Agentic AI: From passive assistance to active execution. With Philippe Ensarguet at Orange

Physical AI has been a demo thing for the last few years. It is no longer. Robots that work in teams, drones that need to be told apart from birds, ports and mines that run on sensors: none of them can wait for a round trip to a data center in the middle of nowhere. AI is not leaving the data center, but it is taking an apartment closer to the edge. In this Sparring Partners podcast, I speak with Sameh Yamany, Chief Technology and AI Officer at VIAVI Solutions, about what edge AI needs to support physical AI, and what that means for operators. Here is what we covered:

Nobody wants data centers near where they live: they use too much power, they bring few jobs, and they are tied to an AI that many people do not understand and fear. Yet we will use AI, and we need data centers to run it. The question is not whether we build them, but where and who gets to decide. In this Sparring Partners podcast, I speak with Mike O’Rielly, Principal at MPORielly Consulting and FCC Commissioner from 2013 to 2020, after 20 years working on tech and telecom policy on Capitol Hill. Mike argues that hyperscalers are losing the PR war to political pressure. As usual, our conversation ranged widely:

For years, the assumption in the US – and not just in the US – was that broadband means fiber, and anything else is second class. In rural areas, that assumption is running into two hard limits: the cost per passing keeps rising as density falls, and Starlink and Amazon Leo now set the price and speed baseline in the hardest-to-reach locations. In Sparring Partners, I speak with Claude Aiken, Chief Strategy Officer and Chief Legal Officer at Nextlink Internet, a rural operator that runs fiber, fixed wireless and mobile across 12 states. Claude was previously President and CEO of WISPA and spent a decade at the FCC, so he has seen rural broadband from the operator, industry and regulator sides. We don't agree on everything, but we agree that headline speed is the wrong thing to chase. Here is what we covered:

Does Open RAN need further opening? Can open-source facilitate and accelerate the adoption of Open RAN? Yes and yes. The OCUDU project was established on these premises to address common challenges that Open RAN deployments face. In Sparring Partners, I speak with Arpit Joshipura, General Manager of Networking, Edge, and IoT at The Linux Foundation, about how we need to move beyond open interfaces (which remain an essential starting point) to open-source RAN platforms that OCUDU builds with open-source software. You may already know this, but Arpit went further and explained what specifically OCUDU does that was missing:

Private networks sound such a good idea. There is a clear need for venues and enterprises to have better on-premises connectivity to support their operations and services. They can augment much-needed indoor connectivity, which today is mostly provided by Wi-Fi. Yet, deployments are only slowly taking off, and that’s the case across geographies. Do we have the technology? Yes. Are they too expensive to deploy? No. So what’s holding us back? Mostly the business models, the (real or perceived) complexity for enterprises to deploy them, and regulatory frameworks geared toward cellular networks owned and run by mobile operators. In this Sparring Partners, I spoke with Maria Lema, Founder and CEO of Weaver Labs, who addresses these issues from the assumption that a private network should be viewed as a strategic investment for the venue or enterprise. Once you accept that, many other pieces fall into place. Listen to the conversation, as I do not want to spoil it, but these are some topics we covered: