Cloud cost is the number one worry for African CIOs. Here is a FinOps playbook that respects the exchange rate
Cloud cost worries 60% of African tech leaders. A practical FinOps playbook on AWS that accounts for currency risk, with eight levers and a monthly rhythm.
Cloud cost is the dominant concern for African technology leaders, drawing 60 per cent of responses in a poll of 400 fintech and enterprise leaders, far ahead of cybersecurity at 18 per cent and regulation at 13 per cent (Africa Hyperscalers, August 2026). The cause is not only usage; it is that AWS bills in US dollars while revenue arrives in shillings, naira, cedi or rand. A FinOps practice that works in Africa therefore has two jobs: reduce the dollar bill, and insulate the organisation from the exchange rate. This playbook covers both.
Why the currency makes this different
A CIO in London who overspends on cloud by 10 per cent has a 10 per cent problem. A CIO in Nairobi whose shilling weakens by 10 per cent against the dollar has the same problem without changing a single resource. Wall Street banks flagged the Kenya shilling as vulnerable in April 2026, and currency moves of that order are a recurring feature of most African markets.
The consequence is that an African cloud budget needs a currency assumption, a sensitivity analysis and a plan for the downside. Finance teams understand this instinctively because they manage it for every imported input. Technology teams often do not, because cloud was presented to them as “pay for what you use” rather than “pay in dollars for what you use”.
The good news is that the engineering levers that reduce cloud cost also reduce currency exposure, because they shrink the dollar number that the exchange rate multiplies. The commercial levers then handle what is left.
The eight levers
Step 1: Rightsizing
Most AWS estates we audit carry 20 to 40 per cent of compute that is oversized for its actual load. Instances provisioned for a launch peak that never recurred, databases sized by guesswork, and development environments running at production scale. AWS Compute Optimizer and AWS Cost Explorer rightsizing recommendations give you the list. The work is in the change management, not the analysis.
Start with non-production. Nobody objects to shrinking a test environment, and the savings fund the appetite for the harder production changes.
Step 2: Savings Plans and Reserved Instances
Once an estate is rightsized, commit to it. Compute Savings Plans give a substantial discount on Amazon EC2, AWS Fargate and AWS Lambda in exchange for a one or three year hourly commitment. The commitment is in dollars, which cuts both ways: it locks in a lower dollar rate, but it is a dollar obligation regardless of what the shilling does.
Our guidance is to commit to the stable baseline, typically 60 to 70 per cent of steady-state compute, on a one-year term, and leave the rest on demand. Three-year terms suit organisations with dollar revenue or a hedged currency position; for a shilling-revenue business they are a bigger bet than the discount justifies.
Step 3: Graviton
AWS Graviton processors offer better price-performance than comparable x86 instances for most workloads. Moving container workloads, Java services, databases on Amazon RDS and managed services such as Amazon OpenSearch Service to Graviton is often a configuration change rather than a migration. For a containerised estate on Amazon EKS it can be done node group by node group with no downtime.
Step 4: Storage tiering
Data accumulates and nobody deletes it. Amazon S3 Intelligent-Tiering moves objects between access tiers automatically based on usage. Lifecycle policies push logs, backups and historical data to S3 Glacier classes. Amazon EBS volumes left attached to stopped instances, and snapshots nobody reviewed, are a reliably recoverable few per cent of most bills.
Step 5: Serverless where the load is spiky
Serverless shifts cost from capacity to usage, which suits workloads with daily or seasonal peaks. Yellow Card reported a 40 to 50 per cent operational expenditure reduction after moving to serverless architectures on AWS. The caveat is that steady, high-throughput workloads can cost more on AWS Lambda than on reserved compute, so the decision is per workload rather than a blanket policy.
Step 6: Containers and consolidation
Africa’s Talking cut infrastructure cost by 50 per cent by moving to containers on Amazon EKS. The mechanism is bin-packing: dozens of underutilised virtual machines become a smaller pool of well-utilised nodes, with autoscaling matching capacity to demand. Combined with Graviton nodes and Savings Plans on the baseline, container consolidation is frequently the single largest saving available to an enterprise with a legacy VM estate.
Step 7: Consolidated billing through an AWS reseller with local-currency invoicing
This is the lever that addresses the exchange rate directly. Buying AWS through an authorised AWS reseller in your market gives you consolidated billing across accounts, an invoice in local currency from a local entity, and, depending on the arrangement, a fixed or periodically reset exchange rate.
The benefits for finance are material:
- One invoice in shillings, payable locally, rather than a dollar card payment or an international wire.
- Withholding tax and VAT handled by a local supplier.
- Exchange rate exposure moved partly or wholly to the reseller, who can hedge at scale.
- Procurement through a local contract, which matters for public entities and regulated institutions.
The reseller arrangement does not change what you build or how you operate your accounts. It changes who sends the invoice and in what currency.
Step 8: Tagging, showback and a monthly FinOps review
None of the above sticks without an operating rhythm. The foundation is a tagging standard: every resource carries an owner, a cost centre, an environment and an application. AWS Organizations tag policies enforce it. Cost allocation tags feed AWS Cost Explorer and AWS Budgets.
Showback then puts each team’s spend in front of them monthly, in both dollars and local currency. Teams that see their bill behave differently from teams that do not.
The monthly review is a one-hour meeting with finance, engineering leads and whoever owns the AWS relationship. The agenda is fixed:
- Spend versus budget, in dollars and in local currency, with the exchange rate used.
- Top ten cost movers since last month and why.
- Rightsizing and tiering recommendations accepted, rejected or deferred.
- Commitment coverage and utilisation for Savings Plans.
- Anomalies flagged by AWS Cost Anomaly Detection.
- Decisions and owners for the next month.
How the levers combine
| Lever | Typical saving on affected spend | Currency effect | Effort |
|---|---|---|---|
| Rightsizing | 20 to 40 per cent of compute | Reduces dollar base | Low to medium |
| Savings Plans | Up to the published discount on committed compute | Locks dollar rate, adds dollar obligation | Low |
| Graviton | Better price-performance per instance | Reduces dollar base | Low to medium |
| Storage tiering | A few per cent of total bill | Reduces dollar base | Low |
| Serverless | Up to 40 to 50 per cent opex on suitable workloads | Reduces dollar base | High |
| Containers | Up to 50 per cent on legacy VM estates | Reduces dollar base | High |
| Reseller billing | No direct AWS discount; removes forex friction | Moves exposure to reseller | Low |
| FinOps rhythm | Sustains all of the above | Makes forex visible monthly | Ongoing |
The percentages for serverless and containers are the published figures from Yellow Card and Africa’s Talking respectively; results for other organisations depend on the starting point.
A sequence that works
Month one: tagging standard, consolidated billing, Cost Explorer and Budgets with alarms, first showback report. Month two: rightsizing of non-production, storage lifecycle policies, Graviton on the easiest workloads. Month three: rightsizing of production, first Savings Plan commitment on the baseline, FinOps review established as a standing meeting. Months four to twelve: container consolidation or serverless re-architecture for the workloads where the business case is clear.
What to do next
If cloud cost is on your board agenda, the first step is a cost review that shows what you spend, in both currencies, by team and by lever. It takes a week and it reframes the conversation from “cloud is expensive” to “here are the eight things we can do about it and what each is worth”.
Shujaa.ai is an AWS Advanced Tier Services Partner headquartered in Nairobi. We run FinOps reviews, implement cost optimisation on AWS and provide reseller billing with local-currency invoicing for East African organisations. See our DevOps and CloudOps services.
AWS Solutions Architects, Nairobi. Shujaa.ai is an AWS Advanced Tier Services Partner headquartered in Nairobi.
