Indosat’s ‘token factory’ platform built together with F5, is a much needed development given the telco industry’s decades of expanding networks and squeezing costs.
According to F5’s SVP for product management in the service provider segment, Ahmed Guetari, a token factory offering is “part of lateral evolution, but you have a few exceptions like (Indosat CEO) Vikram Sinha, who jumps ahead and takes a risk.”
“Everyone else started with having the simplest business model – buying raw compute and renting it out with thin margins.”
What Indonesia’s second largest telco has done with F5 is build a platform that delivers AI services with flexibility, transparency, governance, and sovereignty control, while addressing the complexities of managing a workforce where agents potentially outnumber humans.
An opportunity to leapfrog
As demand for artificial intelligence grows, telecom investments could become a foundation for a new role – providing not just connectivity, but the computing infrastructure and services businesses need to deploy AI.
That opportunity is real, but it will take more than buying GPUs.
Ahmed described how early players in the “Neo Cloud” market built businesses around a straightforward model: secure access to power, buy GPUs, and rent out computing capacity. That approach helped meet an immediate need, but Ahmed pointed out that raw compute is only the starting point for the telco industry.
“It’s necessary. It’s not enough,” he stated.
Modernisation, please
Enterprise customers also need governance, security, flexibility and observability. They need tools to investigate prompts, trace misinformation and respond to risks such as prompt injection. Ahmed said, “To serve enterprise, you need to have a certain capability, what we call enterprise grade. You need to have the tools to go and do forensic (for example).”
For telcos, reaching that next stage means developing capabilities that many have yet to master. AI services depend on cloud technologies such as virtualisation, containers, multi-tenancy, and orchestration. These tools allow computing resources to be allocated among many customers and managed quickly as demand changes.
“You need to have the right software stack that was inherited from the cloud environment, and not all telcos mastered the cloud (technology). All of them tried, but very few (have the) skill set to have the cloud stack, so that’s the journey that telcos need to go through.
“Not all of them have.”
On top of that, tools to manage all this are manual, legacy, and mostly too slow, when AI requires orchestration capability that moves at the speed of request.
Too little, too late?
Telcos have an advantage that gives them “all the assets to win,” including reach, edge locations, real estate and enterprise-customer access.
“It’s not too late,” Ahmed said, arguing that AI adoption remains at an early stage.
His conclusion is optimistic – telcos have assets that could help them compete in the AI infrastructure market. Whether those assets become an advantage depends on how quickly operators can modernise – and whether they choose to compete on value, not just price.
