Indosat: Building an AI-Native Data Centre Strategy

Indosat is reshaping its infrastructure strategy around AI-driven data centres, sovereign cloud and distributed compute, positioning itself as a national digital platform for Indonesia.

Speaking at MWC in Barcelona, President Director and CEO Vikram Sinha outlined how the company is transitioning from a traditional telecommunications operator into what it describes as an AI-native organisation.

Central to this shift is the development of GPU-powered data centre capacity and a broader AI infrastructure ecosystem.

The company already serves close to 200 million customers across Indonesia’s 17,000 islands.

Its next phase focuses on embedding AI across both network operations and digital services, supported by large-scale infrastructure investment.

“We mean it when we say our larger purpose is empowering Indonesia,” says Vikram Sinha, President Director and CEO of Indosat.

“Indonesia is on a journey of becoming a developed nation. Research reports show that AI can be a great enabler. In the early days, all the talk was about building infrastructure – roads and highways – but now it is all about building digital infrastructure. This is mission-critical for a country like Indonesia.”

Scaling AI data centre capacity

At the core of Indosat’s strategy is a significant expansion of data centre infrastructure to support AI workloads.

The company is deploying GPU-powered facilities starting with 10MW of live capacity, with plans to scale to as much as 1GW by 2030.

These data centres are designed to host AI training and inference workloads for both domestic and regional demand.

Indonesia’s access to relatively low-cost power, land and water is seen as a key advantage for attracting large-scale AI deployments.

This positions the country as a potential hub for regional AI infrastructure.

Rather than treating AI as a standalone business, Indosat is embedding it into core operations, using data-driven insights to improve service delivery and operational efficiency.

“If I am not performing well on my core business, I have no right to start a new business,” Vikram says.

Sovereign cloud and local AI models

Alongside data centre expansion, Indosat is developing a sovereign cloud platform supported by locally trained AI models.

At the centre of this effort is Sahabat AI, a large language model designed to understand Indonesia’s languages and cultural context. The model is trained on local data and deployed within Indosat’s infrastructure environment.

“We are not trying to compete with ChatGPT or Gemini,” Vikram explains.

“We are focused on making sure that it understands all languages. We want to make sure that it understands the cultural nuances and understands the real insights. We want to collaborate with ChatGPT and Perplexity, et cetera, at a certain level, but what we want to focus on is the sovereign sensitive data which it is getting trained on and all the local language cultural provision.”

The model is already being used in production environments, including a large-scale spam and scam detection system built in partnership with Tala.

Running on Indosat’s AI infrastructure, the system has blocked close to a billion unwanted interactions within six months.

“As a telco, our job is not only to connect, but also to protect,” Vikram says.

“This is the first use case where we are using AI to solve a real problem at scale. Giving core connectivity is no longer what our role is. If I'm giving you connectivity, I need to give you peace of mind that I'm giving you security, too.”

From centralised data centres to AI grids

Indosat is also extending its infrastructure strategy beyond centralised data centres through an AI grid model.

The company plans to evolve from a limited number of AI factories into a distributed network of up to 55,000 edge locations.

These sites will combine radio access network infrastructure with GPU compute, enabling AI workloads to run closer to end users.

Developed in partnership with Nokia and NVIDIA, this approach allows shared infrastructure between network and AI processing workloads.

“When we started this journey, it was not about creating more efficiency, but moving up from AI RAN to AI grid,” Vikram says.

“That has been the focus. From proof of concept, we are now in the place where we are getting ready to scale up.

“This is a once-in-a-lifetime opportunity for telcos like us to move from just doing connectivity to doing connectivity plus compute, which is intelligence, and to do it at the edge in a sovereign manner.”

Building talent and ecosystem support

Indosat’s infrastructure strategy is linked to broader ecosystem development, including skills and regional inclusion.

The company is involved in national AI initiatives such as the AI Center of Excellence and has established AI Experience Centres across multiple regions, including underserved areas.

“Our approach is AI for all and we are trying to be the great equaliser,” Vikram says.

“We’ve been on a journey in terms of democratising intelligence. We very strongly believe that AI is a great equaliser. We are looking at AI from a growth mindset – how it can empower humans, how humans can lead.”

“When I talk about AI for Indosat, that has been the approach – investing in talent, helping every employee unlock their full potential and unlock growth. The same is happening at a country level. This has been our journey,” he concludes.