Inside G42 and MGX: How Abu Dhabi Is Building the Gulf's Most Connected AI Empire
G42 and MGX are Abu Dhabi's two-part sovereign AI machine. G42 builds the physical infrastructure and deploys frontier AI models. MGX deploys the capital at portfolio scale, backing OpenAI, Anthropic, and xAI while closing the largest dedicated AI fund ever raised at $49 billion. Together they represent the most globally connected sovereign AI operation in the world. This is the definitive analysis of how Abu Dhabi built it, what it means for enterprise AI buyers, and why the UAE leads the world in per capita generative AI adoption.
The UAE's AI story is the most diplomatically sophisticated in the Gulf. Where Saudi Arabia's HUMAIN leads with sovereign scale and chip density, and where Kuwait's Helix and Qatar's Qai have each built focused national platforms, Abu Dhabi has constructed something more complex: a two-vehicle architecture that separates infrastructure operations from capital deployment, allowing each vehicle to move at the speed appropriate to its function. G42 builds and operates. MGX invests and scales. The combination has made Abu Dhabi the most globally connected node in the Gulf sovereign AI network.
This post is the fourth in Integrated.Social's Gulf Sovereign AI Series. For the full comparison of all four Gulf champions, see the Gulf Sovereign AI Series hub page [blocked].
G42: The Infrastructure and Model Layer
Group 42 Holding Ltd, known as G42, was founded in 2018 in Abu Dhabi and is chaired by Sheikh Tahnoon bin Zayed, the UAE's National Security Advisor and brother of President Mohamed bin Zayed Al Nahyan. Its Group CEO is Peng Xiao, who at the World Government Summit in February 2026 articulated the scale of G42's ambition with characteristic precision: "We expect to output up to 100 trillion tokens a day. That is the scale of intelligence we are building for the UAE and beyond."
G42 is a technology holding group operating across four primary domains. The first is AI infrastructure, operated through Core42, G42's sovereign cloud and AI infrastructure subsidiary. The second is AI model development, operated through Inception, G42's research and model development arm. The third is vertical AI applications, including M42 in healthcare and AIQ in energy (a joint venture with ADNOC). The fourth is geospatial AI, operated through Bayanat.
The strategic pivot that defines G42's current era is its 2024 decision to divest all Chinese technology partnerships, remove Huawei equipment from its infrastructure, migrate to Microsoft Azure, and accept US security protocols and export control frameworks. This was not a commercial decision. It was a geopolitical one. By aligning with US regulatory requirements, G42 became the only company in the region to qualify for advanced AI chip exports under US Department of Commerce and Bureau of Industry and Security guidelines. The consequence is that G42 now operates with access to NVIDIA's most advanced chips that no other Gulf AI company can match at comparable scale.
Stargate UAE: The 5-Gigawatt Bet
The most consequential project in G42's portfolio is Stargate UAE, the 1-gigawatt AI compute cluster being built for OpenAI in partnership with Oracle, Cisco, NVIDIA, and SoftBank Group. Stargate UAE is the international anchor of the broader UAE-US AI Campus, a 5-gigawatt, 10-square-mile AI infrastructure hub in Abu Dhabi that will become the largest AI infrastructure project outside the United States.
The project was announced in May 2025 by Presidents Mohamed bin Zayed Al Nahyan and Donald Trump, and its first 200 megawatts are targeted to go live in Q3 2026. OpenAI has stated that Stargate UAE will serve "up to half the world's population," reflecting its positioning as the primary AI infrastructure hub for the Global South, a market of approximately 4 billion people that no American data center can serve with competitive latency.
The chip supply chain for Stargate UAE was secured in November 2025, when the US approved the export of advanced AI semiconductors to G42. The approval covered the equivalent of 21,500 NVIDIA A100 GPUs already accumulated in the UAE (a combination of A100, H100, and H200 chips), plus new licenses for the equivalent of 60,400 additional A100 chips in the form of NVIDIA's GB300 GPUs. These chips are governed by the Regulated Technology Environment (RTE), a compliance framework pioneered by G42 and approved under US export control guidelines.
The UAE-US AI Campus is designed to serve as the compute foundation for AWS, Google, Meta, Microsoft, and xAI, all of which are in discussions to become tenants or strategic partners. The facility's location in Abu Dhabi provides structural advantages that no American data center can replicate: proximity to the Global South's population centers, UAE electricity prices significantly below US averages, and a geopolitical positioning that allows the campus to serve markets where US-origin data center infrastructure faces regulatory or diplomatic friction.
Core42: The Global Infrastructure Operator
Core42 is G42's sovereign cloud and AI infrastructure subsidiary, and it is the operational vehicle through which G42's infrastructure ambitions are executed globally. As of July 2026, Core42 operates 10 sites globally, with a footprint that spans Abu Dhabi, the United States, and Europe.
The US footprint is particularly significant. Core42's Lake Mariner campus in Buffalo, New York, was expanded in June 2026 from 18 megawatts to 60 megawatts of high-performance AI production infrastructure, adding 42 megawatts of capacity in a single expansion. Beyond Lake Mariner, Core42 operates in Dallas, Texas; Sunnyvale and Stockton, California; and Minneapolis, Minnesota, where the Condor Galaxy supercomputers were developed in collaboration with Cerebras. The Maximus-01 supercomputer at Lake Mariner ranks 20th on the global TOP500 list. Core42 holds three of the Top 500 supercomputers globally, including the second and third ranked systems in the Middle East.
In Europe, Core42 established its European headquarters in Dublin in 2025 and expanded AI compute deployments across Italy and France. In June 2026, MGX and Bpifrance announced a partnership with Mistral to expand an AI campus in France, extending the G42 ecosystem into the European sovereign AI market.
The Core42 AI Cloud platform, launched in October 2025, is the commercial access layer to this distributed infrastructure. It allows enterprise customers to provision compute across jurisdictions under a consistent operating model, supporting the full AI lifecycle from large-scale training and fine-tuning to real-time inference. For enterprise AI leaders evaluating infrastructure vendors, Core42 represents a category that did not exist two years ago: a sovereign AI infrastructure operator with a global footprint, US regulatory approval, and the ability to serve workloads that require non-US jurisdiction without sacrificing performance.
JAIS and the Arabic AI Model Layer
G42's model development arm, Inception, has built what is arguably the most important Arabic-language AI capability in the world. JAIS is a bilingual Arabic-English large language model named after Jais mountain, the UAE's highest peak. The original JAIS model had 13 billion parameters and was trained on a 395-billion-token Arabic and English dataset. G42 subsequently launched JAIS 70B alongside 20 other AI models specifically designed to advance Arabic natural language processing.
The strategic significance of JAIS extends beyond the Arabic-speaking world's 400 million users. It establishes G42 as the sovereign AI provider for Arabic-language enterprise applications across government, financial services, healthcare, and media, markets where US frontier models have historically underperformed due to limited Arabic training data. JAIS's open-source licensing creates a developer ecosystem that compounds G42's model advantage over time.
G42 has also expanded its multilingual model portfolio. In December 2025, G42 launched an expanded Hindi-English language model, signaling its intent to serve the Indian subcontinent as well as the Arabic-speaking world. For enterprise AI leaders with operations across the Middle East, South Asia, or Africa, G42's model portfolio represents a frontier AI capability that no US or Chinese provider has matched for these language markets.
MGX: The $49 Billion Capital Machine
MGX was established in March 2024 as a joint venture between Mubadala Investment Company and G42, structured as a private investment fund with a traditional GP/LP architecture. Its mandate is to deploy sovereign capital at the portfolio level across the AI technology stack, targeting $100 billion or more in AI investments by 2030.
On July 1, 2026, MGX announced the final close of MGX Fund I at $49 billion in commitments, exceeding its initial $45 billion target and making it one of the largest dedicated AI investment vehicles ever raised. The fund was raised from institutional and private investors across the Gulf, North America, Asia, and Europe. As of the fund close, MGX had backed 14 companies.
The portfolio is a who's who of frontier AI. MGX co-led OpenAI's $122 billion raise in March 2026. It co-led Anthropic's $30 billion raise in February 2026 and participated in Anthropic's $65 billion Series H in May 2026. It participated in Elon Musk's xAI $20 billion raise in January 2026, prior to xAI's merger with SpaceX. The fund is investing across the full AI technology stack, including semiconductors, AI infrastructure, and AI-enabling technologies and platforms.
The MGX portfolio strategy reflects a deliberate decision to maintain exposure to all frontier AI model providers simultaneously, rather than betting on a single winner. This is the sovereign capital equivalent of a venture portfolio strategy: MGX is positioned to benefit from whichever frontier model provider achieves dominance, while its infrastructure investments through G42 and Core42 ensure that the physical layer of AI compute remains under Abu Dhabi's operational control regardless of which model layer wins.
Microsoft and the US-UAE AI Corridor
The Microsoft-G42 partnership is the most strategically significant bilateral technology relationship in the Gulf. Microsoft made a $1.5 billion equity investment in G42 in April 2024, securing a board seat for President Brad Smith. The total Microsoft UAE commitment is $15.2 billion spanning 2023 to 2029.
The breakdown of that commitment is instructive. From 2023 through 2025, Microsoft invested $7.3 billion in the UAE: $1.5 billion in G42 equity, $4.6 billion in data center capital expenditure, and $1.2 billion in local operating expenses. From 2026 through 2029, Microsoft plans to invest a further $7.9 billion: $5.5 billion in data center capex and $2.4 billion in operating expenses. The 2026-2029 tranche will nearly quadruple local compute capacity.
The Microsoft-G42 relationship has produced institutional infrastructure beyond capital. In February 2025, Microsoft, G42, and Mohamed bin Zayed University of Artificial Intelligence co-founded the Responsible AI Future Foundation (RAIFF) in Abu Dhabi, a foundation promoting responsible AI standards across the Middle East and the Global South. Microsoft has established a Global Engineering Development Center in Abu Dhabi and an AI for Good Lab staffed by PhD-level researchers. Microsoft's AI Diffusion Report, published in November 2025, found that the UAE leads the world in per capita generative AI adoption, with 59.4 percent of the population using generative AI, ahead of Singapore at 58.6 percent and significantly ahead of any European country.
The US-UAE AI corridor that G42 and Microsoft have built is the template for how sovereign AI infrastructure can be constructed within US regulatory frameworks while serving non-US markets. It is the reason why the UAE was the first country outside the US to receive NVIDIA GB300 GPU export licenses, and it is the reason why Stargate UAE is the first international Stargate deployment.
What This Means for B2B Enterprise Leaders
The G42/MGX architecture creates four distinct implications for enterprise AI leaders.
The first is model access at global scale. G42's JAIS and multilingual model portfolio, combined with Core42's distributed infrastructure, means that enterprises with operations across the Middle East, South Asia, or Africa can access frontier-quality AI models in local languages through a single infrastructure provider. This is a capability that no US hyperscaler currently offers at comparable quality for Arabic or Hindi.
The second is infrastructure jurisdiction optionality. Core42's 10-site global footprint, spanning the UAE, US, and Europe, allows enterprise workloads to be distributed across jurisdictions based on regulatory requirements, data residency rules, and cost optimization. For enterprises navigating GDPR in Europe, data localisation requirements in the Gulf, or export control compliance in the US, Core42 offers a single infrastructure partner that can serve all three simultaneously.
The third is the MGX portfolio as a signal of frontier AI direction. MGX's simultaneous positions in OpenAI, Anthropic, and xAI represent the most comprehensive exposure to frontier AI model development of any single investor. Enterprise AI leaders tracking which model providers will achieve long-term dominance should monitor MGX's portfolio additions as a leading indicator of where sovereign capital is placing its bets.
The fourth is the Responsible AI Future Foundation as a governance framework. For enterprise buyers in regulated industries, RAIFF's work on responsible AI standards for the Middle East and Global South creates a governance framework that is neither US-centric nor Chinese-centric. This matters for financial services, healthcare, and government technology buyers who need AI governance frameworks that reflect their regulatory environment rather than Silicon Valley's.
For B2B technology vendors seeking to enter Gulf enterprise markets, G42's relationships with UAE government ministries, ADNOC, and the broader Abu Dhabi institutional ecosystem represent the most direct distribution channel to sovereign procurement. The Agentic AI services at Integrated.Social [blocked] and the Preferred Sources Strategy [blocked] are directly relevant to positioning B2B AI offers for Gulf sovereign procurement cycles.
The Gulf Sovereign AI Series
This post is the fourth in Integrated.Social's Gulf Sovereign AI Series. The complete series covers:
Saudi Arabia's HUMAIN and the full-stack sovereign AI model [blocked] that PIF is building from compute infrastructure to enterprise AI marketplace. Kuwait's Helix Digital Infrastructure [blocked], the energy-first AI platform launched by KIA alongside NVIDIA, KKR, and Vistra Energy. Qatar's Qai [blocked], the trusted AI infrastructure company backed by QIA and Brookfield's $20 billion joint venture. And the Gulf Sovereign AI Series hub page [blocked] for the definitive comparison of all four champions.
Frequently Asked Questions
See the FAQ section below for answers to the most common questions about G42, MGX, and the UAE's sovereign AI strategy.
About the Author
Modi Elnadi is the founder of Integrated.Social, a London-based AI growth marketing agency specialising in Agentic AI strategy, AEO/GEO optimisation, and B2B pipeline generation for technology, financial services, and professional services firms operating in the UK, US, and MENA markets. Modi tracks the Gulf sovereign AI buildout as part of Integrated.Social's ongoing coverage of how sovereign capital is reshaping the global enterprise AI vendor landscape. Connect at integrated.social/about.
