FEATURE
The second pillar is a focus on becoming an‘ AI techco’ – a concept which defines the urgent evolution of traditional telcos into technologydriven, AI-native platforms. We want to create a neo-cloud platform for the region, supporting sovereign workloads for Indonesia and providing strong security capabilities.
The third pillar is our role as a nation-shaper. As a major brand in Indonesia, expectations are high, but what drives us is a want to help build national capabilities. Today, the country is largely seen as a consumption market. With 280 million people, everyone wants to sell their applications here. Our goal is to help the country move from being a consumption market to becoming a creator of infrastructure and co-creation platforms. Our platform which supports this is called Sahabat.
I often reference the‘ five-layer cake’ model which is an idea popularised by Jensen Huang. Indonesia is uniquely positioned across these layers. At the foundation is energy: availability and price. Indonesia has surplus energy and very competitive pricing compared to other South- East Asian countries. Land, water and power are required to build this ecosystem.
The second layer is infrastructure, specifically data centres. We began this journey over a year ago and today we have 10 mW of live capacity, supporting everything from L40 GPUs to H100 systems and we were among the first to go live with Blackwell GB200. Our ambition is to scale to one gigawatt by 2030.
The third layer is chips. We have built a strong partnership with NVIDIA and were one of the first NVIDIA Cloud Partners in Indonesia. This partnership has helped us secure our supply chain during a time when GPUs, storage and networking components are in global shortage. However, all of this means nothing if it does not create value for Indonesia. That is where Sahabat plays a central role.
Initially, Sahabat was launched as a web platform when Jensen Huang visited Indonesia. Recently, we launched the Sahabat mobile app which required securing sovereign storage and infrastructure within Indonesia so training could happen locally.
We use H100 GPUs to train and fine-tune the Sahabat model and collaborate with partners like Google and Alibaba. Instead of investing heavily in building a foundational model from scratch, we build on top of existing models while training sovereign datasets and fine-tuning them to preserve linguistic and cultural nuances.
38
INTELLIGENT CIO APAC www. intelligentcio. com