Why telcos cannot become tech companies (Analyst Angle)

It’s an idea that sounds right — but is it possible?

Every few years, another telecom strategy presentation declares that “our goal is to become a tech company.” The phrase sounds inspiring. It signals ambition, innovation, and a desire to escape regulatory gravity. But it is wrong and potentially value-destructive.

The difference between a telecom company and a technology company is not culture, capital, or leadership. It is physics. Telecom assets behave differently from software assets. They obey different cost curves, scale laws, and risk profiles. When you ignore these differences, you get strategies that sound visionary but destroy value.

If telecom leaders want to understand where real transformation lies, they need to start from first principles, not PowerPoint analogies.

Scope

This discussion focuses on mobile network operators. Fixed-line fiber networks face different constraints: higher per-premises capital costs, even greater geographic fixity, and simpler operations once deployed. The principles of asset fungibility, operating leverage, and cost scaling apply to both. Wireless networks dominate global subscriber and revenue growth, and the interplay between spectrum, coverage, and mobility creates the sharpest contrast to hyperscalers.

I. The physics of hyperscalers versus telcos

Hyperscalers and telecom operators both build large networks of infrastructure. But the way those networks generate return on capital is fundamentally different.

In cloud infrastructure, capacity can be redeployed anywhere in the world with almost no friction. Spare compute in Oregon can serve customers in Singapore. The ability to reassign assets dynamically, what I am calling fungibility, turns fixed cost into a lever for exponential growth.

Telecom networks, by contrast, are anchored to geography. A radio site in Saskatchewan cannot serve a customer in São Paulo. Once built, it is immovable. Every cell site, trench, and spectrum license is a sunk cost defined by location. Utilization is bounded by local demand.

That is the first law of telecom economics: telecom is spatially constrained infrastructure.

II. Four laws of telco economics

Telecom networks are governed by physical laws that define how capital turns into capacity, and capacity into cash flow. These laws are not strategic choices. They are structural realities that shape every investment decision, pricing model, and transformation effort in the industry. Understanding them separates strategy from storytelling. Each law captures a constraint that software economics can bend but not break: geography limits utilization, spectrum scales in steps, coverage expands by duplication rather than pooling, and integration determines whether those costs compound or cancel. Together, they explain why telecom returns flatten as networks grow while hyperscaler returns accelerate with scale.

Law 1: Geography defines utilization

A cell site cost is mostly fixed, while the revenue it can capture depends on local density and on the fraction of each user’s traffic that passes through that site. Busy-hour throughput typically runs four to six times the daily average, producing low time-averaged utilization. In many mature markets, time-averaged RAN utilization sits in the 15–30% band; hyperscaler data centers often reach 70–90% by pooling uncorrelated workloads.

Worked example, traffic-weighted attribution (transparent assumptions)

• ARPU = $50 per month = $600 per year.
  
• Traffic fraction per site = 10%. This reflects users traversing many sites per day.
  
• Urban site users = 300. Rural site users = 30.

Step math

1. Per-user annual ARPU = $600.
   
2. Site-attributable revenue per user = $600 × 0.10 = $60.
   
3. Urban site revenue = 300 × $60 = $18,000 per year.
   
4. Rural site revenue = 30 × $60 = $1,800 per year.

Typical direct site OpEx (lease, power, maintenance, field visits) runs $80,000–$200,000 annually, depending on market and site complexity. The arithmetic shows why individual sites can appear loss-making even when the network as a whole is profitable. Changing assumptions moves the picture. Raise traffic share to 20% or ARPU to $100, and per-site numbers improve. However, the structural point holds: per-site economics are highly sensitive to traffic share, user density, and busy-hour skew, and spare capacity in low-density areas cannot be redeployed to higher-value locations.

Concrete operator examples

• Urban macro in London or New York: high subscriber density and heavy daytime concentration produce much higher utilization than remote rural sites in northern Canada or central Australia.
  
• Stadiums and transit hubs: short duration high intensity. These sites require provisioning for peak but have low time-averaged utilization, driving cost per useful byte much higher than in data-center-like environments.

Law 2: Spectrum economics follow step functions

Spectrum is bought in lumpy chunks. Examples: the U.S. C-band auction raised roughly $80 billion in 2021. T-Mobile paid about $9 billion for wide 2.5 GHz holdings in earlier mid-band acquisitions. Once purchased, spectrum yields near-zero marginal cost per additional user until the network hits congestion. At that point the operator faces a discrete set of investment options: add carriers, densify with small cells, upgrade backhaul, or acquire more spectrum. These options come with six-figure to multi-million-dollar price tags per site or license.

Concrete illustration

• A metro sector may operate at 60% busy-hour utilization and add users with negligible additional CAPEX. When utilization passes an empirical threshold, for example, 80–85%, the operator must densify. Small-cell deployments can cost $50k–$200k per site installed in urban settings when mounting, fiber, power, and permits are included; a macro site densification or a new carrier deployment can exceed $200k–$500k in complex environments — the economics jump, not slope.

Investor implication

• The step-function creates long plateaus of attractive marginal economics followed by sudden reinvestment cliffs. Forecasting returns requires modeling those pieces, not assuming smooth marginal cost curves.

Law 3: Networks scale by coverage, not concurrency

Hyperscalers add nodes to increase pooled capacity. Telcos add sites to increase reach. The economics differ because incremental coverage costs rise as density falls.

Concrete examples

• Rural FTTH costs: $2,000–$5,000 per premises in low-density territories versus $500–$800 in urban areas.
  
• National coverage: the first 90% of population coverage often captures the densest, cheapest users. The last 10% typically contains the most costly to serve per customer. In several OECD countries, reaching the final 5–10% can cost three to five times more per user than the prior coverage tranche.

Operational effect

• Workforce geometry matters. Ten thousand urban-adjacent sites can be serviced efficiently from a handful of depots. Ten thousand widely dispersed rural sites require many more crews, regional depots, and higher travel time overhead, increasing OpEx per site.

Competitive contrast

• Hyperscalers choose regions with demand concentration. Telcos must serve geography because consumers are geographically distributed and regulators often mandate universal service.

Law 4: Integration creates value; isolation exposes cost

Networks are systemic. Integration across spectrum, RAN, core, and backhaul creates operational synergies. Disaggregation can reveal hidden asset value, but it also transfers value to third parties and introduces coordination friction.

Examples and mechanics

• Tower monetization. Many operators sold towers to TowerCos like American Tower or Crown Castle to free capital. This improves balance sheets and converts CAPEX into predictable lease costs. It also transfers upside from tower appreciation to tower investors and imposes recurring rent. The net effect is not uniformly beneficial to operating economics; it depends on how well operator leasing terms, maintenance coordination, and upgrade windows are negotiated.
  
• Neutral-host models. Airport DAS or subway neutral-host systems reduce duplication and lower total cost for carriers, but they require a governance model and often a third-party operator to run the shared system. The gains are real in dense venues, but they do not make rural sites fungible.

Operational caution

• The more you modularize the stack, the more you trade systemic optimization for discrete accounting clarity. Integration lets you optimize truck rolls, spare parts pools, and grading of capacity across layers. Isolation turns those levers into contractual processes and tends to reduce the effective utilization benefits that integrated systems capture.

Together, these laws explain why telcos and tech companies live on opposite sides of the return-on-capital spectrum.

Hyperscalers turn compute elasticity into increasing returns (α > 1): Each unit of infrastructure supports ever-greater workloads. Telcos, constrained by physics, face diminishing returns (α < 1): Every increment of coverage or capacity adds cost faster than it adds revenue. Investors price that asymmetry, rewarding hyperscalers with growth multiples and valuing telcos as infrastructure utilities.

III. Why the “tech company” analogy fails

When executives say, “We need to think like a tech company,” they usually mean “we need recurring revenue and scalable margins.” Software economics do not map onto network physics.

Software scales through replication: every new customer uses the same code.

Networks scale through duplication: every new customer requires more hardware in a specific place.

You can virtualize parts of the telco stack, but you cannot virtualize geography. Digital transformation, such as AI in operations, APIs for exposure, and data-driven pricing, can improve efficiency, but it does not change the denominator of the business: kilometers, kilowatts, and capital intensity. The telco problem is not digital immaturity; it is structural non-fungibility.

IV. Strategic playbook for telcos

Telcos cannot become hyperscalers. Their assets are bound by geography and physics. But they can evolve into hybrid infrastructure operators with platform shaped characteristics. By this, I mean firms that exploit leverage where scale physics allow it (shared infrastructure, AI-enabled efficiency, spectrum reuse) and innovate digitally where abstraction is possible (API exposure, service orchestration, enterprise integration).

The goal is not to imitate cloud economics, but to industrialize the airwaves: turning coverage, capacity, and connectivity into programmable infrastructure that others can build on. Telcos will never scale like the cloud, but they can make the network behave more like software. That is enough to change the game.

1. Monetize usage growth on fixed assets

• Fixed Wireless Access (FWA): Uses existing spectrum and RAN to add 20–40% incremental revenue. Example: T-Mobile FWA serves ~5 million subscribers, generating $3–4 billion annually. Marginal cost is primarily CPE and modest backhaul upgrades.
  
• Enterprise Private Networks: Higher ARPU for dedicated slices of spectrum on shared infrastructure. Applicable to manufacturing, logistics, campuses.
  
• Wholesale and MVNOs: Sell excess capacity to resellers and virtual operators. Smooths utilization, generates revenue with near-zero marginal cost during off-peak hours.

2. Pursue shared infrastructure aggressively

• Active RAN sharing: Deploy one network to serve multiple operators, lowering cost per busy-hour Mbps by 15–25%. Examples: Cornerstone (UK), MBNL (UK).
  
• Neutral-host arrangements: Pool rural coverage or indoor systems (airports, stadiums, subways) to amortize deployment costs.
  
• Tower divestments: Convert owned towers to sale-leaseback arrangements. Improves balance sheet and reduces capital intensity, though long-term appreciation is forfeited.

3. Leverage AI and convergence technologies

Some technologies genuinely shift telco economics toward hyperscaler characteristics, though only partially.These innovations make operations smarter within fixed physics. AI-RAN, for instance, can improve utilization 10–20% and reduce operating expenses 15–25%, mainly by better capacity allocation and predictive maintenance. But they do not change the fact that coverage assets cannot move.

Convergence, in other words, mostly reshuffles value rather than creates new elasticity. It can improve margins and optionality but cannot convert fixed coverage into global scalability.

4. Competing on integration, not infrastructure

If every operator pursues the same hybrid structure, differentiation cannot come from owning capacity alone. It must come from orchestrating experience. Integrating mobility, compute, edge-services, and reliability into propositions that customers value.

Enterprise differentiation is becoming visible in recent results. For example, Verizon Communications reported its Business segment operating income rose 12.7 % year-over-year in Q3 2025, and segment EBITDA margin increased to 23.4 %. These figures suggest that enterprise services (which include private 5G, edge compute, and managed network offerings) are contributing more profitably, even though overall Business segment revenue declined 2.8 % in Q3. Meanwhile, Singtel’s Digital InfraCo and its Paragon platform received a 2025 “Competitive Strategy Leadership” award from Frost & Sullivan for its GPU-as-a-Service offering leveraging edge, compute and network infrastructure. These milestones support the thesis that integration of network, compute and ecosystem is gaining traction. However, publicly-disclosed evidence still falls short of showing full platform-scale economics (for example, incremental EBITDA per dollar of new capex remains undisclosed).

Consumer differentiation remains more nascent but is supported by recent results. Verizon’s Q3 2025 Consumer wireless service revenue rose 2.4 % year-over-year, and ARPA (average revenue per account) reached approximately $147.91, up ~2.0 %. These results indicate that service quality and bundling may be contributing to modest ARPA growth. On the Singtel side, while detailed consumer ARPU uplift is less publicly disclosed, launches such as its 5G+ Mobile Workspace and integrations with enterprise devices signal a strategy to extend beyond connectivity.

Putting this together, the emerging pattern is that operators who can reliably integrate connectivity, edge compute, services and ecosystem partnerships are beginning to outperform segments of their business. But the thesis must be calibrated:

• The results are stronger in enterprise/private network segments than in consumer-bundle segments.
  
• Most operators still do not publish the full set of transformation indicators (e.g., cost per busy-hour Mbps, revenue per deployed site, incremental EBITDA per capex dollar).
  
• Shared infrastructure (towers, RAN, spectrum) remains a cost burden; the differentiation lever is not new assets but new experiences built on existing assets.

Operators that master integration and reliability will have a competitive edge even when underlying infrastructure is shared. The win does not come from owning more towers but from making connectivity invisible and outcomes visible.

5. Measure what drives transformation

Stop tracking transformation by agile squad counts or innovation lab square footage. Measure outcomes:

If these metrics improve, transformation is real. If they do not, you are running theater. Culture programs and agile training do not matter if economics do not change.

V. The real constraint and the real opportunity

Telcos serve geography. Hyperscalers serve workloads. Geography is fixed. Workloads are mobile. That is the boundary.

The opportunity is to industrialize the airwaves where necessary and innovate digitally where possible:

• Spectrum: High fixed cost, near-zero marginal cost, logical flexibility
  
• Core: Centralized, scalable, strong economies of scale
  
• RAN: Distributed, coverage-bound, shareable, incrementally monetizable
  
• Services: Asset-light, software-defined, platform-enabled

Hyperscalers scale code. Telcos scale coverage. Strategy fails when you confuse one for the other.