Ericsson CTO on AI RAN, AI-Native 6G & India’s 5G Edge | India AI Impact Summit

Great to be here. Thanks.

We have been working with AI in the networks for quite some time, but I think you're right there is more focus on it right now in terms of our communication, and it comes from the fact that networks are so critical now when it comes to providing a better AI experience for the users. Be it your generative AI on your smartphone, maybe your AI-based glasses, the agentic AI, and ultimately physical AI. So the networks and how networks are handling AI traffic becomes critically important. So that's a little bit what we talked about recently. Then of course a lot of things are happening in the networks as well. We talk about operators, the leading operators, moving towards autonomous networks. Autonomous networks level four, almost full autonomy. The human is still in the loop making the final decision, but AI is really helping to form the decisions or forming the base for the decisions.

So I think since AI is such a profound technology impact on the networks at all levels, hardware of course, also on the model side, in orchestration, all the way through the stack, there is kind of a full AI strategy since some time in terms of which hardware we can run on. And we have taken the decision actually quite some time ago to have a full hardware-software disaggregation or separation, which means that we can run on different hardware. We can run on CPUs, we can run on GPUs, we run on our own hardware, and that is not a new thing. We talked about that before. But of course now we're also being even clearer about the AI capability that the networks already have. Ericsson's 5G products are AI-powered already today. And we did launch or make visible some of those capabilities just the last couple of days, which are creating better user experience but also saving costs for the operators. It's energy efficiency, spectral efficiency, really using the network in the best possible way. So AI in the network has happened, continues to happen, and will be the path going forward.

So we have AI models for the networks, absolutely. You should think of it as a hierarchical system because the models are much smaller out in the radios and radio access networks and the core networks. The higher up they are, of course, the bigger the models are, the more information they need to draw on. But this hierarchical way of building AI-based systems, agentic AI systems, that is really the way to get the best performance, but also to guarantee that AI stays within the guardrails, that you have trust, security, that you can rely on the AI that you put to work. And I think that comes with the architecture. It comes with how we implement AI on all layers.

I think so. I think the investments that were sort of wisely implemented, fast roll-out of 5G across a nation. I think the number is staggering, some 99.6% of all the districts have 5G in India. So it's a fantastic starting point. If you combine that with what's already in place, the DPI and the Aadhaar and the UPI, and then basically the investments that are happening now on the data center side, if you combine those data centers with a better 5G network, I think it is a perfect digital platform to scale AI use cases. Think about it, your industrial AI, which is starting to happen now in manufacturing, perhaps hospitals, education, farming, agriculture. I think these are all sectors that will now benefit from that stronger 5G infrastructure.