By Harald Kriener, Head of Global Customer Success Management at DE-CIX.
Cloud bills have a habit of hiding their biggest surprises in the smallest line items. A business can spend months optimizing compute usage and tightening SaaS governance, only to discover that a growing share of its cloud bill is being driven elsewhere, by things like data egress charges, inter-region traffic fees, cross-cloud transfers, and inefficient routing paths moving data between platforms, users, and applications. Every cloud migration, AI workflow, backup process, API call, and multi-cloud application generates traffic flowing between platforms, regions, users, and services, and at an enterprise scale, those connections start to matter just as much as the workloads being run. That’s why decisions about connectivity are now just as relevant to FinOps as they are to IT.
FinOps began largely as a discipline focused on cloud consumption, but it’s now concerned with infrastructure strategy as a whole. According to the latest edition of the FinOps Foundation’s annual survey, more than 90% of FinOps teams now manage SaaS and AI spend, up from 63% in the previous year, while many are also becoming more directly involved in cloud and data center architecture decisions. Businesses can no longer afford to separate cloud economics from the networks that connect everything together. As data volumes explode and workloads become increasingly distributed, inefficient routing paths, public Internet dependency, and rising egress fees are turning connectivity into one of the most overlooked drivers of operational cost.
Why connectivity is a FinOps problem
The economics of enterprise IT have changed dramatically in the cloud era. Where infrastructure used to be purchased once and left to depreciate over time, businesses now consume infrastructure continuously, usually across multiple cloud providers, regions, and platforms. That’s unlocked some huge advantages in terms of innovation and scalability, particularly as AI workloads grow and SaaS ecosystems expand, but it’s also introduced a level of financial unpredictability that many organizations are now struggling to manage. Because in highly distributed environments, costs don’t just come from running workloads, but the constant flow of data between them.
That’s how egress fees have become such a growing concern for FinOps teams. Every time data needs to leave a cloud environment and travel elsewhere – which is often – charges begin to accumulate. That could be data traveling to another cloud, an on-premise environment, or another branch or remote end user. Amazon Web Services (AWS) egress fees, for example, can range between $0.05 and $0.10 per gigabyte. That means an organization transferring 10 terabytes of data per day could potentially face monthly egress costs reaching tens of thousands of dollars before storage, compute, or AI processing costs are even factored in. Gartner also estimates that egress fees account for roughly 10–15% of total cloud spending on average.
That challenge is even more pronounced in multi-cloud and AI-heavy environments. AI models, analytics platforms, SaaS applications, and distributed workloads constantly exchange large volumes of data across environments. If those flows rely heavily on the public Internet, businesses inherit all the inefficiencies that come with it – unpredictable routing paths, congestion, latency, limited visibility, and additional security exposure. In other words, organizations aren’t just paying to move data; They’re paying more to move it inefficiently.
Where Internet Exchanges and software-defined routing fit in
As FinOps teams dig deeper into the causes of unpredictable cloud spending, many are arriving at the conclusion that architecture matters just as much as consumption. It’s one thing to ask if workloads are running efficiently, but it’s another thing entirely to understand how those workloads are communicating, where traffic is flowing, and whether data is taking the most efficient path possible between clouds, users, and applications. That’s driving growing interest in private interconnection and Internet Exchanges (IXs) as a way to bring greater control to cloud connectivity. Instead of routing traffic across the public Internet, IXs allow enterprises, cloud providers, and networks to exchange data through direct network interconnection, avoiding many of the bottlenecks and inefficiencies associated with public Internet routing while improving latency, resilience, and performance And because they allow the company network to connect directly with the cloud provider’s own private network, it also significantly reduces egress costs.
Software-defined routing services add another layer of intelligence. Rather than backhauling traffic through local infrastructure or manually managing fragmented cloud connections, businesses can dynamically control how traffic moves between environments and place workloads where it makes the most operational and financial sense. Virtual routing platforms simplify connectivity between multiple clouds and on-premises infrastructure without requiring organizations to continuously deploy and maintain additional hardware, while also giving teams greater visibility into where traffic flows and where costs are accumulating. In AI and multi-cloud environments, where massive volumes of data move constantly, that visibility and control can have a direct impact on both performance and cloud spend.
Smart architecture is becoming the real FinOps advantage
One of the biggest misconceptions around FinOps is that it’s purely a tooling exercise. Dashboards, tagging systems, billing analytics, and usage reports all play an important role, but they only address part of the challenge. The larger opportunity sits deeper within the architecture itself. Businesses can optimize workloads extensively, yet still operate inefficiently if the underlying connectivity model forces data to travel unnecessarily long, congested, or expensive routes, as described above. As cloud environments become more distributed, infrastructure needs to be designed with data movement in mind from the very beginning.
That’s especially important in AI environments where workloads generate enormous volumes of traffic between clouds, data centers, users, inference models, and training clusters. In these situations, decisions around connectivity can directly influence scalability, responsiveness, and operational cost. Software-defined interconnection allows businesses to centralize and virtualize connectivity, automate provisioning, and shift workloads more dynamically between environments based on performance, location, or cost requirements. So instead of treating connectivity as a static utility sitting beneath the business, it needs to be a programmable operational layer that can actively support financial efficiency, resilience, and long-term scalability.
Cloud conversations tend to focus on what businesses are running – AI models, applications, storage, platforms. Increasingly, though, the more important question is how all of those things communicate with each other once they’re live. Data is now in constant motion between clouds, users, edge environments, and AI systems, and every unnecessary hop introduces more cost, more latency, and less control. That’s why connectivity is becoming one of the most important FinOps discussions happening inside enterprises today.