For cloud providers, business is booming. In just the last few weeks, major providers Google and Amazon Web Services have struck deals with the CME Group and Nasdaq, respectively, and a growing patchwork of private, public, and hybrid cloud projects and migrations make up a huge chunk of tech spending in financial services and beyond—globally, $396 billion was spent on public cloud services in 2021, according to Gartner, which projects that figure to reach $482 billion next year.

While the builders and suppliers are making money, some end-users, enamored with cloud’s potential for efficiencies and cost savings, are actually overprovisioning and overspending.

Speaking on a panel last month at the Open Source Strategy Forum in New York, Jennifer Hays, head of engineering efficiency and assurance at Fidelity Investments and chair of the FinOps Foundation governance board, said 35% to 40% of AWS customers overprovisioned or overspent on cloud. (AWS declined to comment on the accuracy of those figures.) Sources in the cloud management and optimization space tell WatersTechnology that cloud overprovisioning and overspending is a pervasive problem that extends to more than just AWS customers.

Bill Murphy, managing partner at Cresting Wave and former CTO at Blackstone, says perceived discounts offered by cloud providers can create overhead issues that lead to overspending. Cloud’s standard model of elasticity, one of its main selling points, promises that customers only pay for what they use when they use it, but it also means there’s rarely a fixed price, making budgeting difficult.

Additionally, Murphy says cloud providers like AWS offer additional price concessions for years-long commitments to utilizing the companies’ infrastructures. “Everybody likes a deal but then people over-buy, and then they’re sitting there with all these unused database instances,” Murphy says. On the other hand, some under-provision.

VMware’s CloudHealth aims to help users gain visibility into their cloud costs and usage metrics, and better govern their related spending and budgeting needs. John McLoughlin, the vendor’s senior director of product management, says the proliferation of cloud technology has allowed for greater implementation of the DevOps philosophy, which involves combining software development and IT operations to shorten development time and provide both continuous delivery and high-quality software.

This entails a balancing act of safety, visibility, and governance versus speed of innovation and getting to market. Sometimes cost efficiency draws the short straw. “What that meant was that the governance and regulations we’re used to having in an older IT world loosened. You can be more innovative; you can drive change more quickly. But that normally raises complexity,” he says.

The shift from on-premises datacenters with limited capacities to elastic public clouds has given vendors freedom to build more applications and solutions, but it also means users may be unsure how much storage and how many tools to utilize. “The complexity gap between management and keeping up with the rising tide of different technologies has driven an efficiency problem where people are nervous about using the right amount of infrastructure to solve a problem,” McLoughlin says.

End-users first grapple with this gap when migrating from a datacenter to the cloud. The process can be as simple as a lift-and-shift—which causes the least disruption but isn’t suitable for legacy-heavy technology or custom-built applications. Or it can be as complex as full-scale re-architecting—which involves breaking existing applications down to their studs and building them back up with cloud-native capabilities, and which is usually the most expensive. Software-as-a-service (SaaS) migration remains a popular, middle-of-the-road choice and involves outsourcing certain applications to cloud services companies—such as Salesforce-hosted CRM—that are better equipped to host and manage them.

End-users often use combinations of these strategies, depending on the systems and applications they’re looking to move. “The myth of the cloud is that it’s inherently cheaper. I think a lot of folks are surprised when they have the climbing cloud bill and their consumption continues to increase,” says Marie Burke, director of product marketing for CloudHealth.

In order to save, spend somewhere else?

There’s a growing market for cloud management vendors, such as CloudHealth, which aren’t cloud providers themselves, but which sell software and services meant to consolidate end-users’ cloud costs and, hopefully, reduce them. Others include Cloudability by Apptio, Densify, and Zesty. AWS, Google Cloud, and Microsoft Azure have similar offerings of their own. The potential problem with enlisting the services of a new vendor to help save costs with an existing one is that any realized savings may still not end up in the end-user’s pocket.

Cloudability uses proprietary machine-learning models, trained on “trillions of rows of data records across cloud providers” to assess consumers’ spending and usage, and identify areas that could potentially be cut, says Eugene Khvostov, the company’s vice president of product and engineering.

“As the data gets ingested—as it goes through the tollbooth, if you will—it identifies the best recommendations using a variety of data points such as how long an elastic block storage volume has been idle or orphaned, or what your total footprint of compute instances looks like, and where you might find efficiencies and recommendations,” Khvostov says.

Obtaining a bird’s eye view of one’s cloud infrastructure, particularly for institutions as large as global investment banks and exchanges, is complex and requires examining hundreds to thousands of different inputs, says Chuck Tatham, chief revenue officer at Densify, a Canadian predictive analytics company that also seeks to optimize enterprises’ cloud and container environments and reduce waste.

Broadly speaking, those inputs include detailed behaviors of applications operating in the cloud, utilization patterns, and the infrastructure they run on. To help customers reduce costs, the company’s software analyzes and compares how busy their cloud applications are each day and night.

Zesty, an Israel-based cloud management platform, uses automation to help customers manage their discounts and match utilization to as close as 100% so that end-users don’t end up paying for something that they’re otherwise not using.

As institutions continue to move to the cloud, continuous adherence to cost savings methods will be optimal. “Keep looking at the other things that are coming up as new workloads are identified to see if there are advanced ways of driving down costs,” McLoughlin says.