AI’s continued unprecedented growth depends heavily on accelerating investment in the physical and digital infrastructure required to train, deploy, and scale AI workloads—from data centers and fiber connectivity to intelligent networks, power, real estate, and accelerated, GPU-based compute. Historically, telcos have not captured a fair share of growth from tech disruptions that substantially increased data traffic such as the rise of video and social media consumption.
Today that dynamic is beginning to shift, giving telcos a new opportunity to play a central role in enabling the AI era. In the process, this could help reignite the industry’s growth.1
The rapid adoption of generative and agentic AI is driving demand not only for centralized compute but also for distributed infrastructure closer to end users. Building and operating this infrastructure require assets that many telcos already have: extensive fiber networks, national footprints, space and presence at the edge, access to power, and experience managing complex, high-availability networks. The question is no longer whether telcos are relevant to AI infrastructure, but where along the value chain they can compete effectively and profitably.
Would you like to learn more about our Technology, Media & Telecommunications Practice?
What’s at stake?
Between 2012 and 2025, global mobile data traffic grew by over 50 percent per year, while telecom service revenues barely increased by 1 percent annually, with hyperscalers capturing most of the economic value.
The AI wave offers telecom operators an opportunity to address that long-standing value capture challenge. Global data center demand could more than triple by 2030, reaching at least 170 gigawatts, driven largely by AI workloads. Importantly, this is not a single “bet-the-company” opportunity. The AI infrastructure value chain offers multiple entry points with different risk, capital intensity, and return profiles, allowing operators to align ambition with balance-sheet capacity and strategic intent.
Within this space, three value pools are particularly relevant for telcos:
- Fiber connectivity for new data centers in core and Tier 2 cities represents a global revenue opportunity of approximately $30 billion to $50 billion by 2030, as new facilities require high-capacity and often dark-fiber2 connections.
- Beyond connectivity, opportunities are present in offering intelligent, software-defined network services to help enterprise customers manage AI workloads, control annual cloud data transfer fees, and meet latency and regulatory requirements. Commonly referred to as egress costs, cloud data transfer fees are estimated at $70 billion to $80 billion annually, creating scope for differentiated telco offerings focused on performance, cost efficiency, and control.
- Telcos may unlock new value by utilizing existing space, power, and cooling for distributed compute needs, including participation in the rapidly emerging GPU-as-a-service (GPUaaS) market, which is estimated to be worth $35 billion to $70 billion by 2030 (excluding hyperscalers). This opportunity is supported by accelerated compute workloads that are growing at more than 30 percent CAGR and expected to represent more than two-thirds of data center demand within five years.
Who are the key players and stakeholders?
The ecosystem is broad and increasingly competitive. Key stakeholders include hyperscalers and colocation providers, cloud operators, and GPUaaS players—many of which prefer dark fiber. Telcos enter this landscape as infrastructure owners and operators, often competing with—and simultaneously partnering alongside—hyperscalers. Other important players include the public sector; regulators that play a central role by shaping fiber access, data sovereignty, and licensing requirements; and key regulated industries that drive demand for sovereign AI.
Additional value chain participants include chip makers, software providers, systems integrators, and partners supporting data center retrofits and cooling solutions. Finally, energy providers and investors are increasingly influential, given power constraints and the capital intensity of AI-ready infrastructure.
What are the recent important developments?
As AI infrastructure scales and decentralizes, several developments are worth watching.
AI data center capacity is expanding and spreading to new geographies. Hyperscalers and colocation providers have announced plans for more than 2,600 new data centers. Roughly one-quarter are to be located in cities with no existing data center footprint, pushing demand into new markets. Procurement procurement preferences are also becoming clearer, with growing emphasis on dark fiber. Overall, the global footprint is expected to approach 11,000 facilities by the early 2030s.
Power and permitting constraints are reshaping where value is accruing. Grid power limitations and long permitting timelines are slowing new builds, increasing the value of existing space and available power. This dynamic favors players that control “ready” assets—particularly sites with assured, contracted power in locations that can support distributed compute for inference.
Telco participation is shifting from pilots to execution. Early movers among operators are signaling a transition from experimentation toward scaled delivery. Some are partnering with hyperscalers to provide dark fiber and edge compute; others are launching GPUaaS offerings or repurposing central offices or spaces with power and cooling to support AI workloads.
What are the biggest challenges?
Despite the opportunity, risks are substantial. There are still uncertainties around demand, particularly across use cases and geographies, as well as the technology and connectivity needs to enable computing at the scale required by AI providers. Competition from hyperscalers and cloud-native players is intense, with superior scale, capital, and developer ecosystems. And it’s still not a given that investments can generate enough compute capacity to meet AI ambitions, while still delivering strong returns.
Commercial and competitive uncertainty also remains. Intelligent network services do not yet have a clearly established market definition, and telecom operators increasingly compete with hyperscalers and network service providers across connectivity, space and power, and GPUaaS offerings.
Capital requirements are high, especially for GPUaaS, dark fiber, and AI-ready data centers, while technology cycles are short and pricing pressure could intensify as supply expands. Many telcos lack the software, sales, and partnership capabilities required to compete beyond traditional connectivity.
The state of AI in 2022—and a half decade in review
What does it take to succeed?
Capturing value from AI infrastructure requires several shifts for telcos:
- Commercial capabilities including dedicated hyperscaler-focused sales teams, faster design and contracting cycles, digitized routing and asset data, and productized connectivity solutions are essential. GPU-as-a-service offerings require advisory-led selling and customer success capabilities. Close partnerships with hyperscalers are key to understand their evolving needs and new emerging technologies in this space.
- Partnering with and leveraging infrastructure investors, software providers, systems integrators, and retrofit specialists is key to lower capital needs, de-risk projects, and accelerate service delivery and facility readiness.
- Disciplined underwriting is important, including evaluation of fiber routes for multi-tenant upside and careful management of cost structures across compute, power, cooling, networking, and memory.
- Risk management for space and power, including prioritizing sites requiring limited retrofit or committing capital only after securing firm tenant demand, is critical; for edge inference, at least 500 kilowatts of power are typically required to achieve economic viability.
What should players be watching?
Telco leaders should monitor where hyperscalers and colocation providers expand into new markets and how power constraints reshape infrastructure economics. Operators have an opportunity to be early movers in the build-out of distributed compute or edge infrastructure, but much will depend on the level of demand for GPUaaS offerings and whether demand for intelligent network services picks up and can be successfully monetized. Ultimately, the biggest risk facing telcos may not be choosing the wrong path—but failing to choose at all.
