Cloud Cost Optimization: AWS vs Azure vs GCP in 2019

By mid-2019, global cloud infrastructure spending had surpassed $100 billion annually, with AWS, Azure, and Google Cloud Platform (GCP) commanding the vast majority of the market. But alongside this growth, a troubling pattern had emerged: organisations were wasting an estimated 35% of their cloud spend on idle resources, over-provisioned instances, and suboptimal pricing models.

For mid-market businesses, this waste was particularly painful. Unlike enterprises with dedicated FinOps teams, most mid-market companies lacked the expertise to navigate the labyrinthine pricing structures of major cloud providers. A single misconfigured auto-scaling policy or forgotten development environment could add thousands of dollars per month to cloud bills without delivering any business value.

The three major cloud providers had each developed distinct pricing philosophies and optimisation tools, making direct comparison challenging but essential for any business serious about controlling its cloud costs.

AWS offered the most granular pricing with on-demand, reserved instances (1-3 year commitments for up to 72% savings), spot instances (up to 90% savings for interruptible workloads), and savings plans. The sheer number of options was both a strength and a challenge — optimising AWS costs required significant expertise. AWS also offered the most mature cost management tooling with Cost Explorer, Trusted Advisor, and detailed billing reports.

Azure provided similar reserved instance discounts and had a significant advantage for organisations already invested in the Microsoft ecosystem. Azure Hybrid Benefit allowed businesses to apply existing Windows Server and SQL Server licences to cloud instances, potentially saving up to 40% compared to on-demand pricing. For mid-market businesses running Microsoft workloads, this often made Azure the most cost-effective option before even considering compute pricing.

Google Cloud Platform differentiated with sustained use discounts — automatic discounts of up to 30% applied to instances running for more than 25% of a billing month, with no upfront commitment required. GCP also offered committed use discounts (similar to reserved instances) and preemptible VMs (similar to spot instances). GCP's per-second billing and custom machine types allowed more precise right-sizing than AWS or Azure's fixed instance families.

Regardless of which cloud provider you use, the highest-impact cost optimisation strategies in 2019 were remarkably consistent:

• Right-sizing: Analyse actual resource utilisation and resize instances to match real workload requirements. Most organisations over-provision by 40-60% because developers default to larger instance types during development and never resize for production.
• Reserved capacity planning: Commit to reserved instances or savings plans for predictable baseline workloads. The 30-72% savings from commitments should be applied to any workload that runs consistently for 12+ months.
• Automated scheduling: Shut down development, staging, and non-production environments outside business hours. A simple scheduling policy can reduce non-production costs by 65%.
• Storage lifecycle policies: Implement automatic tiering from hot to cold storage for infrequently accessed data. Moving old logs and backups to archival storage tiers can reduce storage costs by 80-95%.

Implementing these four strategies alone could reduce a typical mid-market cloud bill by 25-40% without any architectural changes or service degradation. The key was establishing a regular review cadence — monthly cost reviews with clear ownership and accountability for optimisation actions.