McKinsey & Company’s report on strategies for broadband providers in the age of artificial intelligence (AI) points to four ways in which the sector could benefit — but not before providing the context that these companies have struggled during the past decade.
The report, entitled “AI Infrastructure: A New Growth Avenue for Telco Operators,” starts with two stark assessments: While global mobile traffic has accelerated at an annual pace of 60% (from less than 1 exabyte in 2010 to more than 1,200 in 2023), broadband global revenue has grown by only 1% annually (plateauing at about $2 trillion annually in 2023).
Past is not necessarily prologue, however. The report says that broadband providers that move quickly can leverage their vast footprint and ability to manage large networks to meet the needs being driven by generative and agentic AI applications.
The meat of the report is an assessment of the potential benefits and obstacles to four future paths:
Connecting new data centers with fiber: The reality is that there will be a great need for more data centers. Many of these will be in cities where there currently are none. The global market, according to McKiney & Co., is expected to be $30 billion to $50 billion.
There are relatively favorable conditions for broadband providers to aid in data center creation in the U.S. because their existing footprints will reduce the need for permitting and construction that may slow down the process in Europe and Asia.
It seems that broadband providers may be more of a passive participant and help hyperscalers connect their data centers:
“Because hyperscalers have vast data center networks, catering to their needs will be critical. This means telcos can expect a preference for dark fiber, which allows customers to install and manage their networking equipment and optimize and scale bandwidth… Wavelength services on lit fiber may be desirable for temporary and backup lines and tertiary links in markets with a limited supply of dark fiber.”
Enabling high-performance cloud access with intelligent network services: Not only are there demands in terms of capacity, there also is more complexity and variance in the tasks these networks perform. Broadband providers, therefore, could benefit from the need for smart management.
“For telcos, this opportunity provides favorable conditions for turning the tide on a decade of diminishing B2B wireline revenue and establishing healthier, more sustainable B2B income streams as they shift from charging monthly usage and capacity to a more lucrative, value-based pricing model.”
The investments that are being made in software-defined wide-area networks (SD-WANs) will be beneficial, especially as the category continues to evolve. Questions that telecommunications must answer involve product vision and roadmap for intelligent networks? Should providers build or partner? What do the revenue models and monetization strategies look like?
Turning unused space and power into revenue: This path is straight-forward: Broadband providers tend to have unused data center and central office space and power that may help meet AI data center needs. They could provide these assets to hyperscalers, colocation providers and graphics processing unit (GPU) as a service and large enterprises either directly or via lease-back agreements.
The study says that new data centers can take half a decade to build. On top of that, there can be challenges in gaining access to the prodigious amounts of power that will be consumed once the switch is flipped. This gives incumbent broadband players an interesting choice: They can get into the AI compute value chain themselves or participate by “filling the void and capturing revenue from existing assets with minimal investment.”
Building a new GPUaaS business: Like all “as-a-service” offerings, GPUaaS offers customers an asset without the need to build it from scratch. In the case of GPUaaS, there are different models, such as by the hour (called spot contracts), longer arrangements (reserved contracts), and different use cases for the GPUs being offered.
McKinsey & Co. predicts that the GPUaaS market addressed by broadband companies for AI needs could range from $35 billion to $70 billion globally by 2030. Asia and North America are expected to provide the lion’s share of demand.
New business models are not easy. The report suggests that companies attempting to do so should be prepared. “Shifts in four key areas — sales strategies, partnerships, financial evaluations, and communications — are typically needed, with different paths sometimes requiring different operating models.”
The growth in data center demand is so massive it involves types of numbers that are not heard every day. CoBank Lead Economist of Digital Infrastructure Jeff Johnston talked through these numbers during a session at the NTCA–The Rural Broadband Association RTIME conference in San Antonio, Texas last month. Johnston said the average annual data demand increased from 16 Zettabytes in the period from 2011 to 2019 to 71 Zettabytes during the pandemic years of 2020 and 2021. It has grown to 124 Zettabytes since. This is expected to increase more than 200%, to 394 Zettabytes, by 2028. (A Zettabyte equals one billion Terabytes, or one sextillion bytes
