Since online musical instrument and professional audio retailer Sweetwater formalized its cloud strategy around 2016, Johnson has tried to place workloads where they will provide the most value to external and internal customers.
“Everything that has customer contact needs to happen as close to the end user as possible,” he explains. The same logic applies to internal workloads: they belong close to those who use them. AI would not be possible without the cloud. The computing capacity needed for AI already existed – the cloud had it and could provide it. In many ways, AI is perhaps the cloud’s greatest gift to the industry. Both made each other possible,” adds the manager.
Cloud cost management is becoming increasingly complex
Johnson’s biggest problem is consistently controlling costs. “It used to be relatively simple: computing power, storage, data traffic. Now it’s a puzzle. Reserved instances, savings plans, spot pricing, cost per request, data transfer fees between regions – it all adds up quickly. And it’s really hard to predict what your bill will look like until it arrives.”
That’s why FinOps has now become its own discipline.
Basu also considers FinOps to be indispensable. “AI inference costs, data transfers, and growing storage volumes can cause monthly cost fluctuations that surprise even experienced teams,” he says. Cost management is now an ongoing operational discipline and no longer just an occasional review.”
The danger of vendor lock-in is also a constant concern for Johnson. “The deeper you integrate a provider’s native services, the harder it is to switch. That’s not necessarily a bad thing, but it’s a trade-off. I think of it like technical debt. You gain speed today and pay interest later when you want to change direction.”
At the same time, Johnson recognizes that cloud providers are companies that pay attention to revenue and margins and regularly change the rules for discounts and cost models.
Financial efficiency doesn’t happen by itself, he emphasizes: “It requires teams, processes and real investment in FinOps. The tools are there. The real challenge is using them correctly.”
In most companies, financial knowledge lies in accounting and technical knowledge lies in IT, explains Johnson. Bringing both groups together when it comes to cloud cost management is a comparatively new challenge.
“Ten years ago, the model was simple: you requested a capital budget, accounting approved it, hardware was ordered, and IT installed and optimized everything. Done,” says Johnson. “Today it’s a daily task. New services are activated, contracts change, pricing structures are adjusted. The finance department understands the money flows, but not the technology. IT understands the technology, but not the financial levers.”
Every major cloud provider has the appropriate tools, explains Johnson. “AWS Cost Explorer, Azure Cost Management, Google Cloud’s billing dashboards. The data is all there. But many companies don’t use this data. Then the bill comes and everyone is surprised. The tools announced the development. People just didn’t listen.”
Data regulation brings additional challenges
The University of Health Sciences and Pharmacy has students from all over the world. According to Lewis, cloud management has become significantly more difficult due to increasing data protection and regulatory requirements worldwide.
“We need to understand in which cloud region metadata is stored and where it resides permanently. If this data goes into an internally trained model, can we guarantee that it stays in the EU? And if someone requests that their data be deleted, can we reliably identify all storage locations?”
Basu adds that data sovereignty has become another key architectural issue.
“It affects where workloads can run, how data moves between regions, and what can happen to certain data sets. You can’t assume that the default configuration of a hyperscaler will automatically meet all regulatory requirements.”
Private Cloud, Public Cloud oder On-Premises?
AI has sparked a reassessment of workload placement. However, Johnson believes that the question of private cloud, public cloud or on-premises changes more often than many people think.
“I think most of our workloads are in the right place right now. But we only got there because we were willing to challenge standard assumptions.”
At Sweetwater there is no rule that new workloads automatically belong in the cloud. A workload can start in the cloud and later return to the company’s own data center if economics require it.
“The real discipline is to constantly check this, recognize turning points and continuously adjust the size of the environment. The right tool at the right time – that is the only principle that lasts in the long term.”
Basu does not plan to switch to a private cloud because the economics and operational costs do not justify it for his company.
“The right question is: What data residency, latency and control model is appropriate for each workload? That’s a question of data classification, not cloud deployment strategy.”
Lewis, whose university runs around 95 percent of its infrastructure in the cloud, sees few viable alternatives given the demands of AI and modern hardware.
“If you want to train large-scale data lakes and make informed business decisions using machine learning and the intelligence behind it, it is almost no longer practical to rely on your own servers,” says Lewis.
CIOs would have to ask themselves whether they even have the necessary know-how to operate such an infrastructure themselves. Ultimately, you have to make the best of the available options.
Focus on the basics
Companies may be chasing the latest trend—currently agentic AI—but IT leaders should focus on their infrastructure, management and platform planning, Linthicum emphasizes.
“This may be less exciting, but if you want to use AI in the company, you have to solve these problems first.”
Johnson warns that the cloud enables a variety of architectural patterns – and without guardrails, this freedom could become a problem. “Create the guidance document before you need it – not when there are already five teams doing it in five different ways.”
IT managers should also ensure clean tagging and transparency in costs at an early stage. This has to be the first step and shouldn’t be seen as a clean-up project, warns Johnson: “If you can’t clearly understand your spending from day one, you’re already behind.”
Another important step is building an audit program with clear controls over who can make changes to production systems: “The impact of a bad change in the cloud is greater and occurs faster than most people expect – until they experience it themselves.”
The question of skills
The skills required to operate modern cloud environments are evolving faster than most in-house teams can realistically keep up, according to Basu. That’s why International Seaways relies on specialized managed service providers (MSPs) instead of building up in-depth internal expert knowledge in each area.
“This gives us access to current skills without having to constantly retrain or rebuild teams as technology changes. The decisions that protected us in 2020 were made years ago – long before anyone realized their importance. Infrastructure strategy always means preparing for a future that is not yet fully visible.”
It is crucial to make such decisions consciously and on the basis of clear considerations, instead of having to react later under pressure.
Build adaptable organizations
Edge computing adds another layer of complexity, according to Johnson. Those who find the perfect architecture will not be successful. “They are the ones who build organizations that can adapt quickly if the right answer changes tomorrow,” he explains.
The real competitive advantage lies not in the cloud platform chosen, but in how quickly a team can learn and act, says Johnson.
When asked how to make cloud environments less complicated, CIOs offered the following recommendations:
- Treat your cloud architecture like a product, not a project. It requires ongoing accountability and development, not just a one-time implementation.
- Review your decisions regularly. “The right decision in the first year is often no longer the right one in the third. Plan fixed review points,” explains Johnson.
- Assign each workload to a cost center. Basu implemented this; IT continuously checks utilization and sizing instead of waiting for periodic audits.
- Data classification determines regional placement. Before a workload goes into production, legal and compliance departments should be involved from the start.
- Develop cloud-ready solutions. This can be cheaper and less risky than the unchanged migration of heavily customized legacy systems.
- Consolidation is a strategic decision, not a retreat. Fewer, well-governed platforms perform consistently better than a fragmented, multi-cloud landscape.
- Don’t underestimate the governance gap created by AI. Create an AI governance structure now, before technical and organizational debt accumulates.
- Cloud strategy is not just an IT architecture decision. The pandemic has shown that it is above all a decision to ensure corporate resilience. “The organizations that make difficult decisions before a crisis are the ones that emerge unscathed.”
- Link cloud decisions to measurable business outcomes. WHSmith’s Bellendir reports that his company is investing heavily in integration architectures, cybersecurity controls, observability and data governance to better support hybrid environments.
- Place greater emphasis on cloud cost control and operational discipline. This improves transparency about cloud usage and ensures that scaling AI, analytics and digitalization initiatives remains financially viable in the long term.
Although no one can predict whether cloud environments will become less complex in the future, companies will continue to use them.
“The cloud is no longer the future of IT – it is the present. The discussion has shifted from ‘Should we do this?’ to ‘How do we get better at this?’ postponed. And that’s exactly what I spend most of my time doing these days,” Johnson states. (mb)
This article is based on a post from CIO.com
Im Fokus: Cloud Computing
Find out in our free PDF what IT decision-makers need to know about the cloud in 2026.
Download now!
CineVI – shutterstock.com
