Telcos spent years being told it was becoming irrelevant. Hyperscalers were eating its lunch, OTT players were hollowing out its revenues, and its destiny was to be a glorified utility. That narrative is now dead – and what’s replaced it is an opening that most telco boardrooms haven’t fully computed yet.
Let’s be direct about where telecoms has been. A decade of digital transformation programs delivered incremental efficiency gains – not the structural reinvention the industry needed. ARPU flatlined while capex compounded. Consumers couldn’t tell operators apart. And the hyperscalers weren’t just circling – they were moving in.
Between 2020 and 2023, the threat became concrete and specific. Microsoft acquired Affirmed Networks and Metaswitch to build a carrier-grade 5G core, then took ownership of AT&T’s entire Network Cloud platform in 2021 – the first time a tier-one operator handed its network infrastructure software to a hyperscaler.
AWS launched its own Private 5G service in 2022, letting enterprises deploy cellular networks for as little as $14,000 without involving an operator at all; AT&T’s CEO was forced to address the threat directly at an investor event before the product had even formally launched.
Google placed its compute directly at operator network edges and positioned its cloud as the natural home for 5G network functions. By 2023, the question in telco boardrooms wasn’t whether hyperscalers would compete – it was whether there would be anything left to compete for.
So what changed? One thing: generative AI arrived and consumed every dollar and every strategic neuron the hyperscalers had.
“The hyperscalers didn’t pivot away from telco because they lost interest. They pivoted because something far larger came along – and in doing so, they handed telcos the clearest strategic opening in a generation.”
Act I | How the field got cleared
Understand what the hyperscalers were actually doing when they moved into telco territory. They weren’t trying to become connectivity businesses. They were hunting for the next platform layer – the point in the stack where they could control the enterprise relationship, extract the margin, and push everyone else toward commodity. Telco infrastructure was interesting to them as a distribution channel, not as a destination.
Then the AI arms race started in earnest. Microsoft committed hundreds of billions to OpenAI and Azure AI. Google restructured its entire organisation around Gemini. Amazon poured capital into Anthropic and AWS AI infrastructure. The five largest hyperscalers deployed an estimated $197 billion into AI-related infrastructure in 2024 alone – up more than 50 percent on the prior year, with J.P. Morgan projecting the combined figure hits $249 billion by 2026.
That is not a budget line. That is an industry-redefining allocation of capital that leaves no room for side projects.
The retreats followed the capital. Microsoft shut down Azure for Operators entirely in June 2024. AWS killed its Private 5G service in May 2025. Google didn’t stage a clean exit, but its posture shifted from competitor to co-existence partner: at MWC 2026, Google’s Gemini models were powering Nokia’s Network as Code platform – Google supplying AI, operators supplying the network. The telco encroachment story didn’t end with a dramatic industry battle. It ended with the hyperscalers simply having something more important to do.
How the landscape shifted:
2020-21 – Microsoft launches Azure for Operators, acquires Affirmed Networks, Metaswitch, and AT&T’s Network Cloud. AWS announces Private 5G ambitions. Google places compute at operator edges. Existential threat becomes concrete.
2022-23 – AWS Private 5G launches at $14,000 – a direct enterprise bypass of the operator. Microsoft expands Azure for Operators at MWC 2023. Hyperscaler encroachment reaches peak intensity.
2023-24 – Generative AI rewrites Big Tech capital priorities overnight. Hundreds of billions redirect toward frontier models, GPU clusters, data centres. Strategic bandwidth for telco plays evaporates.
2024-25 – The retreats Microsoft shuts down Azure for Operators (June 2024). AWS kills Private 5G (May 2025). Google pivots to AI co-existence partner with operators at MWC 2026. The field clears.
Now – is the window. Hyperscalers capital-locked in AI infrastructure. 44 percent of enterprises increasing connectivity budgets >10 percent in 2026 (IDC). The value layer is open. The question is whether telcos have the conviction to take it.
Act II | Three customer segments; one unfair distribution advantage
Here’s what the “dumb pipe” framing always got wrong. Telcos don’t just serve consumers. They have trusted, embedded, billing-backed relationships with every meaningful segment of the economy – simultaneously. Consumers, SMBs, and large enterprises.
That cross-segment reach is something no hyperscaler, no neocloud, no AI startup can buy. It took decades and billions in spectrum and infrastructure to build. And right now it is sitting undermonetised.
Retail consumers (billing rails, identity verification, device relationships) | Telcos sit between the network and the end user at a point no platform company can replicate without the operator’s cooperation. IDC projects consumers will spend $100 billion via AI agents on smartphones by 2027. Every one of those interactions routes through infrastructure the operator controls. The distribution advantage is already there – it just isn’t being activated.
SMBs (the segment that nobody has yet cracked) | IBM’s February 2026 research is unambiguous: SME AI adoption lags enterprises not because demand is absent, but because nobody has made it simple enough. The complexity barrier is the market opportunity. Telcos already have managed service relationships, simplified billing, and account ownership with millions of mid-market businesses. The first operator that bundles connectivity, compute, and a working AI application for an SMB in a single invoice will find a segment with virtually no switching resistance and years of pent-up demand.
Enterprises (with expanding budgets, in search of trusted partners) | IDC’s 2025 survey of 758 enterprise IT executives: 37.5 percent increased connectivity budgets by over 10 percent in two years; 44 percent plan to do the same in 2026. Large enterprises increasingly need sovereign AI compute – auditable, localised, compliant – that hyperscalers structurally cannot provide. ABI Research forecasts telco GPUaaS revenue at $21.1 billion by 2030. Iliad, Singtel, SK Telecom, and Verizon are already generating real enterprise AI infrastructure revenue. The model works. The question is who scales it.
Act III | IoT – the silent multiplier everyone is underpricing
The IoT argument gets raised in every telco strategy deck and then treated as a future optionality story. It isn’t. It’s a present-tense competitive asset that most operators are leaving on the table entirely.
Telcos are the connectivity substrate for virtually all global IoT – factories, logistics fleets, smart cities, agricultural sensors, connected infrastructure. Each of those deployments generates continuous telemetry streams. The operator that runs the connection owns the data relationship. But right now, that data is being generated, transmitted, and then largely discarded or handed back to the enterprise customer as raw output. That is a catastrophic misread of where the value sits.
The telco that aggregates those streams, applies AI inference at the edge, and returns actionable intelligence to the enterprise customer is no longer selling connectivity. It’s selling outcomes. Connectivity is priced as a commodity. Outcomes are priced as software. That is a fundamental margin transformation, not a product extension.
The infrastructure timing is favourable. Ericsson’s June 2026 Mobility Report shows 5G networks already handling 48 percent of all mobile data traffic, with 85 percent expected by 2031. AI traffic alone is projected to push uplink volumes three times higher by 2031 versus 2025 – exactly the traffic signature that IoT networks and edge AI generate. Telcos are building the physical foundation of AI-native industry right now, whether they frame it that way or not.
“An IoT connection is worth pennies. The intelligence derived from a billion IoT connections in real time is worth a platform. Telcos own the former and are ignoring the latter.”
Act IV | The technology stack that makes it real
The case for telcos isn’t theoretical anymore. But the technology story has moved faster than most strategy articles have caught up with. Here is what the stack actually looks like in mid-2026.
5G SA + network slicing (live, commercial) | Three-hundred and ninety operators are on commercial 5G, with 90-plus on 5G Standalone; 84 commercial differentiated connectivity services are based on network slicing now live globally, up from 65 six months prior (Ericsson June 2026). A hospital, a stadium, and a factory on the same physical infrastructure with contractual performance guarantees. That is a product, not a feature.
GSMA Open Gateway + network APIs (live, accelerating) | Eighty-six operator groups, 300-plus networks, 80 percent of global mobile connections. Standardised APIs – SIM swap, number verification, quality on demand, device location – commercially live across banking, fraud prevention, logistics, and healthcare. Sixty percent of enterprises forecast to deploy network-powered AI solutions by the end of 2026. Telcos are no longer just providing connectivity. They are exposing network intelligence as programmable, monetizable capabilities.
Agentic networks – MCP + A2A (scaling; 2026 inflection) | AI agents are beginning to consume network APIs autonomously via Model Context Protocol and Agent-to-Agent protocols. Telefónica and Nokia are already piloting this: an AI agent invoking SIM-swap and device-swap APIs to prevent bank fraud in real time, without a human in the loop. GSMA’s verdict at MWC 2026: if network APIs are not MCP-accessible, AI agents will route around them. Operators in the flow of the agent economy or bypassed – there is no middle ground.
Sovereign AI infrastructure (scaling, strategic) | At MWC 2026, sovereignty moved from regulatory checkbox to commercial value driver. Ninety-five percent of organisations say private and sovereign AI are important; only 29 percent are acting on it (NTT Data 2026). That gap is the telco opportunity: regulated, local, auditable AI compute that hyperscalers structurally cannot provide. Broadcom’s VMware Telco Cloud Platform 9 enables operators to offer private AI as-a-service with data isolation and compliance built in.
Fixed Wireless Access (growing fast) | Seventy-one percent of FWA providers now deliver over 5G – the largest annual increase in four years (Ericsson June 2026). India overtook the US as the world’s largest 5G FWA market in Q4 2025 with 14.5 million connections. Speed-based tariff plans offered by 57 percent of providers. FWA is solving the SMB and enterprise last-mile problem while generating real ARPU uplift.
AI-RAN (emerging; position now) | GPU-based hardware serving both radio access and AI workloads simultaneously. Still trial-phase in 2026 – no validated ROI benchmark yet. But McKinsey is explicit: wait-and-see risks “irreversible first-mover disadvantage”. Seventy-seven percent of operators in Nvidia’s 2026 survey expect faster-than-anticipated deployment. The window for positioning is now, not after the architecture is universally proven.
Act V | CDRs – the most valuable asset nobody is pricing correctly
Every large telco is sitting on one of the most powerful behavioural datasets on the planet. Call Detail Records – CDRs – document billions of interactions daily: who communicated with whom, for how long, from where, via what channel, on what network conditions. A large operator generates billions of these per day, continuously, across every segment of the economy it serves.
Stack CDRs against billing records, CRM data, network telemetry, and IoT sensor feeds and you have something no hyperscaler can replicate or purchase: a real-time, high-resolution map of economic activity across an entire country, updated every second. The use cases aren’t speculative. Retail enterprises get location strategy intelligence from anonymised mobility data. Financial services get fraud signals from communication pattern analysis.
Governments get population mobility analytics for infrastructure planning and emergency response. Logistics companies get real-time fleet intelligence from network telemetry. Each of these is a revenue line, not a data science experiment.
But the deeper value is structural. IBM’s research makes the point precisely: AI models fine-tuned on proprietary telco data create differentiation that no competitor can replicate, because the training data is structurally inaccessible to anyone outside the operator. A generic foundation model trained on public internet data cannot do what a model trained on a decade of CDRs and network telemetry can do for a specific enterprise vertical.
That is a moat. Telcos should be building it aggressively, not treating their data as a compliance liability.
Key market data points:
Billions+ – CDRs processed daily by a large operator
$21bn – Telco GPUaaS revenue forecast by 2030 (ABI Research)
$200bn – Edge compute market estimate (STL Partners)
$16bn – New economic value AI could create for US telcos by 2030 (McKinsey)
Act VI | Why first movers don’t just win – they lock everyone else out
Here is the structural point that almost every telco strategy discussion misses. The telecom industry is unusually homogeneous. Every national operator – Nordic incumbent, Asian challenger, Latin American operator – faces the same core problem set: commoditised connectivity, ARPU pressure, capex intensity, enterprise customers demanding more, and consumers who can’t tell the difference between networks.
The pain points are not just similar. They are, in most cases, identical.
This homogeneity is normally treated as evidence of the industry’s stagnation. It is actually its most powerful strategic asset. Because it means that any solution built effectively in one market is immediately exportable to every other market. An AI-powered SMB bundle that works in South Korea works in Germany, Brazil, and South Africa. The customer problem is the same. The data architecture is largely transferable. The playbook replicates.
The first operator to build a working CDR-derived enterprise intelligence product can sell that capability globally – to operators in markets where it doesn’t have spectrum, via licensing, joint ventures, or platform agreements. The second operator to build something comparable starts from zero, at full cost, with no head start. The fifth operator buys a licence on someone else’s terms.
McKinsey flags this directly: a wait-and-see posture on AI infrastructure risks “irreversible first-mover disadvantage”. That word – irreversible – deserves to land. This is not a market where being second or third is a defensible position. The network effects of AI platform ownership compound fast and the early positions calcify quickly. Telcos that move now are building something worth owning globally. Telcos that wait are negotiating their place in someone else’s ecosystem.
“The telecom industry’s structural homogeneity – usually cited as the symptom of its stagnation – is actually its most powerful export mechanism. Solve it once in Seoul. Sell it in São Paulo, Stockholm, and Sydney.”
Act VII | What execution actually looks like
The opportunity is real. The assets exist. The technology is ready. The more honest question is whether organisations are structured to capture it – and here, the industry’s track record gives reason for urgency rather than comfort.
McKinsey’s December 2025 survey of 49 telco CxOs is candid: most operators have yet to convert AI investment into structural value – not for lack of ambition, but because pilots have run ahead of operating models. The gap between experimentation and transformation is real, but it is also closeable.
The operators pulling ahead share a recognisable pattern: enterprise-wide commitment rather than isolated experimentation, end-to-end workflow transformation rather than task-level automation, and a data strategy that treats CDRs and network telemetry as product assets rather than operational byproducts. The platform mindset matters too – build once, deploy repeatedly, structure IP so it travels.
That last part is what separates a successful domestic deployment from a global business. And given the homogeneity of the industry’s challenges, global reach is very much on the table for those who move decisively.
Epilogue | The window is open, but won’t stay open
The dumb pipe narrative was always a partial truth. Yes, telcos became commoditised at the connectivity layer. But they never stopped owning the physical infrastructure, the customer relationships, the regulatory standing, and the data flows that AI at scale requires. What changed is that those assets are now strategically relevant in a way they weren’t two years ago – and the companies that would have competed hardest for them are currently absorbed in a capital allocation cycle that will keep them occupied for years.
Enterprise AI needs sovereign low-latency compute. SMBs need someone to make AI accessible without a PhD and a systems integrator. IoT needs a data intelligence layer, not just a connectivity pipe. CDRs need to become products. 5G SA, Open Gateway, agentic networks, sovereign AI infrastructure, and AI-RAN need to become a platform. Every one of these is an executable move for a telco with conviction.
The operators who act in 2025 and 2026 won’t just win in their home markets. They will have built global platforms at a moment when the field was temporarily clear. That is a rare thing in any industry. In telecom, it may be a once-in-a-generation thing.
The rest will be writing think pieces about what they should have done.
Neeraj Kumar is Chief Technology Officer at Zenith System Solutions, where he leads technology strategy across telecom, enterprise, and AI infrastructure domains. He writes on the intersection of network technology and emerging business models in the telecommunications industry. Connect with Neeraj on LinkedIn.
Sources | McKinsey Telco CxO Gen AI Survey (Dec 2025) · IDC Telco Forum 2026 · Ericsson Mobility Report (Jun 2026) · ABI Research Telco AI 2026 · IBM IBV Telecommunications in the AI Era · NTT Data Global AI Report 2026 · J.P. Morgan AI Infrastructure Outlook 2026 · WEF Strategic Role of Telecom Providers in AI Value Chain 2026 · NVIDIA State of AI in Telecommunications 2026 · GSMA Open Gateway MWC 2026 · Juniper Research Top 10 Telecoms Trends 2026
