AI Triggers a Surge in Data Network Infrastructure Needs
Artificial intelligence (AI) is driving a major transformation of data network infrastructure. The surge in inter-data-centre bandwidth requirements affects not only users but also corporations and the interconnects between data centres.
A recent study from Ciena, the US-based provider of network services and software, indicates that global data-centre experts anticipate significant growth in data connectivity needs over the next five years. As cited by BloombergTechnoZ, in a survey of more than 1,300 data-centre professionals across 13 countries, more than half (53%) believe that AI workloads will place the greatest strain on data infrastructure in the next two to three years, ahead of cloud computing (51%) and big data analytics (44%).
Bandwidth demand is also forecast to rise sharply as AI model training and inference require moving large volumes of data. A total of 87% of respondents said they would require fibre capacity of 800 Gb/s or more per wavelength for Data Centre Interconnect (DCI).
AI workloads are reshaping the data-centre industry, from infrastructure to bandwidth demand. AI will also accelerate network traffic growth, forcing operators to adjust their infrastructure strategies.
The Ciena survey also found that as AI computing needs rise, training of large language models will be increasingly distributed across multiple data-centre facilities. 81% of respondents believe AI training will be performed at a number of data-centre facilities.
In Indonesia, an average of 49% of new data centres are expected to be dedicated to AI workloads. AI applications are believed to drive bandwidth increases of at least sixfold over the next five years.
Power availability, cost efficiency, scalability, and reliability are the main factors in selecting new data-centre locations in Indonesia. This shows that AI growth in the country will depend heavily on the readiness of the network infrastructure.
Additionally, 50% of respondents believe that AI applications will drive bandwidth increases of at least sixfold on DCI networks in the next five years.
Yudo Satrio, Head of Network Planning at FiberStar, Jakarta, said on Wednesday (4/3) that the company, which provides optical network infrastructure for residential, commercial, and corporate sectors, observed in 2016 the data capacity they distributed reached 20 giga, rising in 2020 to 8 tera, and jumping in 2026 to 19 tera. By 2027, it is expected to have doubled to 30 tera.
“AI usage in Indonesia will continue to grow and rise sharply. From the consumer side, it will expand to various sectors, so this will require adequate data traffic support,” said Yudo.
Using fibre-optic infrastructure on land and sea stretching from Jakarta to Surabaya, Yudo noted the network length has reached 64,497,562 metres, or about 64,498 km.
During 2025, the number of homepasses reached 3,360,119 points, up from 2,303,553 in 2024, a YoY increase of 45.9%. At the same time, the number of home-connected customers rose to 489,211, from 322,150 in 2024, a YoY increase of 51.9%, faster than homepass growth. In 2026, around 750,000 new homepasses are targeted.
Homepass is a term in the internet service provider industry referring to homes or buildings already reachable by fibre optic networks, thus potential for immediate internet subscription. This indicates the footprint of infrastructure, not the number of active subscribers.
Recent data shows that worker preferences and corporate practices remain skewed towards flexibility. Remote and hybrid work remain relevant.
Telkom’s steps align with global telecoms industry trends, where the separation of infrastructure (InfraCo) from retail services (ServiceCo) is increasingly being adopted.
Indonesia aims to build a Southeast Asia data centre hub by 2030 valued at US$365 billion. Regulation, green energy, and infrastructure are key accelerators.
Through training, workshops, and competitions for students and teachers, the necessary skills are being provided to become competent digital talents.