Cloud adoption in South Africa accelerated sharply over the past five years. The arrival of Azure South Africa North, GCP's africa-south1 region in Johannesburg, and the broader pressure of the pandemic-era remote work shift pushed organisations into the cloud faster than most were ready for.
The result: most have cloud infrastructure. Far fewer have cloud discipline.
If your organisation's cloud bill has grown faster than your business, you are not alone. And if nobody in your team can clearly explain why it costs what it costs, that is a problem with a solution.
The three reasons SA organisations overpay
1. Lift-and-shift without optimisation
The fastest path to cloud is also the most expensive one long-term: take what you have on-premises, move it to virtual machines in the cloud, and call it a migration.
This is not cloud modernisation. It is data centre outsourcing at a premium. You trade capital expenditure for operational expenditure, gain some resilience, and lose almost all of the cost efficiency that cloud was supposed to deliver.
The workloads that actually benefit from cloud economics — auto-scaling, serverless compute, managed services, consumption-based pricing — require architectural rethinking, not just rehosting. Most SA organisations never made that investment because the migration project was declared complete the moment the lights came on.
What it costs: a workload running on a cloud VM that should be containerised or serverless can cost three to five times more than it needs to. Multiply that across a mid-size estate and you are looking at hundreds of thousands of rands in unnecessary monthly spend.
2. No governance, no visibility
Cloud spend is uniquely easy to lose control of. Unlike a capital budget that requires approval and a purchase order, cloud resources can be provisioned in minutes by anyone with the right credentials. A developer spins up a test environment. It runs for six months unnoticed. A data pipeline runs more frequently than it needs to. A storage bucket fills up with snapshots nobody reviews.
Most South African organisations we encounter have no cost allocation tagging strategy, no budget alerts, no regular spend review cadence, and no clear ownership of cloud cost at the team level. Finance sees one invoice. IT sees a console full of resources. Nobody connects the two.
The result is a monthly bill that is accepted rather than interrogated.
What it costs: untagged, ungoverned cloud environments typically carry 20–35% avoidable waste. On a R500,000 monthly bill, that is R100,000–R175,000 leaving the business with nothing to show for it.
3. Vendor-led decisions, not business-led ones
Cloud providers are commercially sophisticated. Their sales motions are designed to maximise consumption, not minimise cost. Committed use discounts (CUDs), reserved instances, and enterprise agreements all sound like savings vehicles — and they are, if you buy the right ones. If you commit to the wrong SKUs, the wrong regions, or the wrong scale, you are locked into spend that does not reflect your actual needs.
Many South African organisations entered cloud agreements based on vendor-proposed architectures and volume estimates that turned out to be wrong. Renegotiating mid-term is difficult. The right architecture conversation should happen before the commercial conversation — not after it.
The South Africa-specific context
Cloud economics in South Africa carry pressures that do not appear in global playbooks:
Currency exposure. Cloud bills are USD or EUR. When the rand weakens — which it does — your cloud spend in rand terms increases without any additional consumption. Organisations that did not model this into their cloud business cases have been absorbing significant unbudgeted cost increases.
Data residency requirements. POPIA and sector-specific regulations (particularly in financial services and health) require that certain data remain within South Africa. This limits architectural choices and, in some cases, forces premium pricing for compliant configurations. Understanding what must stay local versus what can be distributed is a meaningful cost lever.
Connectivity costs. Egress from South African cloud regions to end users is priced differently to global patterns. Architectures optimised for US or European traffic patterns may be significantly more expensive when applied to local usage profiles.
Loadshedding resilience. Multi-zone and multi-region architectures built for resilience against loadshedding add cost. Those costs are sometimes necessary — but they should be sized correctly for the actual risk profile, not built to enterprise-global standards by default.
What to do about it
Start with a cloud cost assessment
Before optimising anything, you need an honest baseline. A structured cloud cost assessment should answer:
- What are you actually running, and what is each workload costing?
- Which resources are idle, oversized, or unassigned?
- What is your tagging and cost allocation coverage?
- Where are your biggest avoidable costs?
This is not a complex or lengthy exercise — a focused two-week assessment of a mid-size cloud estate typically surfaces more than enough findings to justify the work.
Build a governance framework
Cost governance is not a one-time exercise. It requires:
- Tagging standards so every resource is attributed to an owner, environment, and purpose
- Budget alerts at team and service level — not just at the top of the account
- A regular spend review cadence (monthly at minimum, weekly for high-spend environments)
- Clear ownership of cloud cost at the engineering team level, not only in finance
This does not require expensive tooling. It requires discipline and a clear mandate from leadership.
Right-size and modernise strategically
Not every workload needs to be refactored immediately. Prioritise by cost impact: identify the five to ten workloads driving the most spend, assess whether they are appropriately sized and architected, and sequence the modernisation work accordingly.
Common quick wins: switching from always-on VMs to auto-scaling compute, moving appropriate workloads to managed services, decommissioning test environments that have been running for months, and reviewing storage tier assignments.
Think before you commit
Reserved instance and committed use discount decisions should be made against a stable, understood baseline — not against a first year's consumption that includes inefficiency. Commit to what you know you will use. Review commitments regularly. Never let a vendor commitment outpace your architectural roadmap.
The cost of doing nothing
Cloud costs compound. An ungoverned estate does not stabilise — it grows, as new projects are added, old ones are not decommissioned, and consumption increases without corresponding business value. The organisations that address this now, with a structured approach, will have materially lower cost structures in two years than those that continue accepting the bill as it arrives.
This is not a technical problem. It is a governance and discipline problem — and it has a straightforward solution.
CloudNala provides cloud cost assessments and governance frameworks for South African enterprises, public sector organisations, and high-growth businesses. If your cloud bill is growing faster than your business, get in touch.