The global artificial intelligence race is typically framed as a bipolar struggle between the computing clusters of Silicon Valley and the state-backed labs of Beijing. For most nations, the choice is simple: adopt an American model like OpenAI’s GPT or a Chinese alternative like DeepSeek, and hope the cultural translation is close enough.
Vikram Sinha, the CEO of Indosat Ooredoo Hutchison (IOH), believes that “close enough” is a dangerous gamble. As the leader of Indonesia’s second-largest mobile carrier, Sinha is betting that the next frontier of AI isn’t about scale, but sovereignty. He is building Sahabat AI—a large language model (LLM) specifically designed for Indonesian languages and cultural nuances—on the premise that a model trained in San Francisco or Shanghai will always have blind spots when applied to the archipelago.
We see a bold strategic pivot for a telecommunications company. While most telcos are content to be the “dumb pipes” that carry data for Big Tech, Sinha is attempting to move up the value chain, integrating everything from the raw electricity and chips to the final application. However, the project comes with a candid admission: the business case for “cultural sovereignty” is still being written.
The Logic of Sovereign AI
The concept of “Sovereign AI” has moved from a niche government concern to a corporate imperative. For Sinha, the motivation is partly ideological and partly practical. LLMs are not neutral; they reflect the data they are fed. When a model is trained predominantly on English-language internet data, it inherits Western biases and linguistic structures that may not translate to Bahasa Indonesia or regional languages like Batak.
Beyond culture, there is the issue of accuracy. In highly regulated sectors—such as banking, healthcare, and government administration—hallucinations or linguistic misunderstandings aren’t just inconvenient; they are liabilities. This is where the “low resource” language problem becomes a barrier. In AI terms, a “low resource” language isn’t one spoken by few people—Bahasa Indonesia has nearly 300 million speakers—but one that lacks a massive volume of high-quality, digitized text for a model to learn from.
By partnering with the Indonesian tech giant GoTo to develop Sahabat AI (meaning “close friend” in Bahasa), Indosat is attempting to bridge this data gap. The goal is to create a tool that understands the specific vernacular and regulatory environment of Indonesia, providing a level of precision that generic global models cannot match.
Building the ‘AI Layer Cake’
Sinha is not just focusing on the software. He is implementing what Nvidia CEO Jensen Huang calls the “AI layer cake,” a full-stack approach to infrastructure. Rather than relying on distant cloud servers in the U.S., Indosat is pushing “inferencing”—the process of the AI generating a response—to the “edge,” closer to the end user.
To do this, Indosat has partnered with Nvidia to offer GPU-as-a-service, utilizing a cluster of H100 processors to provide on-demand computing power to local businesses in sectors like mining, and banking. Sinha argues that Indonesia possesses structural advantages that the West currently lacks: abundant land, water, and energy. With nearly 800 megawatts of approved power, Indosat is positioning itself to host the energy-hungry data centers required for local AI.
| Metric | Value (Approx.) | Trend/Note |
|---|---|---|
| 2025 Revenue | 56.5 Trillion Rupiah ($3.3B) | ↑ 1.1% YoY |
| 2025 Profit | 5.5 Trillion Rupiah ($320M) | ↑ 12.2% YoY |
| Q1 2026 Revenue | N/A | ↑ 12.1% YoY |
| Average Revenue Per User (ARPU) | 45,000 Rupiah ($2.59) | Post-merger high |
Growth Amidst a Tech Funk
The timing of this pivot is precarious. Indonesia’s tech sector is currently weathering a “valuation hangover.” For years, investors poured capital into “unicorn” startups chasing aggressive growth metrics over sustainable profits. That optimism has cooled, leaving many local startups struggling to find a viable path to profitability.
Even the broader market has been volatile, with the Jakarta Composite Index dropping 17% since the start of the year amid fears of a “frontier market” downgrade. Indosat’s own shares have dipped 9% in the same period, though they are significantly outperforming the local index.
Sinha’s ability to fund these AI ambitions stems from a rare success story in the telecom world: the merger between Indosat and Hutchison 3 Indonesia. While McKinsey estimates that 70% of mergers fail to deliver promised value, IOH has seen consistent growth in revenue and user base. Sinha attributes this to a “maximize, not optimize” philosophy, focusing on customer experience rather than just corporate synergies.
The Billion-Dollar Question: The Business Case
Despite the infrastructure and the patriotic appeal of sovereign AI, the path to monetization remains murky. Sinha is honest about the internal struggle to define the ROI for Sahabat AI. “If I ask my team whether they can make a business case for Sahabat? They don’t know how,” he admitted.
Currently, Sahabat AI functions more as a public utility or an innovation platform than a profit center. By offering it as an open-source tool for local startups, Indosat is essentially seeding an ecosystem. The bet is that by providing the foundational “intelligence,” Indosat will become the indispensable partner for every AI-driven business in the country.
The risk is that global giants like Google—with whom Indosat already partners to offer Gemini AI—could eventually close the cultural gap through better fine-tuning, rendering a standalone local model redundant. For now, Sinha is leaning into the belief that data residency and cultural sovereignty will remain top priorities for the Indonesian government and regulated industries.
Disclaimer: This article contains financial data and market analysis for informational purposes only and does not constitute investment advice.
The next critical milestone for Indosat’s AI strategy will be the rollout of more specific enterprise applications for Sahabat AI, as the company moves from a “platform for collaboration” to a tool with measurable daily active users and revenue streams.
Do you think “Sovereign AI” is a viable business model, or will global giants eventually dominate every market? Share your thoughts in the comments.
Related reading
