“What gets solved in the U.S. or China might not work in Indonesia,” he told Fortune in early April, pointing to the country’s different culture and languages. That leaves the space open for companies like Indosat: “We’re in a pole position to see how we can deliver connectivity plus compute–or intelligence–to millions of people all over the world in a sovereign manner,” he continued.
Sinha is betting that the next phase of AI—running models near the end user, in local languages for local problems—will belong to telecom companies like Indosat in the so-called Global South. Indosat’s CEO, who came to Indonesia after working in India, the Seychelles, and Myanmar, is eager to drive that development through Sahabat AI, a platform for the country’s startups, underpinned by an Indonesian large language model that he argues will avoid the blind spots of a U.S.- or Chinese-trained model.
Still, even Sinha wondered whether he could turn “sovereignty” into a business. “If I ask my team whether they can make a business case for Sahabat? They don’t know how,” he admitted.
Sinha, born in Jamshedpur in eastern India, joined the telecoms business in 2005 with a job at Bharti Airtel. Seven years later, the company dispatched him, at just 37 years old, to lead its business in the Seychelles, a tiny island nation of just 120,000 people off Africa’s eastern coast. He then moved to another island nation, leading Ooredoo’s business in the Maldives, then went on to Myanmar, right as the Southeast Asian country was in the midst of its (ultimately short-lived) democratization and opening.
What Sinha recalled from his time in Myanmar is the average age of his team: 27 years old, all “young guys,” in his words. Yet he felt his time in the country was a rewarding experience. “When I was going to Myanmar, people warned me about the competency gap,” he said. “But if you invest in getting the best out of people, you see a lot of talent.”
“The number one guiding principle we wanted to follow was to look at the merger from a maximize, not optimize, outlook: How could we make one plus one equal 11?” Sinha explained. “When investors and analysts look at mergers, they only talk about synergies, but employees and customers don’t care about that. They care about growth and experience.”
Indosat reported 56.5 trillion Indonesian rupiah ($3.3 billion) in revenue for 2025, a 1.1% increase over the prior year, while profits climbed 12.2% to 5.5 trillion rupiah ($320 million). But those numbers mask a tough year: Sinha notes that the company’s performance was weaker in the first half of the year, only for things to turn around in the second half.
That strong performance has continued into the first quarter of 2026, with revenue jumping by 12.1% year-on-year. (Indosat released its Q1 earnings on April 29, after Fortune‘s conversation with Sinha). Indosat also achieved its highest average revenue per user (ARPU) since the merger, at 45,000 rupiah ($2.59).
“That mindset has to change. We have to build businesses around more sustainable models with things that are more real,” he added.
“Inferencing needs to happen close to the edge,” Sinha said, referring to the deployment of AI models close to end-users rather than in centralized data centers. “Telcos like us can take intelligence to the edge with low latency, and then develop applications which are made in a country, for the country, instead of just buying it from China or the U.S.”
Countries like Indonesia have structural advantages here that the West lacks, Sinha argued. “Countries like Indonesia have power, land and water. Today in Indonesia, I’m sitting with close to 800 megawatts of approved power,” he said. “The U.S. doesn’t have power.”
Still, he admits that raw infrastructure is not enough. “Without human capital, you will never become sovereign,” Sinha said. “Sovereignty is not only about investment or money.”
The case for a locally built model is straightforward, at least in principle. “The LLM is not neutral. And if it is not in your language, it will have bias, cultural nuances and all,” Sinha said. “Every country will focus on protecting the data and cultural sovereignty.”
There’s a practical reason too, beyond the principled one. “The ability for governments, banks, and other regulated entities to use AI relies on accuracy,” explains Pak-Sun Ting, cofounder of Votee AI, a Hong Kong-based startup that has developed an LLM that operates in the Chinese dialect of Cantonese. “If you don’t have accuracy, and people are speaking in a language that models don’t understand, then you don’t have a use case.”
But it’s challenging to build models in languages that don’t boast a lot of material. “Low resource, by definition, means there’s not enough data to build a large language model with natural language processing,” Ting says. The designation has nothing to do with the number of speakers—Bahasa Indonesia is spoken by nearly 300 million people—but with the volume of digitized text, which is small for most languages.
For Sinha, Sahabat looks a bit more like a public service than a business, at least initially. He called it a “platform to innovate and collaborate,” helping to bolster Indonesia’s new AI startups. “We have not promoted this in a manner where we are driving daily and monthly users,” he said.
“We are very confident that a business case will emerge. But yes, there will be doubts in the early days,” he admitted. “You have to go all in and make sure you believe in it.”
