NVIDIA's newly published State of AI in Telecommunications: 2026 Trends survey puts a number on something that's been obvious anecdotally for a while: AI in telecom has stopped being a pilot project and become standard operating practice. Sixty-six percent of surveyed telecom organizations now report active AI use, up from 49% just two years ago in 2024. That's not incremental growth — it's a genuine shift in how the industry runs its operations.
The numbers behind the shift
A few figures from the survey stand out. Eighty-nine percent of telecom respondents plan to increase AI spending in 2026, up sharply from 65% who said the same a year earlier. Generative AI specifically is now in use or under active evaluation at 60% of surveyed organizations, again up from 49% in 2024. And when asked where AI is actually delivering measurable return on investment, the top answers were autonomous network operations (cited by half of respondents), customer service improvements (41%), and internal process optimization (33%).
Put together, that's an industry moving past the "should we try AI" question and into "which operational process do we automate next."
The gap between hyperscale telcos and everyone else
Most of the headline-grabbing AI deployments in telecom — AI-native network cores, GPU-heavy backbone buildouts, agentic systems managing radio access networks — belong to carriers with budgets and infrastructure that a regional or small-market ISP simply doesn't have. The survey found 77% of large operators expect AI-native networks to arrive before full 6G rollout, and 65% already report AI-driven network automation in production. That's a different scale of investment than what most PPPoE and hotspot ISPs are working with.
But the underlying pattern — replacing manual, reactive operations with automated, real-time ones — is exactly what's already happening at the smaller end of the market too, just through different tools. It doesn't take a GPU cluster to stop a technician from manually toggling a router when a client's payment is overdue, or to catch a billing anomaly before it becomes a revenue leak. As Sebastian Barros of Circles put it, telcos are shifting focus "toward moving intelligence across local and regulated infrastructure" — and that logic scales down just as much as it scales up.
What this actually means if you run an ISP
You don't need to become an AI company to benefit from this shift. What the survey really shows is which operational processes are worth automating first, because the biggest telcos have already tested them at scale:
- Autonomous enforcement. The top ROI use case in the survey was autonomous network operations — for a smaller ISP, the direct equivalent is real-time RADIUS control that disconnects and reconnects clients automatically based on payment status, with no operator logging into a router.
- Process optimization over headcount growth. A third of operators cited internal process optimization as a proven AI win. For an ISP scaling past a few hundred subscribers, that means billing and reconciliation that runs itself instead of demanding a bigger back-office team.
- Customer-facing automation. Customer service was the second-highest ROI use case industry-wide — self-service payment and account tools reduce support load the same way, just without a call center behind them.
This is where platforms like XpressRADIUS fit the trend rather than chase it: real-time RADIUS enforcement, automated disconnect and reconnect, and direct M-Pesa and Kopo Kopo integration are the small-ISP version of the same automation the biggest carriers are now spending 89% of their AI budgets to build.
See how automated billing and RADIUS control work for your ISP.
Start a free trial