Abu Dhabi’s artificial intelligence company G42 is developing national infrastructure for autonomous AI systems, alongside plans to build a large-scale computing centre in the UAE.
The centre will be capable of producing up to 100 trillion tokens – which are smaller units of data – per day.
“We call this [the computing hub] the ‘agent factory’,” G42 chief executive Peng Xiao said at a panel at the World Governments Summit in Dubai on Tuesday. He described the initiative as a shift from simply generating AI outputs to governing how AI systems operate independently.
The factory is part of what the advanced technology company has been developing over the past three years. G42 has been building what it calls the “intelligence grid”, a framework that defines how large-scale AI infrastructure is built, interconnected, and governed at both national and global levels, according to the company.
At the foundation of this grid is compute infrastructure that increasingly functions less like traditional data centres and more like “AI token factories”. Output is measured in units of machine intelligence, or tokens, rather than raw processing power.
The agent factory represents a parallel layer to token production, focused on automating the creation and posting of AI agents across sectors such as health care, education, legal services and energy.
G42 says these two layers, token production and agent posting, are tightly linked and together shape how AI capabilities are industrialised at scale.
Mr Xiao gave one example of finding solutions for the Department of Health Abu Dhabi.
“We are deploying as we speak right now, AI for primary care so that doctors can use AI agents now to provide higher quality … and also lower cost access to healthcare,” he said.
Mr Xiao said this will underpin the UAE’s AI capabilities at scale.
“AI can become a lot more autonomous, proactive in supporting your workload, in exploring new areas of discovery,” he said. The idea is to create a framework to oversee AI agents that can act autonomously on behalf of users, even while they are sleeping.
He said the project is being developed in collaboration with Microsoft, and complements major AI infrastructure initiatives such as the 5-gigawatt Stargate AI campus in Abu Dhabi, which is being positioned to be the largest data centre in the world.
G42’s chief executive emphasised that while AI agents can work around the clock, safeguards must ensure their actions align with national standards and ethical requirements.
chief executive of G42
He added that infrastructure alone does not determine outcomes. How AI systems are interconnected, technically, commercially and diplomatically across borders, will strongly influence whether AI adoption delivers broad-based benefits or leads to increased global fragmentation.
Mr Xiao said this evolution requires stronger governance and stressed that AI should be the main priority for all governments that want to see future growth.
“Nation states have to set AI as their number one, not number two, priority to move the needle on the GDP,” he said.
Many countries do not have the natural resources, such as power, to build and implement native compute. But they can take support from G42’s digital embassy initiative.
The digital embassy concept was unveiled last month to create trusted, governed frameworks that allow countries to access advanced AI capabilities.
Countries will be able to retain full sovereignty over their data, policies and regulatory oversight, the company said at the time.
“We are ready to share our resources through this digital embassy initiative, to bring AI technology readiness to nations that may not yet have the infrastructure to deploy it independently,” Mr Xiao said on Tuesday. The initiative is being launched in partnership with Microsoft.
In March, the Abu Dhabi government also signed an agreement with Microsoft and Core42 to enhance the efficiency of government services in the digital sphere.
In November, Microsoft pledged to invest about $8 billion in the UAE between this year and 2029 to support AI and cloud infrastructure development.
The tech company made the pledge after securing licences to export advanced Nvidia A100, H100 and H200 GPUs to the UAE, a milestone for the country’s AI ambitions. The investment follows a $7.3 billion commitment made between 2023 and the end of last year.
