The Kilowatt Network
How AI Is Turning European Telecom into an Energy Industry
For two decades, digital connectivity was framed as a triumph of dematerialization. Data moved through “the cloud”, communication became weightless and networks appeared to scale through software alone. The physical systems that sustained this illusion — substations, cooling plants, diesel generators, copper and fiber — receded into the background, treated as stable utilities rather than strategic constraints.
In reality, telecommunications has always been an energy business disguised as an information service. Every text message, video call or streamed film passes through a chain of powered infrastructure that never sleeps: radio transmitters radiating continuously, switching centers processing packets in real time, fiber links amplified across continents and backup systems ready to sustain operation during outages. The system’s defining feature is permanence. Unlike most industries, telecom networks cannot power down when demand falls; they must remain instantly available everywhere.
Until recently, this energy burden was manageable and predictable. For major European operators such as BT, Deutsche Telekom, Orange, Telefónica and Vodafone, electricity typically represented a modest share of operating expenditure — significant, but not existential. Efficiency improvements in hardware and network design offset traffic growth, allowing companies to carry exponentially more data without proportional increases in power consumption. The infrastructure expanded, but the thermodynamic ceiling remained distant.
That equilibrium is dissolving.
“BT Group consumes almost 1% of all electricity in the UK. The current energy demand underlines how crucial it is that we remove legacy networks such as 3G; they are the biggest culprits of energy waste.”
— Howard Watson, Chief Security & Networks Officer, BT Group (BT Group Manifesto / Newsroom)
Watson’s remark captures a structural shift. Telecom operators are no longer optimizing energy use within a stable system; they are dismantling entire generations of infrastructure to make room for new ones whose power demands are fundamentally higher. The transition from 4G to 5G provided an early warning. While 5G is more energy-efficient per transmitted bit, its dense architecture — thousands of additional cells, massive MIMO antennas, edge processing nodes — can consume significantly more power per site. Efficiency gains at the micro level translate into higher consumption at the macro level because capacity expands even faster.
Artificial intelligence is now accelerating that dynamic beyond incremental change into systemic transformation.
From Pipes to Processors
Traditional telecommunications operated on a simple premise: transport data from point A to point B as reliably as possible. Routers forwarded packets, base stations transmitted signals and the economic model rewarded capacity and coverage. Intelligence resided largely at the edges — in user devices or centralized servers — not within the network itself.
AI inverts that architecture. Networks increasingly analyze, filter, compress, secure and optimize traffic in real time. They predict demand, allocate spectrum dynamically, detect anomalies and support autonomous systems that cannot tolerate latency. Instead of passive conduits, they become active computational environments.
This distinction is crucial from an energy perspective. Moving information consumes power, but thinking about information consumes far more. A router expends energy to switch packets; an AI accelerator expends energy to perform complex matrix operations, maintain high-bandwidth memory access and dissipate heat densities approaching those of industrial equipment. When such processors proliferate across thousands of distributed locations, the network begins to resemble a planetary-scale supercomputer.
The correlation between traffic growth and electricity use therefore breaks down. Data volumes may increase linearly, while computational intensity — driven by AI inference, encryption, analytics and real-time decision systems — grows exponentially. Deutsche Telekom’s and BT’s networks, once optimized primarily for throughput, are evolving into infrastructures where processing power becomes as critical as bandwidth.
The Edge Explosion
Hyperscale cloud computing concentrated processing in vast, centralized facilities designed for efficiency. Edge computing disperses it. To support applications such as autonomous vehicles, smart manufacturing, augmented reality and mission-critical control systems, computation must occur within milliseconds of the user — often within tens of kilometers or less.
This requirement drives the deployment of thousands of micro-datacenters embedded within telecom networks: at base stations, aggregation hubs, urban exchanges and industrial campuses. Each site demands continuous electricity, cooling, physical security and maintenance. Unlike hyperscale facilities, these nodes cannot rely on extreme economies of scale; they replicate infrastructure in miniature across entire continents.
“We believe that more than two-thirds of traffic on our networks by 2030 will be driven by AI. The impact on network planning, configuration and investment will be profound. Parts of the network will need to be re-engineered.”
— Christel Heydemann, CEO, Orange (TelecomTV / Capgemini Conversations for Tomorrow, January 2026)
Heydemann’s projection highlights another shift: AI traffic is not merely larger; it is structurally different. Machine-to-machine communication, sensor streams and generative systems generate upstream flows and real-time interactions that stress networks in new ways. Meeting these demands requires not only additional capacity but fundamentally new architectures optimized for computation at the edge.
The cloud once promised to abstract away geography. Edge computing reintroduces it, binding digital services to physical proximity and, by extension, to local energy availability.
Europe’s Sovereign Energy Gap
If the transformation of telecom into an energy-intensive industry were occurring in a uniform global landscape, the challenge would be formidable but manageable. It is not. European operators confront structural disadvantages that amplify the impact of rising electricity demand.
Unlike many American technology giants, European telcos rarely possess vertically integrated energy assets. They operate across multiple national markets, each with distinct regulatory regimes, grid conditions and pricing structures. Negotiating long-term power agreements becomes a fragmented process rather than a unified strategy. Meanwhile, universal service obligations require them to maintain coverage even in regions where energy infrastructure is costly or unreliable.
“We now carry ten times as much traffic as in 2017 with the same energy consumption… but given the crisis we are in, that is still not enough. European operators need to aggregate their energy demand to support large-scale renewable projects as anchor tenants.”
— Margherita Della Valle, CEO, Vodafone Group (Mobile World Live, 2023–2024)
Della Valle’s call for collective action reflects a deeper issue: energy procurement is becoming as strategic as spectrum acquisition. While hyperscalers sign massive power purchase agreements or invest directly in generation capacity, telecom operators remain largely dependent on external utilities. The result is a “sovereign energy gap” in which Europe’s digital infrastructure lacks direct control over the resources that sustain it.
Net Zero Meets Network Reality
European telecom companies are also among the most committed to decarbonization targets. They publish detailed sustainability roadmaps, invest in renewable energy sourcing and pursue aggressive efficiency measures. Yet the growth of AI-driven services risks outpacing these gains.
Shutting down legacy networks such as 2G and 3G can free capacity, but modernization itself requires new equipment, new sites and new computational capabilities. The paradox is stark: digital technologies are promoted as tools for reducing emissions across the economy, yet the infrastructure enabling them may experience rising absolute energy consumption.
Tim Höttges, CEO of Deutsche Telekom, has repeatedly emphasized this tension between growth and sustainability. Investments in AI hardware promise new revenue streams but also raise questions about environmental impact and long-term operating costs. The pursuit of digital competitiveness cannot ignore thermodynamic realities.
Who Pays for the Power?
Behind the technical challenges lies a political-economic question with far-reaching implications. Telecom operators invest heavily in infrastructure to support services that generate value primarily elsewhere in the digital ecosystem — streaming platforms, cloud providers, social networks and increasingly AI companies. Traffic surges driven by these actors translate directly into energy costs for the networks that carry the data.
“We are the gateway to the future. Nothing will happen in the digital age without us. But it is impossible to face new times with old rules. Our sector is not asking for privileges, but for fairness.”
— José María Álvarez-Pallete, Chairman & CEO, Telefónica (MWC / Telefónica Press Office, 2024)
Álvarez-Pallete’s appeal to fairness echoes the long-running “fair share” debate in Europe over whether large technology companies should contribute more directly to network funding. As electricity becomes a dominant cost driver, the argument gains urgency. If AI services generate massive data flows requiring continuous processing, should the burden of powering the underlying infrastructure fall solely on telecom operators and their customers?
The answer will shape pricing models, regulatory frameworks and the future structure of the digital economy.
Telecom as a Distributed Energy Utility
An alternative trajectory is emerging — one that reframes telecom companies not as passive consumers of electricity but as active participants in energy systems. Their networks already include extensive battery installations to ensure resilience during outages. Aggregated across thousands of sites, this storage capacity is substantial.
By participating in demand-response programs, operators could feed power back into the grid during peak shortages or stabilize local networks through distributed storage. Solar installations at base stations, microgrids for critical facilities and intelligent load management systems could transform telecom infrastructure into a flexible energy resource rather than a fixed liability.
Such a shift would blur the boundary between communications providers and utilities. Telecom companies would manage flows of electrons as well as information, optimizing both for reliability and cost. In regions facing increasing grid volatility, this capability could become strategically valuable — even essential.
Resilience, Security and the Physical Internet
As energy dependence deepens, the resilience of telecommunications becomes inseparable from the resilience of power systems. Blackouts disrupt communication; communication failures hinder recovery from blackouts. Cyberattacks targeting energy infrastructure can cascade into digital outages, while natural disasters can simultaneously damage both networks.
Governments increasingly classify telecom and energy as interdependent critical infrastructures. Ensuring continuity of service requires coordinated planning across sectors traditionally managed separately. The stakes extend beyond commercial performance to national security, emergency response and economic stability.
The Kilowatt Constraint
For decades, the narrative of digital progress emphasized Moore’s Law, bandwidth expansion and software innovation. Each generation of technology promised faster speeds, lower latency and greater connectivity, with physical limits appearing distant or irrelevant. AI has changed that perception by dramatically increasing the energy intensity of computation while dispersing it across the network.
Telecommunications companies now confront a constraint more fundamental than spectrum scarcity or chip supply: the availability, price and reliability of electricity itself. Their future competitiveness may depend less on technological breakthroughs than on access to power markets, grid capacity and energy storage.
The internet once seemed to float above geography, immune to the constraints that shaped earlier industrial systems. It is becoming clear that this was an illusion. Digital civilization runs on physical infrastructure whose limits are measured in megawatts, cooling capacity and material supply chains.
The future of connectivity may therefore be decided not only in laboratories or regulatory forums but in the economics of energy production and distribution. For Europe’s telecom giants, the monthly electricity bill is no longer a background expense. It is a strategic horizon — a boundary that will define how far the digital age can expand.
In that sense, the decisive competition of the AI era may not be for algorithms or even data, but for the power required to sustain intelligence at scale. Bandwidth built the networked world. Kilowatts may determine its future.
Photo credit:
Illustration: AI-generated visual concept (Altair Media)
Caption:
From currency to computation: modern networks transform financial investment into energy, infrastructure and intelligence.
