Companies Spend Three Times More Than They Should on Cloud Costs
CAST AI, an AI-driven cloud optimization company, analyzed infrastructure utilization reports for more than 400 organizations to quantify how much they overspend on cloud costs. These organizations included large enterprises running applications across tens of thousands of CPUs down to small companies running their applications on just a few nodes.
The company used its free cluster analysis tool that provides detailed insight into how a company’s cloud resources are provisioned, as well as specific opportunities for optimization and cost savings.
According to Laurent Gil, co-founder and chief product officer at CAST AI, the company’s advanced AI engine “provides full visibility into how much you’re currently paying for cloud resources as well as how much you would save if those resources were optimized.”
Among CAST AI’s key findings, both in terms of overprovisioning and how much companies overspend as a direct result, include:
- On average, organizations spend 3x more than they should on cloud costs
- The main driver for overspending is the over-provisioning of expensive resources, resulting in significant cost with no material benefit
- Almost two-thirds of money wasted is the result of CPUs and memory that are provisioned but not utilized, combined with the selection of cost-inefficient VMs with expensive CPUs and a too-large memory ratio
- The remaining waste comes from under leveraging the use of spot instances for containers that are qualified to be spot friendly
Provisioning remains a significant challenge for organizations of all sizes. Cloud providers typically have more than 600 different instance types to choose from, so even the most experienced and technically adept DevOps and SRE professionals benefit considerably from automation; it is difficult and time-consuming to select the right types and quantity of VMs while making sure that the infrastructure is continuously rightsized.
With new advances in AI it is now easy to make this process both instant and automated, so that requested, provisioned, and utilized CPUs are 100 percent in sync and remain in sync over time.