Sahabat-AI and the AI Center of Excellence: Indonesian-language AI on NVIDIA Blackwell
Sahabat-AI ("AI Friend") is a collection of open-source Indonesian-language models built on NVIDIA NeMo and NIM microservices, backed by the Indosat-NVIDIA-Cisco AI Center of Excellence running an NVIDIA Blackwell AI factory.
Key takeaways
- Sahabat-AI is a collection of open-source Indonesian-language models built on NVIDIA NeMo and NIM microservices.
- It's designed to support Indonesia's 277M+ speakers across government, industry, universities, and research.
- The Indosat-NVIDIA-Cisco AI Center of Excellence, launched July 2025, runs an AI factory on NVIDIA Blackwell GPU reference architectures.
- NVIDIA founder and CEO Jensen Huang: "Artificial intelligence will democratize technology. It is the great equalizer."
Why Indonesian-language models
Most large language models are trained overwhelmingly on English and a handful of other high-resource languages, which leaves Bahasa Indonesia and Indonesia's many regional languages underserved — worse retrieval quality, worse instruction-following, and cultural context that doesn't transfer well from Western training data. Sahabat-AI ('AI Friend') is Indosat Ooredoo Hutchison's answer: a collection of open-source language models purpose-built for Indonesian language and context, built on NVIDIA NeMo and NIM microservices, targeting Indonesia's more than 277 million speakers across industry, government, universities, and research institutions.
IOH President/CEO Vikram Sinha described Sahabat-AI as embodying 'Indonesia's vision for digital sovereignty and inclusivity' — language that ties directly to the sovereign AI positioning behind GPU Merdeka and, later, Zankore. The initiative launched alongside backing from Indonesia's Ministry of State-Owned Enterprises, GoTo, and Accenture, and was framed by NVIDIA founder and CEO Jensen Huang around AI as 'the great equalizer.'
The AI Center of Excellence, on NVIDIA Blackwell
In July 2025, Indosat, NVIDIA, and Cisco launched the Indonesia AI Center of Excellence, which runs an AI factory built on NVIDIA Blackwell GPU reference architectures using NVIDIA Cloud Partner patterns. Where Sahabat-AI is the model layer, the AI Center of Excellence is the compute and networking layer underneath it — a concrete example of Indosat's NVIDIA Cloud Partner status turning into deployed Blackwell-generation infrastructure rather than a roadmap slide.
Together, the two initiatives are the clearest public evidence that Indosat's NVIDIA relationship covers the full stack — models tuned for the local language, running on GPU infrastructure hosted and operated inside Indonesia — which is the same full-stack pitch behind the larger Zankore platform announced in 2026.