Markets / Industries / Telecom
Industry view · Telecom
Telecom is a $1.85 trillion revenue industry spending a roughly flat couple hundred billion dollars a year on capital, now redirecting that money from concrete-and-steel radio gear toward software that automates the network. The AI-in-telecom software layer is still small — on the order of a few billion dollars a year — but it is the only line item growing at a brisk double-digit clip while the core business is flat.
01 · The thesis
Telecom's predicament is structural: revenue is enormous but barely grows, and a decade of 5G build-out left carriers with heavy capex and thin returns. IDC and Dell'Oro both describe global telecom capex as flat-to-declining into 2026, with capex intensity sliding gradually over the decade. The money that remains is being re-pointed — away from radio hardware, toward AI platforms, automation software, and data-center connectivity. Dell'Oro calls the capex picture 'normalize, not disappear'; the mix is what is changing.
The AI story here is therefore less about new revenue than about defending margin. Verizon executed over 70 million network configuration changes autonomously in 2025; AT&T reports AI-driven technician routing saves it millions a year in fuel; Vodafone's SuperTOBi resolves a large share of service requests in Germany without a human. The second, larger prize is 'telco-for-AI': selling fiber and capacity to the AI build-out, which McKinsey frames as a meaningful new growth avenue for operators. Carriers want to be the network *for* AI, not just a network *using* it.
Nvidia is embedding GPUs into radio access, turning the cell site into an AI compute node and threatening the merchant-silicon status quo.
Autonomous fault detection, configuration, and energy management compress operations headcount and downtime across the core.
Intent-driven assistants deflect the majority of service contacts, the most direct and immediate cost takeout in telecom.
Long-haul and dark fiber to AI data centers is a new growth line that plays to incumbents' existing assets.
Back-office stacks are slowest to move; legacy integration and regulatory load slow agentic automation of provisioning and billing.
02 · Public players & exposure
We plot the listed players on two editorial axes — how exposed each is to AI disruption, against how ready its data, brand and position are to be the answer. The figures in the table are sourced; the placement is our read.
| Company | Stance | The sourced fact |
|---|---|---|
| VerizonVZ | Automating hard | Executed over 70 million autonomous network configuration changes in 2025 and is targeting Level 4 network autonomy. |
| AT&TT | Cost takeout | Reports that AI-driven technician routing saves the company millions of dollars per year in fuel costs. |
| T-Mobile USTMUS | AI-native care | Built IntentCX with OpenAI and is also an AI-RAN 6G pilot partner with Nokia and Nvidia, with a stated goal of sharply cutting human service contacts. |
| NvidiaNVDA | Arms dealer | Investing $1bn in Nokia at $6.01/share to push GPU-based AI-RAN into 5G-Advanced and 6G networks. |
| NokiaNOK | Betting the RAN | Took a $1bn Nvidia investment to add commercial AI-RAN products and stabilize eroding RAN market share. |
| EricssonERIC | Roadmap pressure | Pitching AI-native autonomous-network roadmaps but outside the headline Nvidia-Nokia AI-RAN pact, raising platform-relevance risk. |
| VodafoneVOD | Margin defense | Invested £120m in SuperTOBi; in Portugal first-contact resolution rose from 15% to 60% after deployment. |
| BT GroupBT.A | Shrinking via AI | Cutting up to 55,000 jobs by 2030 with ~10,000 expected to be replaced by AI. |
| ServiceNowNOW | Picks & shovels | Launched telecom-specific AI agents to autonomously run customer-service and network-ops workflows for carriers. |
| ItentialITNL-PVT | Automation pure-play | Network-automation software vendor that has raised $25.5m total, including a $20m Series B led by Elsewhere Partners. |
03 · The two clocks
Three clocks are running at different speeds: customer care is being automated now, the network core over this decade, and the silicon layer is being rebuilt around AI compute.
The fastest clock is customer service. Vodafone's SuperTOBi resolves a large share of service requests without a human, and in Portugal first-contact resolution jumped from 15% to 60%. This is the deflation that telecom CFOs can book this year.
The middle clock is the network itself. Bain and the TM Forum report that a meaningful minority of operators have reached higher levels of network autonomy in select domains, with more expecting to within a couple of years, and network automation is now a large share of operator AI pilots. Self-driving networks are real but uneven.
The slowest and heaviest clock is silicon and RAN. Nvidia's $1bn bet on Nokia and the AI-RAN push reframe the base station as an AI computer — a multi-year, capital-intensive transition that the AI-RAN market is only beginning to fund.
04 · Private flagships
The companies attacking this industry AI-first, with disclosed funding where available:
Nvidia's investment plus its Arc Aerial RAN Computer aims to make GPU-based, software-defined AI-RAN the basis of 5G-Advanced and 6G, with T-Mobile and Dell as launch partners.
An intent-driven decisioning platform built with OpenAI that reads live network and service data to resolve issues, targeting a steep reduction in human service contacts.
Sells resilient long-haul fiber paths between AI data centers, including an AWS agreement, monetizing the carrier's physical assets for the AI build-out.
A generative-AI upgrade to Vodafone's TOBi assistant, deployed across Europe on Microsoft Azure to deflect and resolve service contacts.
Pre-built AI agents that autonomously run labor-intensive customer-service and network-operations workflows for communications service providers.
A multi-domain network-automation and orchestration platform sold to Fortune 500 telecom and enterprise customers, positioning for the agentic-infrastructure era.
05 · Signals
A major U.S. carrier commits to an AI-native customer-experience platform, signaling care automation moving from chatbot to decisioning engine.
Quantifies that self-driving network operations are running at production scale, not pilot scale, at a tier-1 operator.
Reframes the radio access network as AI compute and pulls a struggling RAN vendor into the GPU ecosystem.
Industry surveys point to a large majority of telcos increasing AI budgets — AI moves from experiment to line item.
Confirms the squeeze: total capital is flat-to-down while its composition shifts toward AI and automation.
06 · The exposure read
AI rewards clean, structured advantage and punishes friction. The line runs through who owns the data, the brand and the customer — and who is merely a step the technology can route around.
Sources
TelecomLead — Global telecom revenue $1.85T in 2025 · Fortune Business Insights — AI in Telecommunication market · IDC — Telcos at the crossroads of AI (capex forecast) · Dell'Oro Group — Global telecom capex to normalize, not disappear · Verizon — Architecting network autonomy (70m autonomous changes, Level 4) · Klover.ai — AT&T AI strategy (fuel savings from technician routing) · T-Mobile Newsroom — IntentCX with OpenAI · Nvidia Investor Relations — Nvidia and Nokia AI platform for 6G ($1bn at $6.01/share) · TelecomTV — Nvidia $1bn investment in Nokia AI-RAN · Total Telecom — Vodafone invests £120m in SuperTOBi · Forbes — BT to cut 55,000 jobs by 2030, ~10k via AI · McKinsey — AI infrastructure a new growth avenue for telco operators · Bain & Company — Accelerating autonomous networks (Level 4/5 maturity) · Itential — $20m Series B funding announcement