AI has the potential to facilitate productivity gains, job creation, and new investment across multiple industries. But S&P Global Ratings believes it can only do so with the telecommunications sector providing the necessary infrastructure that enables higher throughput, low latency, and access to energy. In this report, we look at the emerging symbiotic relationship of a transformative new technology and a mature, complex industry.

A New Symbiotic Relationship Is Formed

AI algorithms already consume a lot of data that is transmitted over telecommunication networks, aiding in the development of this nascent technology. For example, "smart cities" that will use AI to enhance safety, sustainability, and quality of life, will require connectivity over wireless and wired networks; while in health care, next-generation remote patient monitoring and outpatient services will be enabled by data collection and analysis that uses telecom infrastructure.

This offers opportunities for the U.S. telecommunications and cable sector at an opportune moment. The sector's credit quality has suffered over the last couple of years because of elevated capital spending (capex) to support network upgrades, rising interest rates, and technology convergence as mobile operators use their wireless networks for in-home broadband and cable providers leverage their wholesale agreements to offer wireless service, resulting in elevated competition. These factors contributed to depressed cash flows and higher debt leverage. AI can change the sector's fortunes, which in turn can lead to wider AI adoption, creating a virtuous cycle. That said, it can also increase credit risk associated with new technology integration, capacity gluts, or low returns on investment.

We believe the adoption of AI will increase dramatically over the next decade and be largely positive for the telecom sector overall, offering new revenue streams and supporting margin expansion. Most telecom and cable operators are already developing generative AI (algorithms used to analyze existing information to create something new) solutions to reduce costs and improve efficiency for their own business operations. However, to support greater network demand and drive revenue growth, the industry will need to increase capacity that could result in higher capex and weaker free cash flow generation. Given that AI is still nascent, the timing and potential impact on credit quality and ratings are uncertain over the intermediate term. Longer term, we expect the credit impact to the sector to vary, depending on how well companies execute and pay for their AI strategies.

Key Opportunities

The telecom sector is known for its complexity and providers face many challenges from a variety of issues, including the interoperability of network architectures, network outages, resource utilization, customer service, cyber-attacks, and the growing demand for bandwidth. Almost overnight, AI has become a potential solution to many of these problems, including:

Detecting billing anomalies and forecasting future trends.   AI can use predictive analysis (also known as machine learning) to detect billing anomalies by creating algorithms to scrutinize records. This allows telecom operators to flag errors or fraudulent activities, and to forecast future billing trends.

Improved customer service.   Chatbots have become increasingly common across multiple sectors and the telecom industry has used them to aid call center associates, which in turn improves agent productivity.

Marketing.   Telecom operators can create personalized messages or virtual media to target an individual customer with AI. It can also determine what type of messaging a customer would be most receptive to, including rewards, celebrity endorsements or limited-time offers.

Improving back-office operations.   AI-driven models can reduce costs and make procurement more efficient. By automating and optimizing processes, telcos can accelerate order fulfillment, reduce expenses, and ultimately improve the overall customer experience.

Network optimization and predictive maintenance.   Using AI to analyze and reroute data traffic to healthy nodes and modify network configuration can make them more efficient long-term. AI-enabled systems can also detect malfunctions and bottlenecks real-time even before they become visible to control centers. Telecoms can implement predictive maintenance strategies that forecast network demand or equipment degradation and failures. They can also use Generative AI in the planning of network expansions and upgrades.

Similarly, AI can model out mobile network strategies in urban environments and help determine the optimal locations for small cells in cities to ensure minimal interference and strong signal coverage. AI-powered drones and video cameras at tower sites can also be used for inspections.

Detecting potential cyber-attacks.   While cyber-attacks threaten many corporate sectors, they have become increasingly prevalent in the telecom industry, including numerous attacks reported by the three major wireless carriers (AT&T Inc., Verizon Communications Inc., and T-Mobile US Inc.) In the telecom sector, networks are interlinked and house vast quantities of customer data and sensitive information. While telecom operators have beefed up their capabilities to reduce these threats in real time, their efforts are still not preventing them completely. Cyber-attacks have compromised sensitive customer data such as tax IDs, PINs, and payment card information, and pose a headline risk to telecoms.

AI-driven applications can theoretically detect future threats and reduce the response time, allowing wireless operators to eliminate the threat before it can access data.

Improving customer churn.   AI-based algorithms can potentially identify customers that are most likely to switch to a competitor by analyzing usage patterns, payment history, and customer service interactions.

Table 1

5G Mobile And AI Technology Are Made For Each Other

To date, "Internet of Things" (IoT) applications that were expected to drive growth from 5G mobile investments (i.e. spectrum license acquisitions and capex) have yet to become meaningful revenue sources for the wireless carriers and the only new revenue stream that has arisen from 5G is fixed wireless access (FWA) for in-home broadband service.

Nonetheless, FWA is an excess capacity business model since spectrum is a finite resource, thereby limiting the number of homes the carriers can target. We therefore expect FWA customer additions will slow over the next couple of years, constraining longer-term revenue contributions from FWA, which is still very small. For example, FWA only accounts for about 1% of Verizon's total wireless revenue. Further, mobile customers have more value to the wireless carrier than FWA since the revenue per bit for a wireless customer is substantially higher than for in-home broadband, which consumes a lot more data.

Chart 1

We believe 5G wireless technology is well-suited to AI. Unlike previous generations of mobile network technology, 5G enables a network to connect almost everything--from machines to devices. It also has better throughput, including multi-gigabit peak data speeds using mid-band or mmWave spectrum, and is capable of ultra-low latency rates, which could fuel AI-based IoT applications, including autonomous vehicles and Edge AI (allows computations to be done close to where data is collected). When AI algorithms process data at "the edge" of a network (the site at which the data is collected), they reduce the need to transmit it over the entire network, which allows the carriers to use bandwidth more efficiently. That said, Edge AI will also require more investment at the edge of the network. This could include the deployment of mini data centers at the base stations of towers, an opportunity that sparked American Tower's interest in the data center assets of Coresite Realty Corp., which it acquired in 2021.

While 5G-supported IoT use-case adoption has been slower than anticipated, several applications could become more realistic with the increasing adoption of AI:

• Virtual reality. While virtual reality has yet to really gain traction in the marketplace, we believe the integration of AI into this space can enable a more immersive user experience, supported by 5G networks that provide ultra-low latency and fast data speeds for users. Further, AI can offer innovations that make virtual reality more compelling than in its current use.
• Smart cities. AI-powered smart cities can improve safety, sustainability, and quality of life. The infrastructure supporting them would include sensors that use 5G wireless networks.
• Drones. Drones over wireless networks with remote control stations and AI-enabled piloting. Amazon has talked about using drones for package deliveries, which could come to fruition as AI evolves.
• Smart homes. AI-powered home security and automation.
• Autonomous vehicles. Autonomous driving, which requires a wireless network with low latency to function well, to date has not been successful given the potential for accidents, network coverage irregularity, along with legal and regulatory hurdles. With self-driving systems, when there is potential for an accident, the system will return control to the human driver without much time to react. AI can be used to detect people, cars, and other objects on the road to develop collision avoidance systems that make autonomous driving safer.

We believe 5G IoT applications can offer a new source of revenue as FWA growth slows due to competition and capacity constraints. In contrast to mobile services and FWA, the IoT business model for wireless carriers will rely more on the wholesaling of network capabilities to drive incremental revenue, in our view. What is unclear is how much data IoT applications will consume and to what extent networks may become burdened by increasing demand for these services.

While mobile carriers are managing available FWA homes to conserve capacity on their networks, if AI-based IoT gains traction, it is unclear if they will need to cannibalize their existing FWA customer base to free up capacity since some applications may consume more data than is currently envisioned. That said, we believe the trade-off would be more favorable since the revenue per bit is likely to be higher for IoT applications than for FWA.

Opportunities Come At A Cost

AI will fuel a new wave of bandwidth demand for telcos

According to S&P Global Ratings technology industry analysts, AI spending by large hyperscalers (large cloud service providers), which are driving AI demand, will continue to grow significantly over the next several years. Total capital spending by the big four (Alphabet Inc., Amazon.com Inc., Meta Platforms Inc., and Microsoft Corp.) is expected to increase 26% in 2024 with AI-specific spending likely to surge over 40% during the year. To accommodate this investment, these companies will need access to more bandwidth, which is best served by telecom providers with a large fiber footprint. (See "U.S. Tech's AI-wakening: Enterprises Tread Cautiously, Hyperscalers Charge Ahead," May 9, 2024.)

We believe issuers that are likely to benefit from AI-related bandwidth demand from hyperscalers include large telcos like AT&T and Verizon as well as other providers with fiber-rich networks such as Lumen Technologies Inc. and Zayo Group Holdings Inc.

In fact, Lumen recently announced it would provide Microsoft with a custom network that includes dedicated access to the Lumen network as well as the installation of fiber on new and existing routes to strengthen the connectivity between Microsoft's data centers and support the increasing demand for AI-related bandwidth. We expect similar deals for Lumen and other telecommunication providers over the next couple of years.

Chart 2

AI investment has proven somewhat more cautious among enterprise customers, and overall adoption of AI has focused on Gen-AI tools such as ChatGPT as well as streamlining operations and improving productivity. A recent survey by Boston Consulting Group found that 90% of business executives cited AI as a top priority, although more than two-thirds were focused on small scale applications. Another study by IDC expects spending on Gen-AI solutions to double to around $40 billion in 2024 and increase another 75% in 2025.

Chart 3

For residential broadband providers, it will be difficult to monetize increasing bandwidth demand.

For cable and fiber-to-the-home (FTTH) providers that derive more of their revenue from residential and small and mid-sized business (SMB) customers, the ability to monetize increased bandwidth demand is uncertain and will depend on their capacity to raise prices or migrate customers to faster data speeds, which could be challenging in an increasingly competitive environment.

Data center demand will grow, as will investment needs

AI technologies require vast computational power, storage space, and low-latency networking for training and running models. As AI adoption grows, the requirement for data centers to house the hardware on which AI systems operate will also increase. While this dynamic should enable continuous growth in a sector that is already experiencing strong demand tailwinds from cloud service providers and an increasing trend towards outsourcing, it will also require new investment to address the supply/demand imbalance. We assume tech companies will continue making significant investments in proprietary data centers, but we also expect sustained demand for capacity from independent wholesale data-center operators. Separately, we believe that interconnected providers, such as Equinix Inc., will benefit from the need to connect and route AI-related data traffic.

That said, generative AI consumes a massive amount of energy. In fact, the International Energy Agency (IEA) forecasts that global data center electricity demand will more than double from 2022 to 2026, with AI playing a major role in that increase. By some estimates a ChatGPT inquiry, for example, uses ten times more energy than a Google search. To address this surging demand for energy, data centers are incorporating liquid cooling to handle increased power densities in their facilities and power usage efficiency (PUE) becomes paramount. In certain cases, this could require capital investments to retrofit an existing facility. For wholesale providers targeting AI-related demand, access to low-cost and widely available power is critical.

Table 2

AI's Impact Should Be Largely Positive, But May Take Time To Evolve

We expect the benefits to credit quality will be largely positive for the U.S. telecom and cable sector. That said, AI is still in its infancy and returns on investment could take time. While revenue opportunities, cost savings, and improved efficiencies bode well for top line growth and margin expansion in an increasingly mature and competitive industry, telecom and cable providers may need to accommodate greater bandwidth demand with network upgrades, resulting in higher capex that could constrain free cash flow. Additionally, there is risk that companies overestimate bandwidth demand, resulting in excess capacity and muted top line growth. Issuers will also need to successfully integrate AI into their operations to achieve cost savings and efficiencies.

Table 3

Related Research

• The Rise Of AI-Powered Smart Cities, May 18, 2024
• Language Modeling: The Fundamentals, Jan. 23, 2024
• Can Generative AI Create A Productivity Boom? Jan. 10, 2024
• Machine Learning: The Fundamentals, Nov. 29, 2023
• Foundation Models Powering Generative AI: The Fundamentals, Nov. 29, 2023
• Metaverse And Generative AI: Envisioning The Future Of Human-Computer Interaction, Nov. 7, 2023