"Should we go AWS or Azure?" is the question we get most often after "build or buy?" — and the honest answer is: it depends on three things, none of which are about the cloud platforms themselves. It depends on your existing licensing, your team's skills, and your specific workload mix.
This is the framework we use with customers, written by someone who's spent the last six years deploying both at production scale across Kenya, Uganda, and Tanzania.
TL;DR
- Pick AWS if you have a Linux-heavy stack, want maximum service depth, or are starting greenfield with no existing Microsoft licensing.
- Pick Azure if you're already on Microsoft 365 Enterprise, run Windows Server / SQL Server / Active Directory, or operate primarily in regulated sectors where Microsoft's compliance posture matters.
- Don't pick GCP as a primary cloud unless you're heavy on data & AI workloads. It's a fine third option, not a default.
- Don't go multi-cloud unless you have a specific regulatory reason. The complexity cost almost always exceeds the resilience benefit.
The regional reality
Both AWS and Azure now have an Africa region. AWS has af-south-1 (Cape Town) since 2020; Azure has Azure South Africa (Johannesburg) since 2019. Both deliver acceptable latency to Kenyan users — typically 50-80ms round trip from Nairobi, vs. 150-200ms from Europe.
For most workloads, the African region is the right default. The exceptions are: workloads that need a service not yet available in the African region (still some — Bedrock, certain ML services, some governance services land in af-south-1 a quarter or two after us-east-1), and workloads where regulator data-residency requirements specifically allow Europe but not South Africa.
When AWS wins
- Service depth. AWS has more services, more variants per service, and a longer track record of running them at scale. If you need something specific (graph database, FPGA compute, satellite ground station), it's almost certainly on AWS first.
- Open-source & Linux. Most open-source ecosystems target AWS first — Kubernetes operators, observability tools, data infrastructure. Less friction.
- Mature multi-account isolation. AWS Organizations and the account-per-environment pattern is a stronger isolation primitive than Azure's resource-group / subscription model.
- Deeper FinOps tooling. Cost Explorer, Compute Optimizer, Savings Plans — AWS's cost-management tooling is genuinely better. Azure is catching up.
When Azure wins
- Microsoft licensing leverage. If you're already paying for Microsoft 365 Enterprise, your Windows Server licences, and SQL Server, Azure's hybrid-benefit pricing makes the same workload meaningfully cheaper than on AWS.
- Active Directory integration. Entra ID (Azure AD) is the lowest-friction option for organisations already running on-prem AD. SSO into your business apps comes for free.
- Regulated industries. Microsoft's investment in regulator-specific compliance certifications (especially for financial services, healthcare, and government) is materially deeper. If your auditor asks for it, Microsoft probably has it.
- Data & analytics on Microsoft stack. If your reporting is on Power BI, your warehouse is Synapse, and your transactional data is in SQL Server — staying on Azure removes a lot of friction.
The honest test is: where would your first hire from the local market be most productive on day one? In Nairobi, that's increasingly AWS for greenfield builds and Azure for organisations with significant Microsoft footprint. Both options are credible for everything else.— Madhu Gummadidari
Multi-cloud — usually no
Multi-cloud sounds resilient. Operationally, it's twice the complexity for marginal real-world benefit. Most "multi-cloud" architectures we audit are actually one cloud running production with a backup CSP that nobody's tested in years.
Genuine multi-cloud is the right answer in two specific situations: regulators that explicitly mandate it, or workloads where the second cloud offers a service the primary one doesn't. Outside those, pick a primary, do it well, and use the saved energy for the work your customers actually feel.
A decision framework
Run these questions, in this order:
- Are you already on Microsoft 365 Enterprise / Windows Server / SQL Server? Default to Azure.
- Are you greenfield, Linux-heavy, container-native? Default to AWS.
- Are you data & AI heavy with Google Workspace? Consider GCP.
- Does the answer to (1) and (2) cancel out? Pick the one your senior engineering hire knows best.
The wrong answer to "which cloud" is rare. The wrong answer to "did we operate it well" is common. Pick a default, hire well, document deeply.