Nvidia’s high-performance hardware has long served as the invisible engine behind the United Arab Emirates’ rapid emergence as a global artificial intelligence powerhouse. Now, the American chip giant is preparing to deepen its footprint in the Gulf nation by offering something far beyond silicon: free software designed to rival the world’s most advanced AI systems, including China’s leading models.
This strategic expansion follows a series of aggressive financial maneuvers by Nvidia. Last month, the company invested $7 billion to acquire a stake in Poolside, a U.S.-based coding startup, while concurrently licensing the specialized tools the firm utilizes to build sophisticated AI models. According to Marc Einstein, an analyst at Counterpoint Research, this upcoming open-weight model is scheduled for release by the end of the year, allowing anyone to download it free of charge. Crucially, the software is specifically optimized for buyers who are already running Nvidia hardware. Few nations fit that description quite as snugly as the United Arab Emirates, where a flagship domestic data center is slated to run on an astonishing 400,000 Nvidia chips.
For years, the UAE carefully positioned itself as neutral ground in the intensifying technological rivalry between the United States and China. The nation routinely acquired hardware from American suppliers while simultaneously utilizing telecommunications and technological equipment from Chinese firms. In an effort to avoid becoming overly dependent on imported software, the UAE even engineered its own indigenous free model, known as Falcon.
However, that delicate neutrality has steadily eroded. The shift became unmistakable when G42, the state-backed artificial intelligence heavyweight in Abu Dhabi, quietly cut its ties with Chinese technology entities to secure access to restricted Nvidia chips under a carefully negotiated pact with Washington. That alignment is poised to narrow even further if the UAE ultimately chooses to adopt Nvidia’s upcoming software ecosystem.
The introduction of Nvidia’s model fills the final missing piece of the puzzle, according to Mohammed Soliman, a senior fellow at the Middle East Institute, a Washington-based foreign-policy think tank. Soliman noted that Nvidia has already captured the market for the underlying hardware. By scaling the factory that produces the models running on those chips and releasing them as open weights, the company is solidifying an end-to-end technological pipeline.
Build or Rent?
As the nation accelerates its technological transformation, the two primary institutions driving the UAE’s artificial intelligence push maintain starkly different philosophies regarding how to acquire advanced systems. The core debate centers on whether the country should build sovereign AI models entirely from scratch or strategically leverage existing technologies developed by external partners.
The Technology Innovation Institute (TII), a prominent government-backed research lab responsible for the creation of the Falcon model, firmly insists on engineering its models from the ground up. In contrast, G42, which manages the foundational infrastructure supporting the UAE’s broader AI initiatives, has thrown its weight behind Nvidia’s push toward accessible, free-to-use models.
Hakim Hacid, chief researcher at the Technology Innovation Institute, articulated the lab’s distinct perspective during a recent interview. He emphasized that as a dedicated model-building laboratory, TII believes maintaining diversity in artificial intelligence perspectives and preserving absolute technological sovereignty requires a robust, homegrown foundation. Hacid argued that the true value of these endeavors is not limited to the commercial models ultimately produced, but rather lies in mastering the underlying technology and retaining total control across the entire AI stack.
Meanwhile, G42 has embraced Nvidia’s collaborative initiatives. While the company did not respond to inquiries regarding whether it plans to utilize Nvidia’s upcoming model, it has acted as an early and enthusiastic supporter of the chip maker’s broader strategy to distribute powerful models widely. In July, Nvidia established the Open Secure AI Alliance, built on the premise that modern nation-states require AI models they can independently inspect and control. G42 immediately joined the coalition, establishing itself as the sole corporate participant from the Gulf region.
Peng Xiao, chief executive officer of G42, underscored this philosophy in a public statement posted on social media platform X at the time of the alliance’s launch. Xiao wrote that frontier open models represent an essential component of the future, functioning alongside frontier closed models, with security, trust, and sovereign control embedded from the start rather than integrated as an afterthought.
According to Soliman, however, adopting an external model does not automatically equate to surrendering national control. He argued that modern AI sovereignty is a matter of degree rather than an all-or-nothing proposition. True sovereignty, Soliman explained, was never about inventing the foundational architecture from scratch; rather, it centers on the model’s weights, the meticulous fine-tuning of domain-specific data, and, most critically at this stage of the industry’s evolution, the domestic capability to serve these sophisticated models efficiently to users within the UAE and other international markets.
This evolving landscape is heavily shaped by shifting geopolitical alignments. The United States government approved export licenses for Nvidia’s most advanced chips to be shipped to the UAE only after G42 divested its financial stakes in Chinese technology firms two years ago. Building upon that diplomatic thaw, Washington officially placed the UAE in July into an exclusive group of nations receiving enhanced favorable treatment under export administration rules, effectively designating the Gulf state as a premier technology ally. Nevertheless, the software powering those high-end chips is largely licensed from U.S. corporations, which retain the authority to dictate permissible use cases and authorized users.
The Falcon project was originally conceived to challenge that paradigm. Developed by TII in 2023, the model was designed as a homegrown, free alternative to proprietary systems originating from the United States and China. Its success was underscored earlier this year when it captured the top spot on Counterpoint Research’s index tracking state-backed foundational models.
Regional dynamics add another layer of complexity. Saudi Arabia, locked in a fierce economic and technological rivalry with the UAE to establish itself as the premier artificial intelligence hub of the Middle East, has pursued an entirely different strategic pathway. Humain, the kingdom’s state-backed artificial intelligence entity, has secured clearance to purchase tens of thousands of Nvidia chips. Earlier this month, Humain introduced a specialized Arabic-language model commissioned from a Chinese laboratory. This release followed another rollout a year prior featuring a model that state officials asserted was built entirely from scratch within the kingdom.
Sam Winter-Levy, a senior fellow at the Carnegie Endowment for International Peace, noted that these regional developments illustrate an absence of rigid full-stack lock-in. He observed that Saudi Arabia is successfully employing Chinese AI models to construct its own specialized systems while simultaneously relying on U.S. hardware to develop and execute them. Winter-Levy added that these states possess powerful economic and strategic incentives to hedge their bets and will likely continue doing so for as long as geopolitical circumstances permit.
Nvidia’s Nemotron
Nvidia has powerful commercial motivations for encouraging sovereign states to adopt its upcoming software release, widely anticipated to be designated Nemotron 4. Earlier iterations of American artificial intelligence models have frequently lagged behind China’s prolific open-source ecosystem. According to an August report published by Hugging Face—a prominent collaborative platform where developers share and download AI models—Chinese open models were downloaded approximately two billion times this year, vastly outperforming any Western counterpart.
To counter this trend and capture greater market share, Nvidia has agreed to a staggering investment of nearly $13 billion to acquire Hugging Face. Because governments and enterprises attempting to build sovereign models almost universally begin their work using free foundational models created by Meta or Alibaba, Nvidia’s extensive capital allocation is strategically designed to ensure that Nemotron becomes the default starting point for developers, according to Einstein.
The multi-billion-dollar investment in Poolside similarly appears less focused on displacing established model families and more directed toward driving long-term adoption of Nemotron within an established ecosystem where Nvidia already commands an unassailable hardware position.
Despite the influx of foreign technology and partnerships, leadership at the Technology Innovation Institute maintains that the UAE can successfully pursue both paths simultaneously. While the foundational research and core development remain proudly Emirati, incorporating the best available global advancements into domestic workflows is entirely feasible, according to Hacid.
As Hacid concluded, true sovereignty does not mean isolation. The nation can continue building its foundational models at home while remaining completely open to the best innovations the world has to offer.
