AI Is Reshaping Telecoms. Can the Industry Keep Up With Its Energy Demands?

This Industry Viewpoint was authored by Sean Varley, Chief Evangelist at Ampere Computing

To deploy AI at the edge, telcos must rethink compute efficiency from the silicon up

Artificial intelligence is quickly becoming the cornerstone of innovation in the telecommunications industry. From intelligent network operations and predictive maintenance to personalized services and RAN optimization for client AI over networks, AI is poised to redefine how telcos operate and compete. According to Fortune Business Insights, the AI in telecommunications market is expected to grow from $2.36 billion in 2023 to $58.74 billion by 2032. But as AI becomes more pervasive across the network stack, a critical question looms: how can the industry deliver the compute power required to run these workloads—especially at the edge—without breaching strict energy, space, and cost constraints?

Telcos are already under immense pressure to modernize legacy networks while reducing emissions, improving reliability, and cutting operational costs. Edge environments, in particular, are uniquely challenged. These deployments demand high-performance processing in limited footprints, often with little room for cooling or extravagant power consumption. Layer AI inference on top of that, and the strain on infrastructure becomes even more acute.

The Cost of Clinging to Legacy Compute in an AI-Driven World

Legacy compute solutions, like x86-based CPUs, have long dominated telecom infrastructure, but they are increasingly ill-suited for this next era of AI-powered networks. Not only are they energy-intensive, but they also lack the scalability and efficiency required to support the distributed, real-time processing AI demands. With the rise of vRAN, Open RAN, and dynamic spectrum management, the complexity of network functions is only growing, along with their compute and energy footprints.

These implications are substantial. The ICT sector already accounts for approximately 3.6% of global electricity usage and 1.4% of total CO₂ emissions, according to Ericsson. As AI accelerates, those figures are expected to climb unless proactive steps are taken to curb consumption.

This isn’t just a sustainability concern—it’s a financial one. Rising energy costs are eroding margins and creating volatility in operational budgets across the telecom industry. As AI adoption expands, particularly at the edge, so too does the energy burden of maintaining outdated infrastructure. This not only accelerates emissions but increases reliance on costly cooling and power systems, further straining environmental and financial resources. In a world where climate impact is under growing scrutiny from regulators, investors and customers, sticking with legacy compute risks leaving telcos on the wrong side of both innovation and sustainability.

AI at the Edge Demands a New Compute Playbook

To stay competitive—and responsible—telcos need a new class of compute: one that delivers AI-grade performance while mitigating the energy drag. Contrary to the common suggestion to use GPUs as a modern approach to these edge-related AI services, GPUs are not the answer. Telecoms are already energy constrained enough without accounting for an energy-intensive GPU infrastructure build out.

This is where next-generation, energy-efficient processors come in. Built on modern Arm-based architecture rather than legacy x86, these processors are engineered from the ground up for efficiency. They consume significantly less power per operation, making them ideal for edge deployments where performance-per-watt is the most important metric.

The result is a solution tailored for the needs of AI at the edge: compact, efficient, scalable compute that doesn’t sacrifice performance. By adopting these CPUs for both traditional network workloads as well as AI, telecom operators can deploy new AI services—from real-time network optimization to automated customer service—closer to the user, without overloading their infrastructure or budgets and without overhauling infrastructure to accommodate GPUs.

What’s more, this isn’t just theoretical. Early deployments of modern, Arm-based processors in telecom infrastructure are already proving their value in real-world scenarios. Operators are seeing reductions in power consumption and cooling costs, while still achieving the compute throughput needed for complex AI inference. This represents a significant leap forward, not only in performance but in the industry’s ability to grow sustainably.

The Choice is Clear: Performance Without the Power Penalty

As the telecom sector races to harness AI’s potential, it must also confront its carbon and energy realities. The path forward demands a strategic rethink of compute infrastructure—especially at the edge. Energy-efficient CPUs are no longer a niche innovation; they are essential tools for future-proofing networks to accommodate AI.

The telecom industry has a clear choice: implement AI with power-intensive x86 and GPU architectures or embrace the new generation of energy efficient Arm-based compute, purpose-built for AI at scale. In the era of intelligent networks, sustainability isn’t a trade-off. It’s a competitive advantage.

About the author

Sean Varley is the Chief Evangelist and leader of ecosystem development and partner programs at Ampere Computing.  His group is responsible for go-to-market, messaging, strategy and execution of developer, partner and revenue ecosystem growth programs. The team covers marketing strategy, strategic business relationships, business planning and developer recruitment for AI Computing in the rapidly evolving Cloud and Edge server markets.   At Ampere, our goal is to provide Products and Solutions that showcase the disruptive advantage our unique processors provide to AI and traditional computing markets. We elegantly integrate Cloud Native SW and HW that help our customers understand and use our products to drastically transform their AI compute investments into highly efficient, high performance, imminently sustainable service infrastructures.

If you haven't already, please take our Reader Survey! Just 3 questions to help us better understand who is reading Telecom Ramblings so we can serve you better!

Categories: Artificial Intelligence · Datacenter · Energy · Industry Viewpoint