Khalifa University, GSMA, AMD open RF AI model to telecom operators

Key Points

1. Khalifa University, GSMA and AMD will open the RFGPT model to telecom operators for evaluation.
2. The model, built by the university's Digital Future Institute, pairs wireless signal analysis with natural-language interaction.
3. Operators get a hands-on test of AI for radio-frequency analysis before committing to network deployment.

The latest:

Telecom operators will be able to put an Emirati-built radio-frequency AI model through its paces at MWC26 in Doha, under an initiative launched by Khalifa University for Science and Technology, the GSMA and AMD. The program lets carriers explore RFGPT’s capabilities and judge whether it fits their own radio-frequency analysis needs.

Details:

• The model: RFGPT was developed by the Digital Future Institute at Khalifa University as a foundation model for radio frequencies. It combines advanced understanding of wireless signals with natural-language interaction, meaning operators can query signal environments conversationally rather than through conventional analysis tools alone.
• The partners: Three parties are behind the initiative: Khalifa University for Science and Technology as developer, the GSMA, the global mobile network industry association, and chipmaker AMD. The announcement did not break down each partner’s individual contribution to the program.
• The venue: The model is being made available to operators during MWC26 in Doha, the mobile industry’s global gathering. No separate timeline was announced for availability beyond the event, nor was a list of participating operators disclosed.
• The university’s framing: Professor Marwan Debbah, senior executive director of the Digital Future Institute, said radio-frequency foundation models offer a new approach to understanding the wireless environment by merging signal comprehension with natural-language reasoning. He said the aim is to let operators test the approach in practice.
• Shaping development: Debbah said operator feedback is meant to help steer how the model evolves, so that it addresses the specific challenges carriers face inside their own networks. That positions the Doha program as a development input as much as a product showcase.
• The GSMA view: Luis Powell, director of artificial intelligence technologies at the association, said operators need opportunities to test emerging AI capabilities against clearly defined use cases, rather than evaluating them in the abstract.
• The AMD view: Philippe Guido, executive vice president and chief commercial officer at AMD, said that as wireless networks grow more complex, operators need new methods to understand radio-frequency environments and make informed decisions based on them.
• The gap addressed: The three statements converge on a single problem: rising network complexity and the absence of practical testing grounds for new AI tools. The initiative is presented as an evaluation track, not a commercial rollout, with no pricing or licensing terms announced.

Background:

Radio-frequency analysis underpins how mobile operators manage spectrum, interference and network performance. Foundation models — large AI systems trained for broad tasks — have mostly targeted text and images; applying them to wireless signals is a newer direction for the telecom sector.

Between the lines:

The initiative is framed around evaluation rather than deployment. Debbah’s emphasis on operator feedback steering development, and Powell’s point about testing against defined use cases, both suggest the model is not yet positioned as a finished commercial product. AMD’s presence alongside a university institute also ties the effort to the compute hardware layer, not just the software.

What’s next

Watch MWC26 in Doha for which operators take up the evaluation, any results they report from testing RFGPT on live radio-frequency environments, and whether the partners announce availability or commercial terms beyond the event.