← Back to Blog Software Development

Why We Don't Recommend Big Cloud Platforms for Most Small Businesses

We get asked regularly whether a client should build on AWS or Azure. Our honest answer, for most small and mid-size businesses, is no. Not because those platforms are bad. Because they're built for a scale and complexity that most businesses will never need - and the cost of that complexity falls entirely on you.

What Big Cloud Actually Looks Like in Practice

AWS alone has over 200 services. A straightforward web application on AWS might touch EC2, RDS, S3, CloudFront, Route 53, IAM, VPC, Security Groups, ALB, ACM, and CloudWatch before you've done anything unusual. Each of those services has its own pricing model, its own configuration surface, and its own failure modes.

We recently reviewed the infrastructure for a small SaaS company - 12 employees, 800 customers - that had been built on AWS by a contractor. Their monthly bill was $3,400. When we mapped what they were actually running, the workload could have been served for under $80 a month on simpler infrastructure. The difference was architectural complexity nobody had audited since the original build.

The Hidden Cost: People

The real cost of big cloud isn't the bill. It's the people you need to manage it.

A properly maintained AWS environment requires someone who understands IAM permission boundaries, VPC subnet routing, auto-scaling group lifecycle hooks, and certificate renewal automation. That's a senior DevOps engineer. In 2026, that person costs $130,000 to $180,000 a year - if you can find and keep one.

Most small businesses end up in one of two situations:

• They hire a contractor to build it, it works for a year, then nobody understands it well enough to change it safely
• They hire someone junior who learns on the job, introduces misconfigurations, and the security debt compounds quietly

Neither outcome is what they were sold when they chose AWS because "it scales."

Deployment Models That Will Consume You

Modern cloud deployments have fractured into staggering complexity. You're now expected to choose between:

• EC2 instances (virtual machines you manage yourself)
• ECS with EC2 launch type (containers on VMs you manage)
• ECS with Fargate (managed containers, but different networking rules)
• EKS (Kubernetes, which is an entire career path by itself)
• Lambda (serverless, but cold starts, timeout limits, and VPC configuration quirks)
• App Runner (managed, but limited control)
• Lightsail (simplified, but you're admitted you don't need AWS)

Each of these has different networking models, different IAM permission structures, different logging configurations, and different pricing. Choosing wrong at the start is expensive to undo. Choosing right requires expertise most small teams don't have in-house.

Multiple Regions, Multiple Availability Zones

The pitch for multi-region, multi-AZ deployments is reliability. The reality for small businesses is operational complexity that rarely pays off. Running active-active across two regions means your database replication strategy, your session management, your cache invalidation, and your deployment process all need to handle split-brain scenarios. For a business whose actual uptime requirement is "don't go down during business hours," this is massive overengineering.

We've seen small businesses spend 40% of their engineering time on infrastructure concerns that have zero impact on their customers.

What We Recommend Instead

Simple, well-understood infrastructure you can reason about and maintain without a specialist on retainer:

• A single well-configured VPS (Hetzner, DigitalOcean, Linode) handles most workloads up to thousands of concurrent users for $20 to $60 a month
• Managed databases (PlanetScale, Supabase, Railway) so you're not hand-managing Postgres backups and failover
• A CDN in front (Cloudflare free tier) for static assets, DDoS protection, and SSL
• Simple object storage (Cloudflare R2, Backblaze B2) for files - a fraction of S3 pricing with no egress fees

This stack is boring. It's maintainable by a single developer. It costs $50 to $150 a month. It handles more traffic than most small businesses will ever generate. And when something breaks, you can understand why without a cloud certification.

When Big Cloud Makes Sense

We're not anti-cloud. AWS, Azure, and GCP are genuinely extraordinary for the right use cases:

• You have compliance requirements that mandate specific cloud certifications (FedRAMP, HIPAA BAA, etc.)
• You have a dedicated DevOps engineer or platform team
• Your workload genuinely requires elastic scaling at unpredictable volumes
• You're integrating deeply with other enterprise systems that already live there

If none of those apply, you're paying a complexity tax for infrastructure built for Netflix. You are not Netflix. Build for where you are now - not for a hypothetical scale you may never reach.

When you're ready to evaluate your infrastructure honestly, let's talk. We'll tell you what you actually need.