When Telkomsel first migrated its critical applications to the public cloud, costs surged rather than dropped. The unexpected rise sparked tense debates between finance and engineering, highlighting a truth many enterprises discover only after committing to the cloud: Migration alone doesn’t guarantee efficiency.
“Four years ago, we migrated our most mission-critical and self-service applications to the public cloud. Costs went up instead. There were many discussions, and it caused quite a panic,” recalled Kristian Wahyu Adi Nugroho, Cloud Centre of Excellence Lead, Telkomsel.
Nugroho presented Telkomsel’s case study during Datadog DASH 2025 in New York City, explaining how the company addressed cloud challenges through FinOps and with the support of observability tools.
Legacy problems
After three decades of operations, Telkomsel now serves 97% of Indonesia’s population, operating 269,000 transceiver stations and covering 158.4 million mobile devices. Predominantly on-premises, the telco faced several hurdles in migrating to public cloud providers such as AWS.
The quickest way to migrate to the cloud, Nugroho noted, is to relocate and rehost. For a more efficient strategy, organisations can replatform or repurchase, but for long-term wins, rearchitecture and refactoring works best.
“Just because you can lift and shift, doesn’t mean you should,” he remarked.
Recognising that not all applications are suitable for migration, Telkomsel adopted a hybrid, polycloud approach.
Nugroho said cost and integration had to be considered carefully. Telkomsel has over 200 applications, but only 15 have been migrated to the cloud. The first step, he explained, is a cost-benefit analysis, followed by considerations such as data residency requirements.
“For example, we’ve invested heavily in the database, but if we migrate it directly to the cloud, it reduces the value of that investment,” he said.
The company also decided to house applications on separate cloud platforms, rather than spreading workloads across multiple clouds.
“For a polycloud approach, we believe one workload is best suited to a single cloud, because setting up an application to run across multiple clouds is also quite challenging. For example, our mission-critical applications are in AWS, while our productivity suite works best in Azure,” Nugroho said.
FinOps strategy
Effective cost management in the cloud requires moving beyond basic monitoring tools to advanced analytics that connect costs with performance and operational metrics. This visibility enables teams to optimise resources, improve application performance, and make more informed engineering decisions.
To prevent cloud bills from escalating, Telkomsel built its FinOps strategy using the Cloud Cost Management (CCM) capabilities in Datadog’s observability and security platform. According to Lutfi Ichsan Effendi, Cloud Engineer, Telkomsel, visibility into actual spending is crucial.
“You can include formulas for custom pricing. You can add enterprise discounts if there are any, and then see whether your cloud requirements are still within budget or not,” he said.
Effendi added that visibility into cost trend analysis enables organisations to proactively monitor whether workloads are increasing or decreasing.
At a deeper level, microservices cost visibility allows analysis by service, deployment, or even at the API level. This links costs directly to business metrics such as revenue or user engagement.
Similarly, adopting infrastructure as code ties changes in provisioning directly to cost impacts, giving developers real-time insights into financial implications.
Cloud considerations
To summarise, Nugroho highlighted five key takeaways from Telkomsel’s FinOps experience. The first is the need for clarity on ownership and accountability.
“Because FinOps spans across several teams — from the CCoE (cloud centre of excellence), operations, developers, and finance — we need clearer boundaries and accountability. Who owns the cost? Who owns the operations?” he said.
Second, organisations must have actionable tools, not just insight.
“Observability alone is not enough. You need to think about adopting analytics tools that can recommend and automate tasks and workloads,” Nugroho said.
Third, FinOps must be treated as a journey rather than a quick fix.
“You must ask yourselves: Where is your journey now, and what steps do you need to take for the ideal FinOps setup? It’s a long journey for us, and in this journey the hyperscalers are supposed to be our partners. We need to ensure we have a good partnership with them, so they can help us. Our partnership with Datadog has supported our FinOps journey,” he said.
Fourth, architecture has a direct impact on cost. Organisations must carefully evaluate before deciding which applications to migrate to the cloud.
Finally, culture is the real challenge. According to Nugroho, the toughest hurdle is not the tooling, but the mindset.
“We often have this mindset of just deploying and then leaving it, but in the cloud we cannot do that. This culture is quite challenging for us, convincing everybody to shift their mindset in the cloud world. We really have to embed cost awareness across the organisation,” he concluded.
