Google Cloud and Nokia have struck a partnership to embed Gemini AI models into the software that runs telecom networks. The deal, announced this week, is built around automating network management — tasks like traffic routing, fault detection, and configuration that operators still largely handle by hand. The companies say the collaboration will push investment into the AI infrastructure needed to make those networks smarter and faster.
What the partnership involves
Nokia will integrate Google Cloud's Gemini AI into its existing network management platforms. That means telecom operators using Nokia gear can tap into generative AI to analyze network data, predict congestion, and even write configuration scripts. Google Cloud, in turn, gets a foothold deeper inside the telecom industry's operational backbone. Neither side disclosed financial terms or a deployment timeline, but they described the work as already underway.
Why telecoms need automated management
Running a modern telecom network involves juggling thousands of cell sites, fiber links, and core routers. Traffic patterns shift by the minute, and a single misconfiguration can cascade into a regional outage. Most operators still rely on engineers to monitor dashboards and manually adjust settings. That approach doesn't scale as 5G and edge computing pile on more complexity. Automating those decisions with AI, the companies argue, frees up engineers and cuts the time it takes to fix problems. The goal is fewer dropped calls and faster service upgrades.
Boosting AI infrastructure investment
The partnership also signals a bet that telecoms will spend more on the computing power needed to run AI models in real time. Network management requires inference at low latency — decisions that have to happen in milliseconds. That means carriers may need to buy more GPUs or cloud instances near their network edges. Google Cloud, which sells custom TPUs and Nvidia GPUs, stands to gain from that demand. Nokia gains a way to offer operators a richer software stack without building the AI models from scratch.
For now, the companies haven't named any carrier customers testing the integrated system. The real test will be whether operators, who are famously cautious about new software in their core networks, trust an AI to handle configuration changes that once required human approval.
