Photo by Brett Sayles on Pexels
The AI Infrastructure Race Is On — But the Finish Line Is Farther Than It Looks
The world’s leading telecommunications companies are in a full sprint to position themselves at the center of the artificial intelligence revolution. Verizon, AT&T, and South Korea’s SK Telecom are among the most aggressive investors, each committing billions of dollars to build out the AI-ready network infrastructure they believe will define the next decade of connectivity. But a sobering new assessment from technology advisory group Omdia is throwing cold water on the hype — not on the vision itself, but on the timeline for returns.
According to Omdia, the headline investment figures being cited across the industry represent ceilings, not guarantees. More critically, the firm warns that revenue growth will trail capacity expansion for years, creating a potentially uncomfortable financial gap that investors and shareholders will need to stomach through the build-out phase.
What These Carriers Are Actually Building
To understand the stakes, it helps to unpack what “AI infrastructure” actually means for a telecom operator. It’s not simply about installing faster antennas or upgrading core networks. The AI backbone these carriers are assembling spans several interconnected layers: edge computing nodes positioned close to end users, high-capacity fiber backhaul to support low-latency data flows, purpose-built data centers with GPU-dense compute clusters, and AI-native network management platforms capable of self-optimization in real time.
Verizon has been notably aggressive in its edge computing ambitions, leveraging its distributed fiber assets and the geographic density of its cell sites to offer enterprises low-latency compute at the network edge. The carrier has positioned its Mobile Edge Compute (MEC) infrastructure as a launchpad for AI inferencing workloads — use cases ranging from real-time video analytics to industrial automation.
AT&T, meanwhile, has doubled down on its fiber strategy as the connective tissue for AI delivery, while simultaneously investing in open RAN architectures that allow software-defined intelligence to be layered into the radio access network. The company’s partnerships with cloud hyperscalers like Microsoft and Google are central to its AI infrastructure thesis, blurring the traditional line between telecom and cloud.
SK Telecom presents perhaps the most ambitious vision of the three. The Korean operator has openly declared itself an “AI company” rather than a traditional telco, investing in large language model development, AI-powered customer service platforms, and even taking equity stakes in AI startups. Its domestic 5G network — already one of the most advanced in the world — is being re-architected as an AI-native platform from the ground up.
The Omdia Warning: Capacity Is Outpacing Revenue
Despite the compelling strategic narratives, Omdia’s analysts are urging caution on the financial trajectory. The firm’s core concern is a familiar one in the history of infrastructure-heavy industries: overbuilding ahead of demand. The worry is that carriers will spend heavily to provision AI-grade network capacity — low-latency edge nodes, high-throughput fiber rings, GPU compute — only to find that enterprise and consumer demand ramps far more slowly than anticipated.
This creates a structural problem. Unlike traditional network upgrades, AI infrastructure carries significantly higher upfront capital costs, particularly when GPU procurement and specialized data center construction are factored in. If utilization rates remain low during the critical first few years of deployment, the return on invested capital could be severely compressed, pressuring already-thin telecom margins.
Omdia’s analysts also point out that the competitive landscape for AI infrastructure is intensely crowded. Telecom carriers aren’t just competing with each other — they’re competing with hyperscalers like AWS, Microsoft Azure, and Google Cloud, all of which have deeper AI engineering expertise, massive existing customer relationships, and the ability to deploy capital at a scale that even the largest telcos cannot easily match.
The Monetization Challenge
One of the thorniest questions facing AI-investing carriers is precisely how they plan to charge for this new infrastructure. Traditional connectivity pricing models — per-megabit, per-subscriber — don’t map cleanly onto AI workloads. Enterprises consuming AI inferencing at the edge, for instance, may value latency and reliability far more than raw throughput, requiring entirely new service-level frameworks and pricing constructs.
Some carriers are exploring consumption-based models tied to compute cycles rather than data transfer, while others are packaging AI capabilities into managed service bundles aimed at enterprise verticals like healthcare, manufacturing, and logistics. But these new business models are largely unproven at scale, and sales cycles for complex enterprise AI services tend to be long and unpredictable.
The Long Game: Why Carriers Are Building Anyway
Despite the cautionary signals from analysts, the carriers pressing forward argue that the alternative — waiting on the sidelines — is far more dangerous. The telecom industry’s history is littered with examples of operators who failed to invest early in transformative infrastructure cycles, only to find themselves disintermediated by more aggressive competitors or technology substitutes.
The argument goes that AI will eventually become as foundational to enterprise operations as cloud computing is today, and that the carriers who own the low-latency, high-reliability network infrastructure closest to where AI workloads run will be uniquely positioned to capture value that pure-play cloud providers cannot.
Whether that thesis holds — and whether it generates the financial returns shareholders expect on a reasonable timeline — remains the defining question hanging over telecom’s biggest AI bets. For now, the backbone is being built. The billions, as Omdia reminds us, are still largely waiting to arrive.
Industry Outlook
Analysts broadly expect 2025 and 2026 to be peak capital expenditure years for AI infrastructure among major carriers, with revenue inflection points unlikely before 2027 at the earliest. The carriers that navigate this gap most effectively — through disciplined capital allocation, smart partnership strategies with hyperscalers, and agile enterprise sales execution — will likely emerge as the defining connectivity platforms of the AI era. Those that overbuild without demand to match may face difficult conversations with investors in the years ahead.
