Cloud Repatriation: Why Enterprises Are Pulling Data Back On-Premise
Rising operating costs, unpredictable pricing, compliance with data sovereignty mandates and laws (such as DPDP Act) are driving organisations toward cloud repatriation – moving workloads and data from the public cloud back to on-premises or private infrastructure.
But leaving isn’t cheap, though. Cloud providers charge steep egress fees to move large volumes of data out of their networks, and those fees act as a financial barrier to exit that can lock customers in (see egress rates below).
“It costs five times more to pull data out of the cloud than putting it on cloud,” said Andrew Huang, Regional Sales Head, Synology.
During a recent media event, Huang said his customers who moved their data to the cloud are thinking about cloud repatriation for multiple reasons.
“With compliance concerns about the DPDP Act, people are [asking questions] about data sovereignty, data ownership, and the regional location of their data on the cloud. They are worried about cloud service providers synching their data with other regions. Some try to pull everything out of the cloud without realizing there is an egress cost; it costs five times more to pull data out of the cloud than putting it on cloud,” said Huang.
(L-R) Asta Liang, Country Manager – SAARC at Synology Inc. / Andrew Huang, Regional Sales Head, Synology (Center) / Davlen Tandersen, Regional Account Manager, Synology Inc.
Standard Egress Rates (per GB):
- Google Cloud: ~$0.12/GB (first 1 TB)
- AWS: ~$0.09/GB (first 10 TB)
- Microsoft Azure: ~$0.087/GB (first 5 TB)
- Oracle Cloud: ~$0.0085/GB (after free tier limits)
(Source: HBS)
Also see: 64% of Indian Organizations Made New Data Protection Purchases Following Security Incidents
Why Egress Costs Are High
- Asymmetric Pricing: Ingress (uploading or moving data into the cloud) is free across major providers, while egress (downloading or moving data out) is heavily metered per gigabyte. [Source: Oracle]
- High Margins for Providers: Compute margins are often single-digit, whereas network bandwidth and egress fees generate profit margins as high as 20% to 30%. [Source: Sedai]
- Scale of Penalties: Moving massive datasets – such as a petabyte of data – can incur one-time egress and network charges ranging from $74,000 to $75,000, not including private circuit or carrier charges. [Source: David Linchitum, LinkedIn post]
Local Inferencing
Another reason influencing cloud repatriation is local data processing and inferencing by private AI models and corporatev AI agents.
Keeping data nearer to on-premise models, often referred to as data gravity or local inference/processing, means executing machine learning, LLM inference, or data pipelines right where the raw data is generated or stored (in a private data center, edge device, or local server room) rather than shipping it across public networks to third-party cloud APIs.
“We recommend hybrid infrastructure to our customers. More important digital assets must be kept on-premise where AI models can read it. And use cloud when you need its flexibility,” said Huang.
