FutureNet Asia-Pacific 2025 – AI in the Telco Crosshairs

Recently, we had the opportunity to attend the FutureNet Asia-Pacific in Singapore, both as a delegate as well as a moderator. The theme for the event was all about Artificial Intelligence (AI) and its relevance and adoption by telcos. Here, we recap some of our key takeaways from the conference. 

The Telco Vision for AI is about a return to relevance and growth

The problems bedeviling the telecom industry are well documented and often discussed. One of the biggest refrains is the current lack of growth in the industry, with an alarming pattern where operators grapple with a scenario where the return on capital invested (ROIC) is trending lower than the weighted average cost of capital (WACC). The lack of growth is also happening at a critical time when the entire industry has been judo-flipped by much nimbler cloud-native players who have captured the higher value growth segments. 

AI is the latest paradigm shift that has caught the telco industry somewhat off guard by the sheer momentum generated by massive investments into large language models (LLMs) and related compute infrastructure. The AI flywheel effect is moving at such a fast pace that telcos are already playing catch up. The rules of the game have shifted at present towards compute and “AI-ready” infrastructure but there are still plenty of opportunities for telcos to tap into. 

The telco vision for AI adoption was in full evidence at FutureNet, with many leading Asian operators present to share their perspectives and insights. Every operator and operator group has their unique environment, network conditions and individual perspectives but there are some common threads on how to approach the AI opportunity. The core vision revolves around using AI to first “fix” the internal efficiency problems within operators while simultaneously building the right infrastructure and capacity to enable AI services to external audiences. Telco opportunities from AI are strongly correlated to a bet on a distributed computing paradigm emerging for AI. 

JFK’s Mission to the Moon as a Corollary for AI

In 1961, John F Kennedy, the US President, gave a clarion call for manned missions to the Moon. At the time, the US was playing catch up to the Soviet Union which had already launched Sputnik and jumped out to a huge lead in the space race. The US did not reach the Moon until the Apollo 11 mission in 1969, several years later. The point is not to portray this as an inordinate delay, but rather, to highlight that after such a bold vision statement, executing on the vision takes significant investment, effort, capacity building and time. As it turns out, years. 

While choosing analogies is fraught with imperfection, the telco industry needs a clarion call like the JFK moment cited above. In many ways, the clarion call has already rung out, approaching more of an industry chorus. But the focus has correctly shifted to the challenging work that has begun but so much remains to be done. 

AI for telcos must be about more than automation

At FutureNet, it was clear that while all operators will grapple with and implement AI in their networks as well as use their network to enable AI, there will be substantial variations and nuances in strategy. First, the common threads.

• An AI-native Telco simply can’t be a Telco wrapped up nicely in machine learning (ML). It must be about enabling business outcomes through automated orchestration.
• AI implementation must be cross-domain to be truly transformative.
• Moreover, AI must be able to leverage all the structural elements that have already been put into place, including cloud, SDN and more.
• To be truly effective, AI must be proactive, and not reactive.
• AI must improve the productivity of the telco employee base. 

For all the above threads to manifest, operators must invest in and implement a roadmap towards a higher degree of “AI-ready” maturity. There was plenty of talk about achieving the near mythical Level 4 autonomous network status under the framework developed by the TMForum by the end of the decade. In other words, operators will not opt for a big bang approach that touches on the core logic of their networks, but rather execute on AI incrementally, first with things like anomaly detection, then energy savings and so on. But beyond talk of Level 4, the clear import of many of the conversations was that AI adoption must be about more than automation in the network, it must enable business outcomes!

AI for telcos must be built on a bedrock of unified data 

At FutureNet, there was plenty of discussion beyond vision statements about how to implement AI and leverage the technology. One of the biggest themes focused on the urgent need to build a unified foundational layer of data that cuts across all telco domains. There are plenty who will view this as a dissonant theme, and cite the significant investments made by operators towards building a unified data architecture. For sure, there is no disputing the progress made. However, investments in data lakes without an ability to access "clean”, cross-domain data would render it into something more akin to a data swamp. 

Breaking down silos across domains is crucial, but more important for future AI-readiness is the recognition that AI is only as good as the data available. Operators have long dealt with a fragmented array of systems which have made data difficult to consume. To this end, data must be structured and cleaned into formats that are easily consumable. This effort is also important as an AI will not be able to make inferences about scenarios that it has not previously dealt with. Operators must use the opportunity to create a foundational data layer that is easily consumable or any AI that they implement will be tapping into a black box of sorts. For mobile operators, a bedrock of clean and unified data is the most critical foundational work they must complete before larger ambitions of implementing agentic AI and building AI factories.