As telcos prepare to support AI workloads at hyperscale, their role is shifting from being a connectivity supplier to becoming integral to a converged AI infrastructure that supports critical workloads from the edge to the datacentre to the cloud. Pallavi Mahajan, the chief technology and AI officer at Nokia, tells George Malim, the managing editor of Agile Telco, how she sees telcos enabling and thriving from the AI supercycle as they build on their trusted networking capabilities to play their part in the complex models and opportunities that the various iterations of AI demand.
Mahajan has shaped industry-defining platforms and innovative business models across AI, infrastructure and networking. Her experience spans the full compute spectrum – from accelerated computing across the datacentre, cloud and edge to leading the first exascale supercomputer deployment, Frontier, and architecting software-defined networks. Today, as chief technology and AI officer at Nokia, she is driving the shift to AI-native, distributed systems where compute, networks and intelligence converge. Her remit spans technology, AI and platform strategy, steering Nokia’s roadmap and scaling its role in next-generation infrastructure.
Before Nokia, Mahajan held senior leadership roles at Intel, Hewlett Packard Enterprise and Juniper Networks. In each role, she led at the frontier across AI datacentres, high performance computing and autonomous networks. Her work has consistently focused on platformisation at scale – building systems designed for performance, adaptability and impact. The result: a track record defined by turning complexity into platforms that power the future of AI infrastructure.
George Malim: The AI supercycle has largely been framed as a compute story composed of GPUs, frontier models and hyperscaler capex. From your vantage point leading AI and technology at Nokia, what is missing from that framing?
Pallavi Mahajan: The compute-only framing isn’t wrong, it’s just incomplete. In frontier AI, the unit of compute has evolved from chips to entire AI factories. The real breakthrough for the AI supercycle isn’t bigger silicon, it’s the network, and that is what turns many machines into systems.
George Malim