Telecom operators to invest billions more in AI cloud infrastructure

Global spending on AI cloud infrastructure by telecom operators will more than double in the coming five years. According to a new report from Analysys Mason, investment in the sector is set to reach a total of $17 billion by 2025.

The telecoms sector is undergoing a transformative moment, in the UK and globally. Amid difficulties rolling out 5G services, rising costs and spiralling complaints, providers are looking to find new ways to improve customer attraction and retention – while in Britain, the merger of VodafoneThree means that just three major mobile network operators (MNOs) are left at the top of the market. If smaller challengers hope to gain market share in that environment, they will need to find ways to outflank their larger opponents, with agility and responsiveness to customer needs. 

All of this has contributed to AI becoming a cornerstone of telecoms operators’ strategic agendas. Champions of the new technology argue that it could elevate customer experience and strengthen brand loyalty, while also speeding up automation at scale, reducing costs and supporting new services and business models in the process. 

According to research from Analysys Mason, this could see telecoms operators investments in AI cloud infrastructure hit $17 billion by 2030. Forecasting a CAGR of 17% from 2024 until the end of the decade, the researchers believe investments aimed at powering both internal AI use cases and AI-related services for enterprise and consumer markets will power this new era of spending.

The report notes that customer care, marketing and productivity use cases have so far driven most of operators’ initial investments in AI cloud infrastructure. But looking ahead, next-generation services and business models, including ‘as-a-service’ offerings of network and compute infrastructure are being viewed as a way to expand revenues.

As-a-Service business models see product use offered as a subscription-based service, rather than as something outright owned and maintained by the customer. Originating from the infamous software-as-a-service (SaaS) concept that appeared in the 2010s with the advent of cloud computing, the template has expanded to many other sectors of the economy, where operators are hungry to find new ways to transform one-off purchases into continued streams of revenue.

AIaaS for telecom operators

Analysys Mason suggests that a sign of things to come in this regard can be found in the Asia-Pacific region. Network operators there are leading the charge in AI cloud infrastructure investment to support B2B and B2C services, including AI-as-a-service (AIaaS), as well as sovereign AI and large language model (LLM) training in local languages. 

This includes Chinese operators, particularly China Telecom and China Mobile, who are significantly scaling up their AI computing capacity through major investments. While access to advanced AI chips remains constrained by export controls, local chip alternatives are advancing rapidly, with Huawei emerging at the forefront of this effort – while DeepSeek recently proved it was possible to at least match the services of OpenAI, with significantly less expensive chips and energy demands.

These changes might also offer up opportunities for smaller telecoms operators. In the shape of DeepSeek, the emergence of smaller, domain-specific AI models may reshape AI cloud investment strategies, forging chances for more niche operators to deliver AI services in a less generalist manner – and leading to more intense competition in the AI cloud services market in the process.

However, as is still so often the case when discussing AI, Analysys Mason admits that much of this is still hypothetical. For all the hype around the potential of B2B offerings and as-a-service pivots in the telecoms sector, uncertainty remains over whether operators will successfully monetise their AI investments – of if “the current momentum could lead to a bubble”. 

The researchers conclude, “Success requires more than accumulating GPU capacity; it depends on building a strong developer ecosystem, delivering public cloud-like service experiences with automated, cost-efficient operations and complying with sovereignty requirements. While some operators may succeed independently in certain markets, others will likely need close collaboration with partners such as public cloud or GPUaaS providers. However, even if operators struggle to monetise AI through enterprise services, proven and substantial efficiency gains in internal operations use cases could still drive continued investment.”