미국 통신사의 설비 투자가 감소, AI 대기업이 주역으로 부상

US is still world's biggest market but capex dipped 8% in 2024, to $81B, and 1Q25 guidance was weak; telcos shifting to monetization amidst policy uncertainty, inflation fears, and the generative AI bubble.

This brief examines the near-term outlook for telecom capital expenditures (capex) in the US, based on the 1Q25 earnings of key operators, including AT&T, Verizon, Charter, Comcast, T-Mobile (DT), Lumen, Frontier, and more. It analyzes how current spending patterns compare with previous forecasts, highlights the main forces shaping investment decisions, explores implications for vendors, and discusses trends in emerging areas such as AI, large language models (LLMs), and data centers.

The US market closed 2024 with $80.5 billion in telco capex and $505.2 billion in revenues, representing 27% and 28% of global industry totals, respectively. The USA's capex to revenue ratio, or capital intensity, was in the 17-18% range in 2022-23. That was well above historic levels. The ratio started to moderate in 2024. For full-year 2024, capital intensity was 15.9%. Based upon 1Q25 earnings calls and other datapoints, that will fall further. US telcos are looking to monetize recent investments, and more focused on conserving capital and cutting costs. Economic and policy uncertainty continues to be high, with inflation and recession fears stemming from the presidentially contrived trade emergency.

As telcos conserve capex, they must see irony in the massive investment bubble in the data center market. Telcos are struggling to attract the same investor interest. However, telcos are finding other ways to benefit from AI in the US. They aren't positioned well to ride the GenAI wave, but they are deploying AI-based technologies to deliver operational efficiency.

For instance:

• AT&T: using AI to improve call center automation, software development, and digital acquisition. The company says that investment in automation is helping offset labor costs and drive self-service efficiency across sales and support channels.
• Verizon: integrating AI into customer care and network ops; use cases include predictive maintenance, customer service automation, and internal process efficiencies.
• Comcast: AI is deployed across billing automation, network troubleshooting, and customer service, reducing truck rolls and call center loads. Machine learning models are used to predict customer churn and drive targeted promotions.
• Charter: has invested in AI for years, focused on frontline efficiency and self-service; it is reducing repair calls and truck rolls. Internal tools are used to optimize operations, enhance agent support, and reduce OPEX.
• DT/T-Mobile: AI is integral to DT's digital transformation, aiming at Euro-800 million in group cost savings by 2027. AI is used to optimize fiber rollouts, automate customer service (e.g., chatbots solving 50% of issues without humans), and remotely manage 75% of routers. DT says that AI supports operations across service, coding, sales, and mobile RAN monitoring.
• Lumen: Lumen is positioning itself as a backbone provider for the AI economy. Lumen Digital platform uses AI for automation and real-time network control.

Over the next few quarters, US telcos will remain tightly constrained on capex. Weaker economic growth will hit revenues. Rising government debt and tariffs will cause inflation to spike. Fewer international visitors to the US, less foreign investment, both lead to weaker demand. Tariffs may also cause shortages or high costs in certain component parts used in building networks. Labor costs will rise, as the surge in investment from an adjacent sector (data centers) will make it harder for telcos to keep their staff. But telcos have been wrestling with such issues for years, and they will survive this as well.

Research Coverage

Organizations mentioned:

Table of Contents

Summary

Market background - USA

1Q25 results

AT&T

Verizon

Comcast

Charter Communications

T-Mobile (DT)

Frontier Communications

Lumen Technologies (ex-CenturyLink)

Dish Network (EchoStar)

Appendix