How businesses are cutting cloud spend to fund AI investment

Kevin Naicker, Head of Cloud at DVT.

---

The 2026 Flexera State of the Cloud Report estimates that 29% of public cloud spend is wasted, while 81% of respondents are already using generative AI. For technology leaders, the pressure is no longer simply to curb a growing cloud bill. It is to recover wasted spend without slowing the investment needed for AI, data and modernisation.

Cloud was expected to lower infrastructure costs and increase agility. Instead, many organisations are facing escalating monthly bills, limited visibility into what is driving them and growing pressure to justify further technology investment.

As a contributor to the FinOps Foundation, the global community for cloud financial management, DVT works with organisations to improve visibility, accountability and efficiency across cloud environments. DVT combines cloud architecture and engineering expertise with FinOps practices to identify immediate savings and improve the long-term efficiency of cloud platforms.

Naicker says the problem often stems from businesses continuing to manage cloud environments as if they were traditional data centres.

"Many early cloud migrations simply moved applications into the cloud without redesigning them. Resources remain sized for peak demand, workloads run when they are not needed and organisations miss many of the efficiencies cloud platforms can provide," says Naicker.

For South African organisations, the challenge is often intensified by foreign currency exposure, ageing applications and legacy platforms that consume resources without delivering proportional business value.

The first step is to establish where cloud spend is going, who owns it and which resources are delivering value.

Reviews of cloud environments often uncover the same problems: over-provisioned resources, under-utilised infrastructure, unused licences and development or test environments that remain active without clear ownership. In some cases, cloud accounts have changed hands over the years, leaving no one accountable for the associated costs.

"We typically focus on finding the first 20% of savings. That process also gives us a much better understanding of the customer's applications, workloads and architecture," says Naicker.

Those savings can deliver an immediate financial benefit, but the review often reveals broader opportunities to modernise applications, automate processes and improve cloud efficiency.

Cloud cost control does not sit solely with the finance department. Engineering and architectural decisions determine much of the spend, while finance needs to understand whether that spend supports the organisation's priorities.

"Finance teams talk in rands and dollars. Engineering teams talk in compute, storage and clusters. The FinOps Framework provides a common language that helps both sides understand the trade-offs and make better decisions," says Naicker.

DVT helps bring these teams together by linking technical usage and ownership to financial information and business priorities. This makes it easier to decide what to resize, retire, automate or redesign.

The result is clearer accountability and a better basis for deciding where cloud investment should increase, where it should be reduced and how its return should be measured.

The cost challenge is extending into AI. Flexera's 2026 State of ITAM Report found that 59% of organisations reported an increase in wasted AI software spend, a higher proportion than for public cloud software, SaaS or infrastructure as a service and platform as a service.

Reducing avoidable cloud costs gives organisations room to fund AI, advanced analytics and modern data platforms without simply adding to the technology budget. The aim is not indiscriminate cost-cutting, but moving money from resources that deliver little value to initiatives that can improve growth, service or productivity.

"Legacy systems often continue to support critical business processes. But organisations also need the flexibility to invest in emerging technologies that create new opportunities for growth," says Naicker.

As AI adoption accelerates, organisations must account for new cost drivers, including GPU infrastructure, model usage and token consumption. These costs can rise quickly when ownership, usage and business outcomes are not visible.

"We are seeing businesses spend significantly more on AI resources than in previous generations of cloud computing. Understanding and managing those costs will become just as important as managing traditional cloud infrastructure," concludes Naicker.