs, startups and regulated enterprises that cannot move sensitive workloads offshore. The physical presence of high‑end GPUs reduces latency and regulatory friction for in‑country AI services.
• Economic and skills potential: Microsoft’s skilling commitments tied to the investment (including multi‑year training targets) aim to seed the workforce needed to operate and build services on top of the new infrastructure.

Geopolitical and export‑control dimensions​

The authorization to place Blackwell‑class hardware in the UAE sets a precedent for how advanced AI accelerators can be exported under conditional regimes. This is a diplomatic as much as a technical act: the U.S. considered national‑security trade‑offs before permitting the shipments, and the licenses define operational guardrails that will shape future cross‑border compute flows.

---

Commercial and operational impacts for enterprises and IT teams​

For Windows‑centric enterprises and Azure customers the announcement changes procurement and architecture choices in several concrete ways:

• Sovereign hosting becomes actionable: Organizations operating in finance, healthcare, energy and government can more credibly procure in‑country Copilot and inference services with onshore processing guarantees. Contractual and eligibility details will determine practical reach.
• Hybrid AI patterns: Expect hybrid topologies where sensitive data ingress remains on‑premises or in local clouds, while GPU‑heavy training and large‑context inference burst into Khazna/Azure capacity.
• Procurement checklist changes: Buyers should demand explicit audit rights, portability guarantees, incident response SLAs, and energy‑sourcing commitments before signing long‑term purchase agreements tied to local capacity.

Practical, ranked steps IT leaders should take now:

• Inventory workloads that require onshore processing and map them to Microsoft’s product‑residency timelines.
• Insert auditability, data portability, and vendor‑exit clauses into contractual RFPs and MOUs.
• Require transparent energy and sustainability commitments (PPAs, delivery milestones) for any capacity-dependent pricing.
• Harden multi‑region failover and backup strategies — sovereign hosting reduces legal friction but does not eliminate service outage risk.

---

Risks, unknowns and cautionary points​

• Timeline uncertainty: Public statements place initial capacity coming online before the end of 2026, but delivery depends on substations, transformer builds, supply chains for cooled racks, and GPU shipments. Treat the timeline as conditional and phased.
• GPU‑equivalent arithmetic is an approximation: Microsoft and media outlets report GPU inventories and “A100‑equivalent” figures as shorthand. These equivalencies are useful proxies but do not precisely map to training throughput or inference capacity for a given workload. Performance depends on interconnect topology, memory pooling, and model parallelism; treat the raw counts as illustrative rather than definitive. Flagged as an approximation and not an absolute inventory.
• Governance and transparency concerns: The effectiveness of the intergovernmental assurance frameworks and any binding oversight mechanisms will be judged by third‑party audits, breach disclosures, and the specificity of contractual audit rights — none of which were released in full at announcement time. Without independent, verifiable audit trails, public trust in “sovereign” assurances will remain conditional.
• Concentration and vendor lock‑in: Deep integration between a sovereign operator (G42/Khazna) and a single hyperscaler (Microsoft) concentrates risk. Buyers should negotiate portability, evacuation and contingency options to avoid being locked into a single supply chain or vendor.
• Energy and sustainability delivery risk: Promises of renewable supply or low-carbon operation require contractually firmed delivery dates and public carbon‑intensity metrics at the facility level. Absent those, sustainability claims remain aspirational.

---

What to watch next (signals that will validate progress)​

• Publication of site permits, PPA contracts, or grid interconnection agreements indicating the utility timelines and firmed energy supply.
• Third‑party audit summaries or redacted compliance attestations showing export‑control and hardware custody processes for licensed accelerators.
• Microsoft product announcements that convert infrastructure into product guarantees (for example, formal availability dates for in‑country Microsoft 365 Copilot processing for qualified UAE customers).
• Public disclosures of capex breakdowns or Khazna financial filings that illuminate the cost per MW and financing structure.
• Independent coverage of hardware racking and commissioning (photos, customs filings, or lease notices) that corroborate compute arrival timelines.

---

Strengths and strategic value — a critical assessment​

• Scale and capability: A 200 MW commitment is a genuine, material step toward making the UAE a competitive AI infrastructure hub. The power envelope is large enough to host thousands of GPU‑dense racks when deployed with modern cooling, giving local researchers, enterprises and public agencies access to frontier compute.
• Integrated strategy: Pairing compute with product residency commitments, skilling programs, and a Responsible AI institutional layer is a sophisticated play that recognizes regulation and skills are as crucial as raw FLOPS.
• Policy precedent: The export licensing approvals are strategically important: they show how frontier hardware can flow to trusted partners under conditional frameworks, which could become a template for allied nations seeking onshore compute without breaching export constraints.

At the same time, notable limitations and risks temper the upside:

• Opaque operational detail: The lack of public, granular capex, site and governance documentation reduces the ability of customers and external observers to validate claims. Independent auditability is key to converting a program into durable trust.
• Execution complexity: Building hyperscale AI‑capable halls, securing energy at scale, and integrating very high density GPU hardware is technically non‑trivial; any slippage in one area (power, cooling, supply chain) will cascade into delays.

---

Bottom line for WindowsForum readers and IT decision‑makers​

Microsoft and G42’s 200 MW data‑center expansion through Khazna Data Centers is a major, credible commitment that materially shifts the AI infrastructure landscape in the Gulf. It makes in‑country, Azure‑grade AI hosting more attainable for regulated customers and signals a new operational model where hyperscalers deploy frontier hardware under intergovernmental assurance frameworks.
However, operational delivery — not headlines — will determine whether the program realizes its promise. Enterprise buyers should treat Microsoft’s product‑residency statements as promising but conditional, and insist on contractual auditability, portability, and firmed energy and resiliency commitments before signing long‑term agreements tied to this capacity. The GPU numbers and A100‑equivalent tallies are useful for scale context, but they should be interpreted as proxies for performance rather than literal hardware inventories.

---

Conclusion​

This 200 MW expansion anchors Microsoft’s $15.2 billion UAE program in a tangible infrastructure commitment and marks another step in the geopolitically complex, technically exacting race to localize frontier AI compute. The announcement combines engineering ambition, diplomatic negotiation over export controls, and a coordinated push to pair capacity with governance and skills investments. For enterprises and IT leaders, the practical takeaway is straightforward: the capacity will create new possibilities for sovereign AI services — but procurement teams must demand the transparency, contractual protections and technical detail necessary to convert those possibilities into operational reality.

---

Source: TechInformed Microsoft, G42 announce 200-megawatt UAE data center expansion - TechInformed

 

[ChatGPT]

Microsoft and Abu Dhabi’s G42 have committed to a 200‑megawatt expansion of data‑centre capacity in the United Arab Emirates, a move packaged inside Microsoft’s broader $15.2 billion investment commitment to the country and delivered through G42’s Khazna Data Centers — a strategic push that combines frontier GPU compute, sovereign cloud promises, and a Responsible AI governance agenda.

Background / Overview​

The announcement is the latest, most concrete step in a deepening strategic partnership between Microsoft and G42 that stretches beyond typical vendor‑client ties into co‑investment and co‑governance. Microsoft has publicly framed the UAE program as a multiyear package of capital and operating expenditures that began in 2023 and runs through 2029, with roughly $7.3 billion already invested and $7.9 billion earmarked for 2026–2029. That $15.2 billion envelope now incorporates a specific, measurable infrastructure deliverable: 200 MW of IT load that Khazna Data Centers will build and operate, with initial capacity expected to start coming online before the end of 2026. This expansion is explicitly tied to enabling advanced AI workloads in‑region. Microsoft and partners have secured U.S. export licenses allowing shipment of high‑end NVIDIA accelerators — described in public reporting as GB/Blackwell‑class systems — to UAE facilities. The partners present this as a package: local compute capacity + product residency promises (notably for select Microsoft 365 Copilot processing) + governance and skills investments, including the Responsible AI Future Foundation and an Abu Dhabi AI for Good Lab.

---

What “200 MW” actually means — technical translation​

Power is the limiting resource​

When vendors announce a figure like 200 megawatts (MW) for data‑centre capacity they are referring to IT load — the electrical envelope available to serve server racks, accelerators, networking equipment and the cooling infrastructure that keeps them within thermal limits. IT‑load MW is the single most important engineering constraint for modern AI‑first data centres: it determines how many high‑density racks can be supported, whether liquid cooling is required, and what kinds of fault‑tolerant electrical topologies must be in place.
At conservative estimates (around 10 kW per rack) 200 MW of IT load translates to roughly 20,000 racks. For GPU‑dense AI deployments — where single racks commonly draw 10–40+ kW each — 200 MW can support tens of thousands of accelerators when paired with appropriate cooling and power distribution. These are order‑of‑magnitude calculations; the final usable compute depends on rack density, cooling approach (air vs direct‑to‑chip liquid vs immersion), redundancy (N+1, 2N), and the mix of accelerators deployed.

GPU families and “A100‑equivalents”​

Public statements around this program repeatedly translate compute inventories into “A100‑equivalent” numbers for shorthand. That shorthand can be misleading: newer GB/Blackwell‑class accelerators (GB200/GB300 families) and Blackwell Ultra systems deliver materially higher performance per rack and per watt than older A100 or H100 generations. Consequently, an “A100‑equivalent” metric is best treated as a performance proxy rather than a literal GPU count. Expect deployments to favor rack‑scale, liquid‑cooled pods optimized for the GB/Blackwell topology when the objective is large‑context model training and low‑latency inference.

---

The operational package: more than metal and megawatts​

Product residency and Azure services​

A core commercial promise tied to the expansion is product‑level residency for certain Microsoft services — notably Microsoft 365 Copilot — allowing qualified UAE organisations to keep processing and inference inside UAE Azure‑grade infrastructure. That matters to regulated sectors (finance, healthcare, government) that require demonstrable in‑country processing for privacy, compliance and contractual reasons. These are phased commitments and eligibility rules will define practical coverage; infrastructure presence alone does not automatically equal service availability. Enterprises should request concrete feature‑level availability timelines and audit rights before relying on product‑residency claims.

Governance and responsible AI​

Microsoft, G42 and Mohamed bin Zayed University of Artificial Intelligence (MBZUAI) have launched the Responsible AI Future Foundation, and Microsoft has expanded its AI for Good Lab presence in Abu Dhabi. These bodies are positioned as instruments to pair capability growth with policy, safety research, and governance frameworks aimed at ethics, explainability, and domain‑specific risk mitigation. They serve two functions simultaneously: public assurance for partners and customers, and a platform for regionally relevant standards and research. That said, the effectiveness of such institutions depends on transparency, independent oversight, and enforceable accountability mechanisms.

Talent and local capacity building​

Microsoft reiterated its commitment to skill one million people in the UAE by 2027, tying infrastructure expansion to workforce development and local engineering capacity. The reported expansion is accompanied by existing initiatives such as the Global Engineering Development Center and the Abu Dhabi AI for Good Lab, which are intended to feed local industries with trained cloud and AI talent. These are important signals for long‑term ecosystem building; converting training commitments into sustainable employment and product development requires credible program design, local hiring pipelines, and measurable placement and retention metrics.

---

Geopolitics, export controls and tech diplomacy​

The move is as much diplomatic as technical. The U.S. Commerce Department’s decision to approve exports of advanced NVIDIA accelerators to support UAE deployments sets a precedent in the geopolitically sensitive landscape of frontier AI hardware controls. That licensing indicates U.S. willingness to enable allied partner deployments under defined safeguards rather than imposing blanket embargoes — a model of conditional export that balances industrial cooperation with national‑security concerns. Observers will watch the governance and verification arrangements that attach to these licenses, as they could inform future export control frameworks. This arrangement also demonstrates how hyperscalers and national actors can negotiate frameworks that allow high‑end compute to operate outside U.S. jurisdiction while retaining contractual and technical safeguards. It’s a template that other nations and firms will study closely.

---

Environmental and grid implications​

Energy supply and sustainability questions​

A 200 MW IT load implies a substantially larger grid requirement after accounting for facility overheads, redundancy and cooling plant (PUE). Delivering reliable power at scale requires substation upgrades, long‑term energy procurement (PPAs) and often on‑site generation or storage for firming renewables. Public materials accompanying Khazna’s prior projects emphasize modular, energy‑efficient design and partnerships with local utilities on supply, but the announcement did not publish detailed PPAs or firming strategies for this specific 200 MW tranche. That omission matters: scaling AI compute without clear, auditable plans for renewable procurement and carbon‑firming invites legitimate scrutiny from customers and civil society.

Cooling and facility design​

High‑density GPU racks push datacentre designs toward liquid cooling (direct or immersion) to keep Power Usage Effectiveness (PUE) acceptable. Expect Khazna’s AI‑optimized builds to incorporate these thermal strategies, but operationalizing them at hyperscale requires mature supply‑chain logistics, qualified facilities staff, and proven failure‑mode mitigations. Large‑scale immersion deployments change maintenance, fire safety, and hardware lifecycle processes compared with traditional air‑cooled halls.

---

Economic and market impacts for the UAE and regional cloud buyers​

Boost to the local digital economy​

From an economic lens, onshoring frontier compute and announcing skills programs supply critical ingredients for a national AI value chain: lower latency for domestic services, sovereign hosting for regulated customers, and infrastructure for startups and research labs. The government’s objective to grow the digital economy’s GDP contribution is better served by visible infrastructure commitments like this one. But infrastructure alone does not guarantee local value capture; policy, procurement choices, public‑private partnerships, and local entrepreneurship ecosystems determine whether the UAE retains an outsized share of AI value creation.

Positioning and competition​

For Microsoft and Azure, the expansion strengthens the company’s value proposition in the Gulf and broader Middle East for customers prioritizing local residency and auditability. Competitors and regional providers will respond with their own offers or partnerships, affecting pricing, procurement terms and multi‑cloud buying decisions. For enterprises, the strategic calculus now includes sovereign‑capable Azure services as a tangible option rather than an aspirational promise.

---

Risks, unknowns and points of caution​

• Timeline risk: Public statements set a target of initial capacity before the end of 2026, but megawatt‑scale builds depend on substations, transformer delivery, grid upgrades, and hardware supply timelines. Treat the timeline as conditional and phased rather than a single date‑certain rollout.
• Opacity on locations and contracts: Exact site locations, capex breakdown per MW, and contract details (who pays for what, guaranteed availability SLAs) were not disclosed with the headline announcement — information enterprises and regulators will need to evaluate real operational readiness.
• Environmental externalities: Without published, auditable PPAs, firmed renewable plans, and emissions‑reduction milestones, large AI deployments risk becoming power‑intensive installations with limited transparency on climate impact. Customers with ESG mandates should demand clear energy sourcing commitments.
• Concentration and single‑operator risk: The Khazna/G42–Microsoft model concentrates significant sovereign capability in a small set of actors. That concentration can accelerate capability delivery but also raises questions about procurement diversity, market competition and long‑term resilience.
• Export‑control conditionality: The U.S. export licenses that enable GB/Blackwell shipments are attached to conditions. Any future change in policy or geopolitical dynamics could affect hardware replenishment and upgrades, potentially constraining long‑term operations.
• Governance and auditability: The Responsible AI Future Foundation is a positive governance step, but its impact will depend on independent oversight, publication of standards, and mechanisms for third‑party audits. Without those, governance signals risk being perceived as headline framing rather than enforceable guardrails.

---

Practical guidance for enterprises and IT leaders (Windows‑centric and Azure customers)​

• Confirm product‑residency eligibility and feature timelines: Request a day‑by‑day availability matrix for Microsoft 365 Copilot features and which "qualified" customers will be supported in‑country and when.
• Insist on contractual audit rights: Security, export‑control compliance and data residency claims should be auditable by third parties or by independent compliance firms.
• Demand energy and sustainability commitments: Include PPA milestones, renewable delivery schedules, and carbon accounting in procurement contracts if ESG compliance matters.
• Prepare hybrid failover architecture: Do not assume sovereign hosting eliminates the need for cross‑region geo‑replication and failover; design for multi‑region backups and clear RTO/RPO SLAs.
• Include exit and portability clauses: Ensure data and model portability provisions are explicit, including timelines and technical specifications for migration.
• Validate governance frameworks: Ask for the Responsible AI Future Foundation’s published standards, audit schedules and participation terms for external researchers and civil‑society oversight.
• Reassess supplier concentration risk: If critical services or workloads will rely heavily on a single operator or region, plan contingency and multi‑vendor options.

---

Why this matters for WindowsForum readers​

WindowsForum’s audience spans enterprise Windows administrators, cloud architects, systems integrators and IT managers who must translate strategic infrastructure news into operational decisions. The Microsoft–G42 200 MW expansion is a practical inflection point: it materially increases in‑region high‑performance compute, which can lower latency for Windows‑centric cloud services, enable in‑country Copilot processing for regulated organizations, and change procurement calculus for enterprises that previously relied on offshore regions.
For on‑prem + cloud hybrid deployments, this announcement shifts the balance toward nearby sovereign cloud options for large inference workloads and model hosting. That can reduce egress costs and legal friction but does not remove architectural obligations: robust backup, DR planning and contractual protections remain essential.

---

Final assessment — strengths and strategic upsides​

• Scale and capability: 200 MW is a meaningful, credible addition to UAE AI infrastructure and can support large‑model training and low‑latency inference when paired with GB/Blackwell accelerators.
• Integrated strategy: Pairing compute with product residency, governance institutions and skills programs increases the odds that capacity translates into usable services for regulated sectors rather than being idle metal.
• Diplomatic precedent: The successful export licensing shows a model for trusted, auditable offshore deployments of frontier AI hardware under U.S. authorization. This can unlock cooperative models for allied nations.

Final assessment — principal risks and what to watch next​

• Delivery execution: Substation builds, supply chains for liquid‑cooled racks, and GPU shipments are friction points that can delay timelines. Treat the “before end of 2026” claim as conditional until substations are commissioned and third‑party attestations are available.
• Transparency gaps: Missing public detail on energy sourcing, capex allocation and site locations leaves stakeholders with open questions on sustainability and resilience.
• Governance depth: The Responsible AI Foundation is a positive sign, but momentum should be measured by published standards, open governance processes and independent audits — not only announcements.

---

The Microsoft–G42 200 MW announcement is a significant chapter in the UAE’s ambition to become an AI hub. It pairs tangible infrastructure scale with product‑level residency promises, governance initiatives, and a publicized commitment to skilling. For enterprises, the moment shifts a long‑standing question — “Can we run frontier AI onshore?” — from theoretical to practical, but not yet definitive. The path from press release to production will be determined by execution fidelity: substations built, GPUs racked and validated under clear audit regimes, renewable energy contracted and disclosed, and governance frameworks that demonstrate independent oversight. For IT leaders, the pragmatic posture is to prepare, validate, and contract defensively — to take advantage of the new sovereign options while guarding against timeline slippage, energy uncertainty, and governance opacity.

---

Source: Communications Today Microsoft, G42 announce 200 MW data centre capacity expansion | Communications Today