The Rival on the Other Shore
G42 is the entity that Saudi AI competes against most directly, most visibly, and most consequentially. Founded in Abu Dhabi, backed by the UAE’s royal family and operating within the MBZUAI ecosystem that Crown Prince Mohammed bin Zayed has built over the past decade, G42 represents the UAE’s answer to the same fundamental question that Humain is trying to answer for Saudi Arabia: can an Arab country become a genuine participant in the global AI technology landscape, not just a customer of it?
The UAE-Saudi AI competition is not zero-sum — both countries are building capacity that serves the broader Middle East and, in the UAE’s case, a global client base. But it is real, it is well-resourced on both sides, and its outcome will determine which Gulf state emerges as the default AI partner for the developing world’s sovereign AI programs. For vendors, investors, and policy professionals tracking the region, understanding G42’s position requires understanding both what it has built and how it has navigated the political environment that nearly derailed it.
The Microsoft Alignment: What $1.5 Billion Signals
Microsoft’s $1.5 billion investment for a 5% stake in G42 — announced in 2024 — was not primarily a financial transaction. The valuation it implies is significant (approximately $30 billion), but the strategic signal it sends is more important: G42 has secured the explicit endorsement of the world’s largest enterprise technology company, with all the compliance, governance, and geopolitical alignment that endorsement implies.
The Microsoft investment came with conditions that G42 publicly accepted: selling its investments in Chinese technology companies, removing Huawei equipment from its data centers, and reorienting its technology partnerships exclusively toward US and allied vendors. G42 sold its stake in ByteDance’s parent company, negotiated the removal of Huawei network equipment from Core42 data centers, and restructured its board to include representatives acceptable to US national security reviewers.
This restructuring was not voluntary in the sense of being strategically preferred by G42’s leadership independent of external pressure. The US government — operating through Commerce Department export control processes and intelligence community communications to Microsoft — had made clear that companies with significant Chinese technology partnerships would face increasing difficulty accessing advanced US AI chips and cloud partnerships. G42’s Chinese investments were a liability that threatened to cut off access to NVIDIA GPUs and hyperscaler partnerships without which G42 could not operate at frontier AI scale.
The decision to accept Microsoft’s conditions and sell the Chinese investments was, in retrospect, the most consequential strategic decision G42 has made. It chose US alignment over hedged neutrality, and the payoff has been substantial: NVIDIA GPU access under the BIS Tier-2 framework, Microsoft Azure partnership for cloud AI services, and the credibility with Western technology companies that the MBZUAI research reputation alone could not provide.
For Saudi Arabia’s AI strategy, the G42 restructuring is instructive. Saudi Arabia faces the same choice implicitly — the US AI Diffusion framework’s Tier-2 classification means that large NVIDIA GPU deployments require a US-Saudi alignment that constrains other partnerships — but Saudi Arabia’s starting position is stronger than the UAE’s was. Saudi Arabia does not have comparable Chinese technology investments to divest, and Humain’s partnership architecture has been built from the beginning on US technology alignment.
Jais vs. Allam: The Arabic LLM Competition
G42’s most direct product-level competition with Saudi AI is in Arabic large language models. The Jais model — a 70-billion-parameter Arabic LLM developed by MBZUAI (Mohamed bin Zayed University of AI) and deployed through G42’s Core42 infrastructure — has been positioned as the leading open Arabic LLM by parameter count and benchmark performance.
Jais was released as an open-weight model, meaning that researchers and organizations can download and deploy it without API fees. This open release strategy mirrors the approach of Meta’s Llama models and reflects MBZUAI’s academic culture, which prioritizes research impact and citation count alongside commercial deployment. The 70 billion parameters of Jais exceeds the 34 billion parameters of Saudi Arabia’s Allam, and on several Arabic language benchmarks Jais has shown competitive or superior performance.
The Saudi response has been measured rather than reactive. SDAIA has continued Allam’s development — Allam-2 development is ongoing — while emphasizing the sovereign AI deployment stack that Allam enables rather than engaging in a direct parameter-count comparison. The more sophisticated benchmark competition is not parameters but performance on Saudi-specific language tasks: Gulf Arabic dialect understanding, Arabic legal and regulatory text interpretation, and Arabic language interfaces for Saudi government services.
This application-specific competition is harder to score with public benchmarks but more commercially relevant. A model that scores 3% better on academic Arabic NLP benchmarks but cannot reliably process Saudi Ministry of Finance document formats is not the better product for SDAIA’s purposes. The Allam-Jais competition is ultimately decided not by MBZUAI papers but by which model is more embedded in the government and enterprise applications that matter to Arabic-language users.
The geographic split has commercial logic: UAE-based enterprises and the international Arabic-language market (North Africa, Levant, diaspora) are natural Jais customers through G42’s commercial channels; Saudi government ministries and Saudi enterprises are natural Allam customers. The competition is most intense in the middle: Gulf Cooperation Council enterprises that could plausibly use either, and international Arabic-language AI applications where both models compete.
Core42, MBZUAI, and Inception: G42’s Architecture
G42 operates through three primary entities that have distinct but complementary roles.
Core42 is G42’s cloud and AI infrastructure subsidiary, responsible for the GPU clusters, data centers, and cloud platform that underpin G42’s AI services. Core42 has built one of the largest GPU deployments in the Middle East — NVIDIA H100 clusters at Abu Dhabi facilities — and is expanding through the same BIS Tier-2 framework that Saudi Arabia uses for its NVIDIA deployments. Core42’s infrastructure serves both G42’s internal model training and inference requirements and external customers through an AI cloud platform.
MBZUAI is the Mohamed bin Zayed University of AI — a graduate research university in Abu Dhabi that functions somewhat like a UAE counterpart to KAUST’s AI Initiative. MBZUAI has recruited international AI researchers, publishes openly, and has produced the Jais model as its flagship applied output. The university’s research quality is generally regarded as strong; it has attracted faculty from top US and European programs and has a publication record competitive with mid-tier US research universities.
The KAUST-MBZUAI comparison is instructive. Both are Gulf research universities with strong AI programs, strong endowments, international faculty, and Arabic NLP as a priority. KAUST has fifteen years of history and a broader STEM mandate; MBZUAI is AI-focused from inception and has the advantage of being newer — it was designed knowing what AI research requires in ways that KAUST’s original architecture, designed in 2007, was not. The Arabic LLM competition between Jais and Allam is, at its research layer, substantially a competition between these two institutions.
Inception is G42’s AI product and application subsidiary, responsible for taking the models that MBZUAI develops and the infrastructure that Core42 provides and building commercial AI applications. Inception’s product portfolio includes AI applications for healthcare, government services, and enterprise automation — the same domains that Humain is targeting in Saudi Arabia. The UAE’s more open economy and international business orientation gives Inception a different commercial context than Saudi AI applications, but the underlying technology stack is directly comparable.
UAE vs. Saudi AI: Measurement and Trajectory
The “who leads” question in Gulf AI is asked frequently and answered inconsistently, because the two countries are leading in different dimensions that matter differently to different observers.
Saudi Arabia leads on capital commitment. The $77 billion Humain commitment, the $5.3 billion AWS deal, the $10 billion Google Cloud deal — Saudi Arabia’s AI investment announcements from May 2025 alone dwarf anything the UAE has committed at a comparable moment. Saudi Arabia’s sovereign wealth fund capacity, oil revenue base, and Crown Prince’s personal investment in AI as a national priority create a capital mobilization capability that the UAE, with its smaller economy and more diversified investment portfolio, does not match.
The UAE leads on deployed AI. G42’s products are in operation — Core42 is serving real customers, Jais is being used by Arabic-language developers globally, Inception’s healthcare AI is deployed in UAE hospitals. The UAE’s more open economy, more diverse expatriate workforce, and longer history of international technology partnerships mean that AI applications have been deployed at scale, tested against real user needs, and iterated through commercial feedback loops for longer than Saudi Arabia’s equivalent programs.
The measurement challenge is that capital commitment and deployed AI are not the same thing, and the relationship between them is not proportional. Saudi Arabia’s $77 billion in announced commitments will not all convert to deployed AI systems on any short timeline; announced deals represent intentions, not operations. G42’s deployed systems, while smaller in total capital investment, represent real operational AI that is generating data about what works and what does not.
The trajectory question is more interesting: which approach generates more sustainable AI capability over the next decade? Saudi Arabia’s approach — massive capital investment, anchor partnerships with every major US AI company, aggressive talent recruitment — accelerates deployment if execution matches commitment. UAE’s approach — building from deployed operational experience, maintaining international partnerships, developing research credibility — generates capability more slowly but with more validated product-market fit.
G42’s Chinese Technology Divestiture: A Template
The forced divestiture of G42’s Chinese technology investments — Huawei equipment removal, ByteDance stake sale — is worth extended analysis because it established a template that other Gulf AI players are navigating.
The core dynamic: US export control policy for advanced AI chips is conditioned, de facto, on the technology alignment of the recipient. A company that is deploying Huawei telecommunications equipment in its data centers — equipment that the US government has designated as a national security threat — faces difficulties accessing NVIDIA GPUs under the BIS framework, because the combination of advanced AI chips and potentially compromised network infrastructure creates an unacceptable intelligence risk from the US government’s perspective.
G42’s response was to choose: US alignment and NVIDIA access, or Chinese technology partnerships and exclusion from frontier AI. The choice was made decisively. The execution was uncomfortable — Huawei equipment removal is expensive, and the ByteDance stake was likely sold at a discount given the circumstances — but the strategic outcome has been positive. G42 post-divestiture has access to NVIDIA hardware, Microsoft partnership, and the US government relationships that enable continued expansion.
Saudi Arabia has not faced the same explicit pressure because Saudi Arabia has not made comparable Chinese AI technology investments. Saudi Arabia has engaged with Chinese companies — Huawei has telecom infrastructure in Saudi Arabia, Chinese EV companies are present — but the AI infrastructure layer has been built on US technology from the beginning. Humain’s partnership architecture — OpenAI, Google, AWS, xAI, AMD, Qualcomm — is entirely US-aligned. This is partly strategic choice and partly path dependency: Saudi Arabia’s Vision 2030 was designed in close partnership with US advisors and consulting firms who shaped the technology architecture.
Who Wins the Arab AI Capital Competition?
The competition between G42 and Humain for the title of “AI capital of the Arab world” will be decided by multiple metrics over the next decade, and different metrics point to different winners.
If the metric is deployed AI applications serving real users at scale, the UAE currently leads and the gap will not close quickly. G42’s three-year head start in commercial AI deployment — Core42, Jais, Inception’s products — represents operational experience that capital alone cannot replicate. Saudi Arabia’s AI applications, however well-funded, will spend the next two to three years in development and early deployment before they generate the operational data that G42’s products already have.
If the metric is AI infrastructure scale — compute capacity, data center square footage, GPU count — Saudi Arabia will overtake the UAE within two to three years given the capital differential. The DataVolt facility at NEOM, SDAIA’s Hexagon data center, Humain’s infrastructure commitments, and the hyperscaler deployments will, when operational, represent more total AI compute than the UAE currently has.
If the metric is international AI partnership depth — relationships with the US companies building frontier AI, research publications cited by the global AI community, Arabic-language AI model quality — the competition is closer than the capital differential suggests. MBZUAI’s research quality is competitive with KAUST. Jais’s open release has generated more international developer adoption than Allam’s more restricted deployment. G42’s Microsoft relationship is deeper and more operationally integrated than any equivalent Saudi-US partnership.
The forward competition trajectory suggests convergence rather than clear victory. Saudi Arabia’s capital will bring its infrastructure and talent up to parity with UAE’s current position within three to five years. The UAE’s operational experience advantage will be partially offset by Saudi Arabia’s scale advantages as Saudi AI systems accumulate their own operational data. The Arabic LLM competition will continue for years, with the winner determined by application embedding rather than benchmark performance.
For the international AI companies and investors who are choosing between UAE and Saudi AI partnerships, the practical implication is that both are worth maintaining. The Gulf AI market is large enough for two major players; the UAE and Saudi Arabia are not exclusive partners for any major technology company. Microsoft-G42, Microsoft-Saudi partnerships; Google-UAE infrastructure, Google-Saudi $10 billion commitment; AWS-UAE deployments, AWS-Saudi $5.3 billion deal — the major US AI companies have made exactly this calculation already. The smart money is on both, and the competition between them will drive both to higher capability than either would achieve without the rivalry.