Na serious matter we dey talk about cloud money for Africa. According to PwC 2025 Africa Cloud Business Survey, 88% of organisations for continent plan to increase their cloud budgets this year. But wait o, many of them no even sabi exactly where that money dey go. Na real paradox of modern cloud economics – the more you spend, the harder e dey to see wetin you dey pay for.
At the same time, Africa cloud maturity don accelerate well well. The PwC survey find say over 86% of organisations report medium or high cloud maturity for 2025, up from 61% just two years earlier. But wetin this progress reveal na new set of problems. The challenge no longer be how to enter cloud; e don become how to stop haemorrhaging money once you don enter.
And while the usual cloud spend suspects – idle resources, overprovisioning, shadow IT – don dey well-documented, the real waste no always be the obvious waste. E dey quieter, more structural and often plenty harder to fix. When you rent compute from hyperscaler, you no dey get bespoke environment. You dey choose from menu: small, medium, large. The problem be say most real-world workloads no dey map cleanly onto those options.
“If you need more than small but you no quite fill medium, you must take the medium,” says Marco Vieira, solutions architect at Nutanix. “You dey pay for resources you no go ever use.” At small scale, that gap dey tolerable. But once you dey run 50, 80 or 100 virtual machines, that wasted capacity on every single instance start to add up. This na when “micro waste” become serious cost burden.
Micro waste na new kind of problem for organisations. Deleting orphaned resources and switching off idle instances both important, but na reactive fixes. Micro waste, says Vieira, dey embedded in the architecture from the outset. Na something you no see for dashboard; e just dey quietly built into every invoice.
The answer, he say, na to right-size the total resource pool rather than try to fit individual workloads into predefined instance types. “We fit eliminate that micro waste by matching the actual resource requirement, not just picking the closest available size,” he say, adding that this na more precise approach to provisioning, and one wey pay off most at scale.
There’s related cloud cost problem wey equally common and equally invisible: peak provisioning. Businesses design workloads for worst-case scenario. End-of-month payroll runs, seasonal retail spikes, overnight batch jobs – the moments define the infrastructure, even if they represent only small fraction of actual operating time. “But for 25 days of the month, e dey run at half that capacity,” says Vieira. “You dey pay for the peak, continuously, whether you need am or not.”
Na point wey Ian Engelbrecht, field CTO at Veeam Software EMEA, echo. Teams prepare for peaks, rather than patterns. “If infrastructure dey permanently provisioned for those peaks, inefficiency dey embedded,” he say. E sound obvious when you think about am, but the fix na dynamic rightsizing and e rarely dey done properly.
This mean continuously adjusting resource allocation to match actual usage instead of locking in worst case capacity forever. Cloud platforms support this natively through tools like AWS Auto Scaling, Azure Monitor and Google Cloud Recommender, yet dem dey largely unused.
Cloud optimisation rarely come from one big decision. Duolingo na great example of how e actually work for practice. Back for 2024, the language learning app, wey get more than 500mn users, ask its engineering team to reduce AWS spend without breaking anything. The cloud bill dey high and the margin for error dey slim.
Wetin the team find when dem start look na, for hindsight, predictable: storage nobody need, instances bigger than the workloads require, and databases wey become costly because nobody review dem for while. Working through am methodically, dem migrate caches and databases onto more cost-efficient processor architecture, rethink their EC2 instance pool and move individual microservices, including the Python monolith wey power much of the core product, onto better-suited infrastructure.
“At first, we dey small scared,” says Jean-Paul Bonny, senior software engineer at Duolingo, “but that just speak to the efficiency of moving to better architecture.” The savings add up across the stack, with 25% reduction in caching costs, 50% drop in database spend within three months and 10% reduction in compute costs, all without changing single line of application code. “Na cost savings win and reliability win come together,” says Bonny.
Wetin both micro waste and dynamic rightsizing get common be say neither show up if cloud cost optimisation dey treated as one-time exercise. You no fit simply audit the environment, cut the waste and move on. “E be about continuously reviewing, just like you go do for your own environment,” says Vieira. “The review and optimisation process need to be continuous thing.”
The logic dey straightforward – workloads change and business requirements shift – so service wey right-sized for January fit be badly over-provisioned by June, or vice versa. Without continuous loop of monitoring, analysis and adjustment, even well-optimised environment drift back towards waste.
Nutanix Cloud Manager, na design around this idea. E run ongoing reports wey flag which workloads dey sit for right environment and which should move, repriced or retired. “E go tell you exactly which workloads should be where, for the right reasons,” says Vieira. “Maybe e be time to move something from on-prem to cloud because e go run more optimally there. Or maybe get workloads for cloud wey should come back in-house.”
The point here be say not every workload get value, yet every workload get cost. With cloud costs, visibility need come before everything else because you no fit optimise wetin you no see. “Accountability work best when e dey visible,” says Engelbrecht. “The aim no be only protection, but informed control.”
Visibility without action na just dashboard. Wetin organisations need, says Engelbrecht, na governance structure wey make acting on wetin dem see the default, not the exception. One of the most common myths about cloud be say e dey inherently cheaper than on-premises infrastructure. E fit be, but e no always be. Flexera 2025 State of the Cloud Report find say organisations still dey overshoot cloud budgets by average of 17%.
“The total cost of ownership na really wetin need to dey considered,” says Vieira. “Not just the front-end cost per hour or per day, but everything wey dey enter am – skills, risk, performance, availability, business priorities.” The calculation change depending on the workload, the organisation and the time horizon. TCO model, for example, look very different from monthly bill comparison.
“For us, e no be about whether e dey on-prem or in-cloud,” says Vieira. “E be about the right fit for the customer.” AI dey become one of the most significant cloud cost drivers and unlike the waste problems wey come before am, e dey arrive before organisations get any idea wetin dem dey spend or why.
GPU instances, API calls and SaaS subscriptions dey provisioned at speed, with small governance and even less clarity on wetin business value dem suppose deliver. And now, all this dey show up as bill shock. “Unvalidated AI adoption don enter the mix,” says Engelbrecht. “AI tools dey deployed before business value dey proven.”
Na familiar pattern. If anything, na the same dynamic wey give rise to cloud sprawl for first place: fast adoption, slow governance, deferred accountability. AI and that na really scary thing,” says Vieira. “You need to sabi wetin business objective you dey try meet and whether AI fit actually get you there.”
Rushing workloads into public cloud AI services before model dey validated risk both cost overruns and data exposure. His advice na to start with on-prem proof of concept, validate the model and then scale into cloud when the use case dey proven and the data governance dey in place.
Cloud value for Africa dey shaped by geopolitics, regulation and foreign exchange – and 89% of organisations don already adjust their cloud approach in response. “For this environment, realising cloud value depend on how effectively organisations respond to shifting geopolitical, regulatory and economic pressures,” says Mark Allderman, Africa cloud and digital leader at PwC South Africa.
“Cloud strategies must dey flexible enough to adapt to new compliance requirements, resilient to currency and cost fluctuations and aligned to national sovereignty expectations.” The PoPI Act for South Africa, Nigeria NDPR and Kenya Data Protection Act all add compliance complexity to wetin already be technically demanding environment, and with 45% of organisations say geopolitical shifts dey directly affecting their cloud strategies, the pressure to get this right dey only grow.
None of this make cost optimisation easier. “For 20 years, that conversation no change,” says Vieira. “The only way to mitigate am na to invest in tooling wey help people run these services without steep learning curve.” The right tools create visibility and control, but without commitment to continuous review, even the best-optimised environment drift back towards waste.
This na why cloud cost optimisation don outgrow the idea say e just be technical problem with technical fix. The real issue na lifecycle management. “For cloud, for theory, get infinite amount of resources and so people just tend to deploy and say, ‘No worry, we go sort am next week’,” says Vieira. “That cost shock come about very quickly. Cloud definitely no be deploy-and-forget concept.”
Do you have a news tip for NNN? Please email us at editor @ nnn.ng
