Skip to main content
artificial intelligence

The Largest AI Data Center Outside the U.S. Is Being Built in Abu Dhabi. Most of the Jobs Aren't About AI.

By John Hugo

The $30 Billion Bet That Changes OpenAI's Global Footprint

On May 22, 2025, OpenAI announced Stargate UAE, the first international deployment of its Stargate infrastructure platform and the inaugural project under its new "OpenAI for Countries" initiative. The numbers are stark: a $30 billion, 10-square-mile AI campus in Abu Dhabi, scaling to 5 gigawatts of power capacity by 2030. That's roughly the entire electricity consumption of Croatia, dedicated to running over one million NVIDIA accelerator chips. It is, by every available measure, the largest AI infrastructure project outside the United States — and it signals a structural shift in how OpenAI plans to grow.

The deal was not a simple land lease. It was negotiated in close coordination with the U.S. government and formalized during President Trump's visit to the UAE. The partnership structure reveals the strategic calculus on both sides. G42, the Abu Dhabi-based AI company chaired by Sheikh Tahnoon bin Zayed Al Nahyan, holds a 60% stake and serves as the local operating partner, reflecting the UAE's insistence that its sovereign AI infrastructure remain under national control. OpenAI holds 20%, contributing its models, training methodologies, and research expertise. NVIDIA holds 12% as the primary chip supplier, and Oracle holds the remaining 8% for cloud infrastructure and enterprise integration. Cisco and SoftBank round out the alliance.

The dual-investment structure is what makes this more than a real estate deal. Alongside the 1-gigawatt Stargate UAE cluster (with its first 200 megawatts expected online in 2026), the UAE committed to investing in U.S. Stargate infrastructure, building on the U.S.-UAE AI Acceleration Partnership announced the same week. The UAE separately pledged $1.4 trillion in U.S. investments. The message: this is a two-way technology corridor, not an outsourcing arrangement.

For OpenAI, the strategic logic is about compute distribution and geopolitical positioning. The company's existing Stargate infrastructure is concentrated in the United States. Stargate UAE gives OpenAI a sovereign foothold in a region that sits within a 2,000-mile radius of roughly half the world's population, covering South Asia, Africa, and Central Asia, markets where latency from U.S.-based clusters is a real constraint. It also positions OpenAI ahead of competitors: China's efforts to build AI infrastructure partnerships in the Gulf were hampered by the same U.S. export restrictions that led G42 to divest its Chinese technology ties in 2024, and European sovereign AI efforts have moved more slowly.

The UAE, for its part, gets something no other country has: access to OpenAI's most advanced models (including future generations not yet publicly released) deployed on sovereign territory. The entire UAE population will receive ChatGPT Plus subscriptions, a global first. Omar Al Olama, the UAE's Minister of State for AI, Digital Economy and Remote Work Applications, framed it at the Machines Can Think summit in January 2026: the campus was "larger than Monaco" and cost more than $30 billion, up from initial estimates in the $20 billion range. "The UAE kept exporting to ensure that countries that do not have the ability to build these models are able to leverage on what the UAE has created," he said.

Construction broke ground on March 20, 2026, in Abu Dhabi's Masdar City technology zone, with South Korea joining as a strategic technology partner. The first phase, 200 megawatts, is targeted for Q3 2026. Full buildout to 5 gigawatts is expected by 2030. OpenAI's own job board already lists a Technical Deployment Lead role based in Abu Dhabi, a signal that the hiring ramp has started. The question is no longer whether this project will reshape OpenAI's global footprint, but how fast the workforce needed to run it can be built.

5 Gigawatts of AI Compute — and the Engineers Needed to Build It

The Stargate UAE campus broke ground in Abu Dhabi on March 20, 2026, and the numbers attached to it are difficult to process at first pass. The facility will span 10 square miles inside the Masdar City technology zone. At full build-out, it will consume 5 gigawatts of electricity, roughly what Croatia or Lithuania uses as a country. The project houses over one million NVIDIA Blackwell Ultra and successor AI accelerator chips in a cluster that, by raw processing power, exceeds every commercial AI data center currently operating in Europe combined.

That scale doesn't just describe a building. It describes a workforce problem.

Power first. The campus needs dedicated generation on-site: natural gas turbines for baseload, a 1.5-gigawatt solar array, battery storage, and a dedicated transmission corridor approved by the Abu Dhabi Department of Energy connecting to the national grid. Backup capacity comes from the Emirates Nuclear Energy Corporation's Barakah nuclear plant. Each of those systems — gas turbines, solar integration, battery management, grid interconnection, nuclear backup — requires its own engineering teams for design, commissioning, and ongoing operations. A 5-gigawatt power build is not a data center project with some electrical work bolted on. It is an energy project that happens to serve a data center.

Cooling second. One million advanced GPU racks generate heat at a volume that air conditioning cannot handle. The campus is deploying advanced liquid cooling systems. Designing, installing, and maintaining liquid cooling at this scale — piping networks, heat exchangers, coolant distribution units, leak detection, redundancy planning — is a discipline that most traditional data center operators have never had to staff for. It pulls in mechanical engineers, fluid systems specialists, and HVAC engineers who understand thermal load at the megawatt level.

The data center layer itself. The campus is structured in three phases. Phase One, targeting completion by the end of 2027, delivers the first two data center buildings at 1 gigawatt of capacity plus core power and cooling. Phase Two, expected mid-2028, adds three more buildings, the solar array, and primary research labs. Phase Three, finishing around 2030, brings the facility to its 5-gigawatt design capacity and adds advanced manufacturing and testing facilities. Each phase requires construction-phase engineers — people who can read a data center floor plan in Revit, design containment systems, lay out rack elevations, and commission UPS systems — followed by operations-phase staff who run the facility day to day.

The job postings already appearing in the Abu Dhabi market reflect this. Khazna Data Centers is hiring critical engineers specializing in chillers. Naukrigulf lists data center engineer roles requiring experience with UPS systems and cooling equipment. These are not speculative listings. They are the leading edge of a hiring wave that will intensify as Phase One construction accelerates through 2026.

What 5 gigawatts means in workforce terms. DigitalDubai.ai reports the project is expected to create roughly 6,000 direct jobs during construction and over 3,500 permanent positions once fully operational. The partnership agreement mandates that at least 30% of the permanent workforce be UAE nationals, which has already triggered a scholarship pipeline: G42 is sending 200 Emirati students annually to universities in the US and UK specifically to build talent for the campus.

Infrastructure System Scale Engineering Disciplines Required
Power generation & distribution 5 GW (gas turbines, solar, battery, grid interconnect) Electrical, power systems, grid integration, energy storage
Liquid cooling Advanced liquid cooling, high-density GPU racks Mechanical, HVAC, fluid systems, thermal engineering
Data center buildings 3 phases, 10 sq mi campus Structural, construction, commissioning, BIM/Revit design
GPU cluster operations 1M+ NVIDIA accelerator chips Systems, network (400Gb/s InfiniBand), hardware integration
Backup & redundancy Barakah nuclear plant interconnect Nuclear safety, reliability engineering, grid resilience

The takeaway is straightforward: building frontier-AI compute at this scale is not primarily an AI problem. It is an infrastructure problem. The engineers who will make Stargate UAE operational are power systems engineers, cooling specialists, construction project managers, and commissioning leads — people who can deliver a gigawatt-scale energy and cooling backbone before a single GPU rack goes live. The AI talent gets the headlines. The infrastructure talent gets the campus built.

What Roles Are Actually Open in Abu Dhabi

The Stargate UAE project isn't just a compute play. It's a workforce buildout that reveals, in concrete terms, what frontier-scale AI infrastructure demands from the people who construct and run it. The roles being recruited across OpenAI, G42, and NVIDIA in Abu Dhabi map directly onto the technical realities of a 5-gigawatt campus, and they tell a different story than the research-heavy hiring most people associate with these companies.

OpenAI's own careers page lists a Technical Deployment Lead — UAE based in Abu Dhabi, one of the clearest signals that the company is staffing on the ground for Stargate operations. The role sits within a broader hiring push: OpenAI's careers portal shows the company adding roles spanning identity engineering, trust and safety, and strategic sales, but the Abu Dhabi deployment lead is the one tied directly to the physical infrastructure buildout.

G42, as the primary developer and operator of the facility, is the largest local employer in the cluster. The company's careers portal lists openings for engineers working at "the intersection of digital and physical intelligence," including a Smart Contract Engineer role focused on machine-verifiable commitments and autonomous coordination across decentralized environments. G42's LinkedIn presence positions the company as a hub for AI and cloud computing talent in the Gulf, and the Stargate project gives that recruiting pitch real infrastructure behind it.

NVIDIA's role is more specialized but no less critical. The company is supplying its Grace Blackwell GB300 systems for the campus, each combining 72 Blackwell Ultra B300 chips with 36 Grace CPUs, which means demand for engineers who can deploy, optimize, and maintain GPU clusters at a scale that doesn't yet exist anywhere outside a handful of US facilities. The Abu Dhabi campus will be the first place many of these systems operate outside American soil.

JobXDubai's analysis of the UAE AI job market identifies five role categories seeing the sharpest demand growth: data engineers, AI software developers, machine learning engineers, cybersecurity specialists, and data scientists. Mid-level AI roles in the UAE currently pay between AED 250,000 and AED 400,000 annually (roughly $68,000–$109,000), per JobXDubai's reporting, though the scarcity of qualified infrastructure engineers is already pushing those figures upward.

What's notable is the seniority mix. This isn't an entry-level hiring wave. The Technical Deployment Lead role at OpenAI, the smart contract and protocol work at G42, and the GPU-cluster operations NVIDIA must staff all point to mid-career and senior engineers, people who have built and run large-scale systems before. The 200-megawatt first phase going live in 2026 means the most intense hiring window is now through late 2026, with the full 5-gigawatt buildout sustaining demand well into the second half of the decade.

For engineers watching the AI sector, the signal is straightforward: the biggest infrastructure buildout in AI history is recruiting in Abu Dhabi, and the roles are less about training models and more about keeping the lights on (literally) at a facility larger than Monaco.

Why Abu Dhabi Won This Deal

The UAE didn't end up as the site for OpenAI's first major international Stargate deployment by accident. In May 2025, the US and UAE signed the AI Acceleration Partnership, a bilateral framework that explicitly committed both governments to building a 5 GW AI technology cluster in Abu Dhabi, anchored by a 1 GW AI data center built to US security standards. The agreement, announced during President Trump's state visit to Abu Dhabi, tied the project to a broader $1.4 trillion UAE investment commitment in the US over the following decade. This wasn't a vendor selection. It was a geopolitical alignment codified at the highest level of both governments.

The strategic logic runs on three tracks. First, power. Abu Dhabi can deliver industrial-scale electricity at a pace and cost that most US and European jurisdictions cannot match. Building a 5 GW compute cluster in Texas or Virginia would require navigating years of permitting, grid upgrades, and local opposition. The UAE's state-backed energy infrastructure compresses that timeline dramatically. Second, capital. Abu Dhabi's MGX, the state-backed AI investment firm managing over $100 billion in assets and tied to Mubadala and G42, is a direct financial participant in the Stargate project. The money is local, patient, and aligned with the timeline. Third, geography. Abu Dhabi sits between the compute-hungry markets of South Asia, the Middle East, and Africa, regions where demand for AI inference capacity is growing but where no comparable infrastructure exists. The cluster is designed to serve regional computation demand, not just host US companies' overflow.

For the global AI workforce, this matters because it breaks a long-standing pattern. Frontier-scale compute infrastructure has been concentrated in a handful of US locations (Northern Virginia, Oregon, Texas) with smaller clusters in Ireland, Singapore, and the Netherlands. Abu Dhabi's 5 GW cluster is larger than any single US site, and it's being built with a deliberate workforce development mandate. The UAE's National AI Strategy, in place since 2017, made the country the first in the world to appoint a dedicated AI minister. That policy infrastructure is now translating into hiring demand.

The catch is the talent gap. AI salaries in the UAE match or exceed US levels, reflecting the country's willingness to pay for scarce skills. But demand for AI professionals (especially in data-center engineering, power systems, and GPU-cluster operations) is outpacing local supply. G42 has built four times the computing capacity in the US compared to the UAE, largely because the US talent pool for operating that infrastructure is deeper. The Stargate project is, in part, an attempt to close that gap by creating enough concentrated demand to attract and retain engineers who would otherwise stay in Austin or San Francisco.

The 2026 Ramp: What the First 200MW Phase Reveals

The 200-megawatt first phase of Stargate UAE is now a physical construction site, not a slide deck. More than 5,000 workers are deployed on the Abu Dhabi campus, over 100,000 cubic metres of concrete has been poured, and the steelwork alone weighs roughly 1.5 times the Eiffel Tower. All long-lead equipment has been procured. First mechanical systems are already on site. The target: operational delivery in the third quarter of 2026.

That timeline compresses the hiring window in ways most infrastructure projects don't. A conventional hyperscale data center recruits its operations team 12 to 18 months before go-live. Stargate's 200MW phase is on track to flip that sequence: construction and workforce buildout are running in parallel, not in series. The facility needs roughly 100,000 NVIDIA Grace Blackwell GB300 chips powering about 1,400 servers at Phase 1 alone, and each rack is expected to exceed 100 kW, demanding direct-to-chip liquid cooling that most Gulf-region technicians have never worked at scale.

The workforce math gets starker when you look beyond Phase 1. The full 1-gigawatt cluster is expected to reach completion around 2029, with the broader 5-gigawatt campus spanning 19.2 square kilometres, roughly nine times the size of Monaco. Industry analysts project the next 300 to 400 megawatts coming online in 2026–2027, scaling to 600–700 megawatts by 2027–2028. Each tranche requires its own wave of power engineers, cooling specialists, network architects, and GPU-cluster operators. The ramp isn't a single hiring surge. It's a four-year staircase.

OpenAI's own careers page already reflects the early signal: a Technical Deployment Lead role for Abu Dhabi is listed alongside the company's broader hiring push. That's the tip of a workforce buildout that will need hundreds of specialized engineers on the ground before the first training workloads run in late 2026, and thousands more if the campus hits its 5GW target on schedule.

What Frontier-AI Infrastructure Jobs Actually Pay in 2025

The money is real. An AI engineer in Dubai with one to three years of experience earns around AED 215,000 a year, according to SalaryExpert's survey data. Push past eight years and that figure climbs to roughly AED 350,000. Entry-level roles start near AED 9,000 per month, while senior engineers report closer to AED 36,000, per JobX Dubai's September 2025 update. None of it is taxed.

But the compensation story is more nuanced than the gross numbers suggest. The UAE Lifeguide's 2025 salary guide puts the monthly range for AI engineers at AED 30,000 to AED 45,000 for full-time, in-office roles, with machine learning engineers and AI product managers occasionally breaking AED 55,000 at the top end. Data centre engineers, the people who keep the physical infrastructure running, sit lower — Glassdoor's median estimate for Dubai is AED 9,000 per month. The gap between a GPU-cluster software role and a power-systems engineering role can be a factor of three or four, even on the same campus.

What the job actually involves depends heavily on which layer of the stack you occupy. GPU cluster operations engineers spend their days managing job scheduling across thousands of NVIDIA accelerators, troubleshooting network fabric issues at the rack level, and working with site reliability teams to keep training runs from failing at hour 40 of a 48-hour job. It is closer to hyperscale DevOps than to research engineering. Career transition guides on aicareerfinder.com describe a path from GPU cluster engineer into higher-paying AI infrastructure architect roles, with the NVIDIA Certified Professional AI Operations (NCP-AIO) credential cited as a differentiator by multiple certification prep sites.

Data centre engineers and power-systems specialists face a different rhythm. The Stargate UAE campus will eventually demand 5 gigawatts of capacity. Someone has to design the electrical distribution, the cooling loops, the redundancy architecture. These roles are less about code and more about mechanical and electrical engineering at scale — coordinating with local utilities, managing contractor timelines, and ensuring uptime SLAs that frontier-AI training workloads make unforgiving. Abu Dhabi's summer ambient temperatures above 45°C add a design constraint that most US-based engineers never confront.

Benefits packages in the UAE typically include private health insurance, annual flight allowances, housing and transportation stipends, and end-of-service gratuity under UAE labour law. Senior hires relocating from the US or Europe may negotiate retention bonuses or relocation allowances. Stock options are rare outside multinational R&D divisions and a small number of AI-focused startups.

OpenAI's own hiring activity in the region signals what the company values on the ground. OpenAI's careers page shows a Technical Deployment Lead position based in Abu Dhabi, one of the few listings tied directly to the UAE rather than San Francisco. That role sits alongside US-based positions paying $293,000 to $515,000 a year, a reminder that the Abu Dhabi campus is an extension of a compensation structure still anchored to Bay Area benchmarks, adjusted for local cost of living and tax treatment.

The career trajectory for engineers entering through projects like Stargate UAE is still being written. The first 200 MW phase will need hands-on operators and commissioning engineers now. The full 5 GW buildout will need programme managers, vendor-relationship leads, and people who understand both the compute layer and the construction layer well enough to coordinate between them. Engineers who can bridge that gap (who can talk to both the CUDA kernel team and the cooling-tower contractor) are the ones who will define what this workforce looks like by 2027.


Working in AI? Zero G Talent tracks the openings: browse AI jobs, openings at OpenAI and World Labs, and the people building the field.

Ready to Start Your Space Career?

Browse artificial intelligence jobs and find your next opportunity.

View artificial intelligence Jobs