Can telcos capitalize on the cloud and AI infrastructure building boom?

Artificial intelligence has sparked a cloud computing infrastructure goldrush that most telcos are not taking part in. But what will the new national resources mean for their business in markets where these investments are being made?

For the most part, telecom operators are not joining this spending spree, with notable exceptions including SK Telecom, NTT, and Reliance Jio. Some operators, such as Orange and Swisscom, have made AI-related infrastructure investments to offer “sovereign AI infrastructure” services, according to Chris Silberberg, Research Manager, EMEA Telecommunications Insights at IDC.

But overall, telecom capital expenditure is in decline.

“Buying GPUs is expensive, so it isn’t feasible for many telcos to put this as an investment priority above the significant investments they are already making in their fixed and mobile networks,” said Silberberg.

Deep-pocketed hyperscalers are driving the infrastructure boom, having pledged tens of billions in just the first half of this year to build up data center and computing capacity to support AI around the world.

Indeed, Dell’Oro expects hyperscale cloud capital expenditure to grow 17% in 2024 driven by investments in AI and accelerated computing. Meta, for one, has increased its expected capex range for 2024 to $35 - $40 billion as it ramps up infrastructure spending to support AI. Meta’s capex for full-year 2023 was $28.1 billion.

The recent flurry of investment news shows how spending is on the rise.

Microsoft announced plans to invest €4 billion ($4.3 billion) on cloud and AI infrastructure, AI skills training, and accelerating startups in France. The infrastructure part of the plan includes deploying up to 25,000 graphics processing units (GPUs) by 2025, expanding its data center footprints in Paris and Marseille and building a new data center campus in the Grand Est region in the northeast of the country.

This followed commitments from Microsoft to invest $2.9 billion in hyperscale cloud computing and AI infrastructure in Japan; $3.3 billion to expand U.S. cloud and AI capacity with a new data center campus in Wisconsin; $2.1 billion in Spain by the end of 2025 and €3.3 billion in Germany in the next two years.

AWS is also expanding in France with a pledge to invest an additional €1.2 billion in the country. In Spain, AWS committed €15.7 billion to expand its cloud infrastructure in Aragón for its Europe Region. The fresh injection of funds dwarfs the €2.5 billion that the hyperscaler had originally planned in 2021. Telefónica is among the AWS enterprise customers that use the Aragón facility.

AWS also committed to invest S$12 billion ($9 billion) in its cloud infrastructure in Singapore between 2024 and 2028 as part of its new AWS AI Spring Singapore programme aimed at supporting the country’s Smart Nation and National AI Strategy 2.0 goals. The intent of the AWS project is to “fast-track” AI and generative AI (GenAI) adoption, working with the Singapore government, public sector and enterprises.

In Japan, AWS announced in January that it would invest ¥2.3 trillion ($14.3 billion) by 2027 in its cloud infrastructure in Tokyo and Osaka, bringing its total capex investment in the country to ¥3.8 trillion since 2011.

In the same month, Google announced that it would spend $1 billion to build a new data center on a 33-acre site in the UK. Google has also pledged to invest $2 billion in Malaysia to build its first data center and Google Cloud region.

Where does the AI infra boom leave telcos?

According to Silberberg, the telcos that can afford to build sovereign AI infrastructure offerings could see a revenue boost “with a clear and simple business case,” although it is difficult to know the size of the opportunity at this point.

“For other telcos, the route to monetization is much trickier and runs through a partner-driven approach that offers targeted AI services to businesses,” he said.

“The telco could supply the preferred local compute and connectivity infrastructure with some additional security services, but then systems integrator, application and data center or cloud partners would supply additional central compute, application and implementation services,” he explained.

A recent report from analyst firm Omdia advises communication service providers (CSPs) to “approach cloud as an opportunity to reimagine their roles in the technology ecosystem” and that partnering with cloud providers opens new business models.

An example of these models includes marketplaces where CSPs buy and sell services or provide a channel to market. Others are in B2B opportunities, where hyperscalers and telcos co-invest in new services for enterprises or where their complementary services are sold “side-by-side.”

As users of cloud services, telcos and other enterprises, can benefit from having availability of AI and cloud services in-country and closer to their operations. This helps with data sovereignty requirements and regulatory compliance for data storage while reduced latencies can improve customer experience, explained Roy Chua, Founder and Principal at AvidThink.

Telcos use public cloud for various IT workloads. Some are starting to move network workloads to public cloud environments, such as Telefónica Germany’s deployment with Nokia of 5G standalone core software on AWS.

“The expanded data center footprint offers cloud-enabled telcos more options for workload placement, considering factors like cost, proximity, compliance, and resiliency,” said Chua.

There are also possible opportunities to “provide connectivity to and across these data centers,” he said, but noted that hyperscalers often use their own infrastructure.

“Cloud on-ramps provided by regional or local telcos and inter-exchange carriers may benefit from increased enterprise and consumer demand for connectivity to these data centers,” he said.

Operators are also looking to AI and automation to drive down operating costs. Chua said that the AI computing capabilities will help operators with using AI, machine learning, and GenAI for their own operations or for developing new services.

“As major hyperscalers compete for dominance in the AI market, telcos are likely to become more dependent on them,” he said. However, he expects operators to have partnerships with multiple hyperscalers and AI model builders “to avoid single-vendor lock-in.”

“Considering the crucial role of data in AI workflows (both GenAI and other types), telcos may be compelled to move increasing amounts of data into hyperscaler clouds for convenience. However, telcos will also likely push for hyperscalers to accommodate on-premises options, which all hyperscalers currently offer to varying degrees,” he said.