Nvidia’s chips are driving the United Arab Emirates’ rise as an artificial-intelligence power. Next on offer is free software, a model built to rival China’s best.
Last month, Nvidia spent $7 billion to buy a stake in U.S. coding startup Poolside and license the tools it uses to build AI models. Poolside’s new model is due by the end of the year for anyone to download for free, and it is aimed at buyers already running Nvidia hardware, Marc Einstein, an analyst at Counterpoint Research, told Rest of World. Few countries fit that description better than the UAE, whose flagship data center will run on 400,000 Nvidia chips.
For years the UAE positioned itself as neutral ground between the U.S. and China, buying chips from the former and, until recently, equipment from the latter. It built its own free model, Falcon, to avoid relying on imported software. That neutrality has been fading since G42, the state-backed AI company, cut its Chinese ties to win Nvidia’s chips. It would narrow further if the country also opts for the new model. “The last missing piece was the model layer,” Mohammed Soliman, a senior fellow at the Middle East Institute, a foreign-policy think tank in Washington, told Rest of World. “Nvidia already owns the chips. Now it is scaling the factory that produces the models that run on those chips, and it is releasing them as open weights.”
Build or rent? #
The two institutions leading the UAE’s AI push each have their own view on whether to build sovereign AI models from scratch or lean on existing technology from other companies.
The Technology Innovation Institute, the government lab behind Falcon, insists on building its own models from the ground up, while G42, which runs the UAE’s AI infrastructure, has aligned itself with Nvidia’s push for free models.
“As a model-building lab, we believe that maintaining diversity in AI perspectives, and preserving full technological sovereignty, requires a strong, homegrown foundation,” Hakim Hacid, TII’s chief researcher, told Rest of World. “The value is not limited to the models we ultimately produce, but it lies in mastering the underlying technology and retaining control across the entire AI stack.”
G42 did not answer Rest of World’s questions about whether it would use Nvidia’s model. The company has been an early supporter of Nvidia’s push to give away powerful models. In July, Nvidia formed an alliance around the argument that countries need AI models they can inspect and control, and G42 joined immediately, becoming the only Gulf company to do so.
“Frontier open models are an essential part of that future, alongside frontier closed models, with security, trust and sovereign control designed in from the start rather than bolted on later,” G42’s CEO, Peng Xiao, wrote on X at the time.
Building on someone else’s model need not mean giving up control, Soliman said. Sovereignty in AI is a matter of degree rather than an all-or-nothing choice, he said.
“Sovereignty was never about inventing the foundational model from scratch,” Soliman said. “It is about the model’s weights, how you fine-tune the data, and, most importantly at this stage, your ability to serve these models to users in the UAE and other markets.”
The U.S. government approved sales of Nvidia’s most advanced chips to the UAE after G42 sold its stakes in Chinese technology companies two years ago. In July, it placed the UAE in the group of countries it treats as its closest technology allies. The software running on those chips is mostly rented from U.S. companies, which decide what the models may do and who may use them.
Falcon was meant to be the exception. TII built it in 2023 as a free alternative to models from the U.S. and China, and it ranked first this year on Counterpoint’s index of state-backed models.
Saudi Arabia, which is competing with the UAE to become the region’s AI hub, has taken a different route. Humain, the state AI company, is cleared to buy tens of thousands of Nvidia chips, and introduced an Arabic model this month that it had commissioned from a Chinese lab. A year earlier, it had launched one it said was built from scratch in the kingdom.
“It shows there’s little full-stack lock-in: The Saudis are using Chinese AI models to build their own models, while using U.S. chips to develop and run them,” Sam Winter-Levy, a senior fellow at the think tank Carnegie Endowment for International Peace, told Rest of World. “These states have strong incentives to hedge and will continue to do so for as long as they can.”
Nvidia’s Nemotron #
Nvidia has its own reasons for wanting countries to take the upcoming model, expected to be called Nemotron 4.
Earlier versions of U.S. AI models have lagged well behind China’s free models, which were downloaded roughly 2 billion times this year, far ahead of any Western rival, according to an August report by Hugging Face, the platform where developers share and download AI models.
To change that, Nvidia has also agreed to pay almost $13 billion to buy Hugging Face. Governments building their own models mostly start from a free one made by Meta or Alibaba, and Nvidia’s spending is meant to make Nemotron the base they pick instead, Einstein said.
The Poolside deal “appears less about displacing incumbent model families and more about increasing future adoption of Nemotron within an ecosystem where Nvidia already has a strong hardware position,” Einstein said.
TII says it can do both. While the foundation stays Emirati, anything better from elsewhere can still be put to work, Hacid said.
“Sovereignty does not mean isolation,” he said. “We build our foundation at home, while remaining open to the best of the world.”
Additional reporting by Divsha Bhat in Dubai.