Singapore’s ComfortDelGro is preparing to move selected applications from public cloud services back to its own private cloud. Its upgraded VMware environment supports production containers, with applications on Azure being considered first and AWS workloads following later. The transport group also wants to run most of its AI workloads internally.
The strategy brings attention to an issue that can be obscured by the language of cloud adoption: where a workload runs should depend on its requirements and continuing costs. A decision made when an application was launched may need another look as usage, data volumes and operating priorities change.
The work behind repatriation
Moving an application is more complicated than copying its files to another server. It may depend on managed databases, identity services, networking arrangements and monitoring tools supplied by a cloud platform. Those connections have to be understood before the application can work reliably elsewhere.
A private environment also needs people to maintain it. Hardware capacity, security patches, backup procedures and recovery testing become part of the operating model. A comparison based only on a public cloud invoice can overlook these responsibilities and make a migration look simpler than it is.
AI introduces another layer of planning. Some workloads are experimental and irregular; others run repeatedly against sensitive internal information. These patterns can produce very different requirements for computing capacity, data movement and access control. Treating every AI project as the same type of workload makes cost planning less useful.
ComfortDelGro’s approach concerns selected applications and an evolving infrastructure strategy. It should not be read as proof that public cloud services have lost their value. The more useful question is which environment fits a particular application, and whether the organisation can operate that environment well.
A practical comparison for Bangladesh
Bangladeshi companies considering a similar change can begin with an inventory of their applications and dependencies. They should compare full operating costs, including staff, support, resilience and the time required to make changes. Expected growth matters as much as current usage.
Data control is another practical consideration. Keeping information in an internal environment may help meet a company’s requirements, but location alone does not provide security. Permissions, logging and recovery procedures still need attention.
The Singapore example shows why infrastructure choices deserve periodic review. A useful cloud strategy explains where applications belong, how costs will be measured and who will keep them running. That gives businesses a clearer basis for investment than a blanket commitment to one deployment model.
