AWS vs. Azure vs. GCP: what actually matters for a mid-size company

The "which cloud is best" debate misses the point — for most mid-size companies, the right platform is determined by three practical factors that have nothing to do with raw feature comparisons.

The three questions that actually decide it

First: what's your existing tooling ecosystem? A company already deep in Microsoft 365 and Active Directory has a real integration advantage on Azure that's expensive to replicate elsewhere. A company with a data science team already fluent in BigQuery has a similar advantage on GCP. This existing-tooling gravity matters more than any abstract feature comparison, because migration and retraining costs are real and often underestimated.

Second: what's your actual workload shape? AWS has the deepest and most mature service catalog overall, which matters if your needs are broad and varied. GCP tends to have an edge specifically for data analytics and ML workloads — BigQuery and Vertex AI are genuinely strong. Azure has the edge for anything that needs to integrate tightly with an existing Microsoft enterprise stack.

Third: where's your team's actual expertise? The "best" platform run by a team with no experience on it will underperform a "second-best" platform run by people who know it deeply. We weight this factor heavily in recommendations — a migration to a theoretically superior platform that your team doesn't know yet has a real productivity cost in the first 6-12 months that needs to be honestly counted.

Where the pricing comparison actually matters (and where it's a red herring)

Raw compute pricing across the three major clouds is close enough that it's rarely the deciding factor on its own — the differences that matter are in reserved instance/committed use discount structures, egress costs (which vary meaningfully and can matter a lot for data-heavy workloads), and managed service pricing for the specific services you'll actually use. We build a workload-specific cost model rather than relying on generic pricing comparisons, because generic list-price comparisons rarely reflect what a specific company will actually pay after discounts and workload-specific service costs.

A concrete example

A manufacturing client evaluating a cloud migration had an existing Microsoft 365 environment, a moderate compute workload, and a small internal IT team with no deep cloud specialization on any platform. We recommended Azure — not because it was objectively "best," but because the Active Directory integration eliminated a whole category of identity management work, the team's existing Windows Server familiarity transferred directly, and the migration timeline was roughly 30% shorter than our estimate for an equivalent AWS migration, purely due to reduced learning curve and integration work.

Where "multi-cloud" gets recommended when it shouldn't

Multi-cloud strategies are sometimes genuinely justified — regulatory requirements, specific best-in-class service needs, avoiding vendor lock-in for a critical system. But for most mid-size companies, running production workloads split across two or three clouds multiplies operational complexity (separate monitoring, separate IAM, separate cost management) without a proportional benefit. We push back on multi-cloud by default and ask clients to articulate the specific requirement driving it before we build for it.

How Ndakum approaches it

We're genuinely platform-agnostic in our Cloud Engineering work — the recommendation comes from your existing tooling, workload shape, and team expertise, not from which platform we're most comfortable selling.