AIrev and Qualcomm tie agent sovereignty to where agents run
Listen to this article
Read by Anchor
The question raised by this announcement is not which model is most capable, but where the agent actually runs. Abu Dhabi based AI company AIrev announced a strategic agreement with Qualcomm Technologies to run its OnDemand agent platform on Dragonwing hardware. The first step begins with integrating the platform with the Dragonwing AI On-Prem Appliance, a standalone box that keeps the model and inference workload inside the client's own environment.
The agent goes to the dataThis is the practical framing presented by the announcement: rather than transferring data to a mandatory cloud service, the agent definition itself can run in a sovereign data centre, on an enterprise server, or at the network edge. AIrev defines its platform as a low-code or no-code operating system for building, deploying, and managing autonomous agents. In this context, the stated value is not a comparison of model capabilities, but the choice of execution location.
The source categorises Dragonwing under Qualcomm's products for industrial and enterprise use, rather than mobile phones. The announced appliance is designed to keep the model and inference, the process of handling AI inputs and producing outputs, inside the customer's facility. Wassim Chourbaji, senior vice president and president for Europe, Middle East and Africa at Qualcomm Technologies, said making OnDemand available on Dragonwing products broadens how enterprises and governments can deploy capable and efficient agents.
Reported figures require proper attributionAIrev says its products processed 6.8 trillion tokens in the second quarter of 2026, roughly six times the figure from a year earlier. It also claims that OnDemand has exceeded four million users, hosts more than 300 specialised agents, and supports over 50 languages. These figures originate from the company and are presented by the source as reported, meaning they are insufficient on their own to evaluate product quality or the actual adoption of the new appliance.
Yet they illustrate why the agreement highlights deployment location. An on-premises appliance offers an alternative to cloud hosting for where inference workloads reside. This represents the sovereign dimension of the announcement: an operational architecture that can be placed inside the client's environment, rather than generic claims about infrastructure ownership. Coverage notes that the company was founded in Abu Dhabi in 2024, while the agreement spans Dragonwing products, beginning specifically with the on-premises appliance.
What was omitted matters just as muchThe material contains no pricing, availability timeline, or confirmed client for the joint product, nor does it list the models the appliance will run. It would therefore be inaccurate to treat the agreement as an announcement of active deployment across any specific government entity or enterprise. The subsequent test is straightforward: will details regarding availability, models, and clients emerge following this announcement? Until they do, the confirmed reality remains an announced integration between an Abu Dhabi based agent platform and Qualcomm enterprise hardware, with a clear focus on where AI runs.