The promise was irresistible. Pay only for what you use. Scale on demand. Turn capital expenditure into an operating line. Spend your engineering time on the product instead of the hardware.
For many organisations that promise was kept, and it is worth saying so plainly: public cloud changed the industry, and for startups, global products and fast experiments it is still hard to beat.
But a growing number of CIOs and CFOs are facing an uncomfortable reality. What began as a cost optimisation has become one of the largest and fastest-growing items in the IT budget — and one of the hardest to predict.
The invoice is not the estimate
Cloud decisions are usually made on the visible prices: compute, storage, a managed database. The bill arrives with everything else. Data transfer. Egress. API requests. Snapshots and retention. Load balancers. Monitoring and security tooling. Replication between regions.
Individually, each charge looks small. Together, they are why so many organisations find that the invoice looks nothing like the business case.
The cost of leaving
Egress deserves its own paragraph. Moving data into most public clouds is free or close to it. Moving it out is not. Every backup copied elsewhere, every second provider, every migration and every large dataset served to users carries a toll.
The effect is subtle and powerful: once data is in, it becomes expensive to move. That is not a technical property of the cloud. It is a commercial one.
Growth becomes someone else’s revenue
Infrastructure you owned was predictable. You knew the hardware, the maintenance and the support, and you could plan a budget around them. Consumption pricing swaps that for a bill driven by user activity, data growth, transfers and every new deployment.
That flexibility has value. It also means the better your business does, the more of it flows to your provider — and the less certain your finance team can be about next quarter. An entire discipline, FinOps, now exists largely to manage the complexity of the bill itself. When you need a team to understand the invoice, simplicity has become a project.
Most platforms are priced to grow with you. That is the problem.
Repatriation is not retreat
Many organisations are now moving selected workloads back out of public cloud. The word for it, repatriation, sounds like a reversal. It is better understood as maturity.
It is not a rejection of cloud. It is a decision to place each workload where it delivers the most value: steady, data-heavy and regulated systems on infrastructure with predictable economics; bursty, global and experimental work where elasticity earns its premium.
Why local alternatives are getting a second look
Local and private clouds are attractive for reasons that go beyond price:
- Predictability — clearer pricing, fewer line items, budgets finance teams can plan around.
- Control — your data residency, your governance, your access rules, visible rather than assumed.
- Proximity — lower latency for systems that mostly serve users in the same country, from banking to government services.
- Accountability — support from engineers in your time zone who know your environment, rather than an escalation path across continents.
For many finance teams, predictability is now worth more than unlimited elasticity.
The real question is ownership
Underneath the cost debate is a bigger one. The industry settled on a trade: you can have a cloud that works, or a cloud that is yours. Speed and modern services came with a condition — hand over the estate, the data gravity and the terms of the relationship.
That trade is no longer the only option. It is now possible to run a private cloud with the self-service experience, catalogue and automation people expect from public cloud, on infrastructure you control, without taking on a hyperscaler’s engineering overhead.
That is what Oasis Enterprise Cloud is for. It runs in your data centre or in a colocation facility you choose, on x86 or ARM hardware from the vendor you already buy from, with no lock-in to any infrastructure vendor. Business units run as isolated tenants with showback and chargeback. And the platform is licensed at a flat fee per deployment, banded by scale — so growth in your estate is not a growth in your platform bill.
Place workloads, not loyalty
None of this means leaving public cloud entirely. The strongest strategies are built on placement rather than platform loyalty. Ask, workload by workload:
- Does it need predictable cost more than elastic scale?
- Does it carry residency or regulatory obligations?
- How much data does it move, and what does moving it cost?
- Does it serve a global audience, or a local one?
- If it had to move tomorrow, could it?
And when you evaluate a local alternative, hold it to the same standard: transparent pricing with no surprise charges, clear answers on who operates it and where, local support you can reach, and the services you actually need — compute, Kubernetes, managed databases, object storage, backup and disaster recovery, security and monitoring. Predictability should not cost you capability.
The next phase of cloud will be defined by optimisation, not migration. The organisations that do best will be the ones that decided, deliberately, which parts of their cloud they want to own.